Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Università degli studi di Torino Facoltà di Scienze Matematiche Fisiche e Naturali DOTTORATO DI RICERCA IN FISICA XVII CICLO DEVELOPMENT AND TEST OF SEGMENTED ELECTRODE IONIZATION CHAMBERS AS BEAM MONITOR FOR HADRONTHERAPY Advisor: Prof. Cristiana Peroni Candidate: Alberto Boriano Index Index................................................................................................................................................. 1 Introduction ...................................................................................................................................... 6 1 The cure of cancer with radiation ................................................................................................ 7 1.1 The significance of radiobiology in radiotherapy ................................................................ 7 1.1.1 The role of radiotherapy in the management of cancer ............................................... 7 1.1.2 The role of radiation biology ....................................................................................... 7 1.1.3 The time-scale of effects in radiation biology ............................................................. 8 1.1.4 Response of normal and malignant tissues to radiation exposure ............................... 9 1.1.5 Response curve, dose-response curves and isoeffect relationship ............................... 9 1.2 DNA damage and cell killing............................................................................................. 11 1.2.1 Initial process of radiation damage ............................................................................ 11 1.2.2 Radiation damage to DNA ......................................................................................... 12 1.2.3 Cell death in mammalian tissues ................................................................................ 15 1.3 The oxygen effect............................................................................................................... 15 1.4 Particle beams in radiotherapy ........................................................................................... 18 1.4.1 Biological effects depend upon LET ......................................................................... 19 1.4.2 The biological basis for high-LET radiotherapy ........................................................ 21 2 Radiotherapy with hadrons ........................................................................................................ 22 2.1 Proton interactions with matter .......................................................................................... 22 2.1.1 Introduction ................................................................................................................ 22 2.1.2 Proton interactions with electrons: energy loss.......................................................... 22 2.1.3 Proton interaction with electrons: energy loss distribution ........................................ 25 2.1.4 Proton interactions with nuclei: scattering ................................................................. 29 2.1.5 Proton interactions with nuclei: nuclear reactions ..................................................... 31 2.1.6 Proton dose distribution ............................................................................................. 32 2.1.7 Fragmentation ............................................................................................................ 35 2.2 Accelerators for hadrontherapy .......................................................................................... 37 2.2.1 Introduction ................................................................................................................ 37 2.2.2 Cyclotron .................................................................................................................... 37 2.2.3 Synchrotron ................................................................................................................ 38 2.2.4 Example of dedicated designs .................................................................................... 38 2.3 Beam spreading .................................................................................................................. 40 2.3.1 Passive spreading ....................................................................................................... 40 2.3.2 Scanning ..................................................................................................................... 41 2.4 Monitoring of hadrontherapy beam ................................................................................... 43 2.4.1 Passive scanning systems ........................................................................................... 43 2.4.2 Active scanning systems ............................................................................................ 43 2.4.3 Development of segmented anode ionization chambers ............................................ 44 2.4.4 Electronic read-out: the VLSI chip ............................................................................ 46 3 The GSI test: experimental setup and results ............................................................................. 48 3.1 The GSI facility in Darmstadt ............................................................................................ 48 3.2 The experimental setup ...................................................................................................... 49 3.3 The data acquisition system ............................................................................................... 50 3.3.1 The operating system VxWorks and the integrated environment TornadoII............. 50 3.3.2 The data acquisition system: hardware ...................................................................... 51 3.3.3 The detector read-out ................................................................................................. 52 3.4 Results ................................................................................................................................ 53 3.4.1 Charge collection efficiency ...................................................................................... 53 3.4.2 Calibration .................................................................................................................. 54 1 3.4.3 Space coordinates resolution with a steady beam ...................................................... 55 3.4.4 Space coordinates and fluence measurements with a scanning beam ....................... 57 3.4.5 Performance with a Treatment Planning System (TPS) ............................................ 58 Conclusions .................................................................................................................................... 61 4 The IBA test: experimental setup and results ............................................................................ 62 4.1 The CRC facility at Louvain-la-Neuve .............................................................................. 62 4.2 The Pencil Beam Scanning (PBS) ..................................................................................... 63 4.2.1 Low Level Regulation ................................................................................................ 63 4.3 The experimental setup ...................................................................................................... 65 4.3.1 The beam line setup ................................................................................................... 65 4.3.2 The structure of the data acquisition .......................................................................... 66 4.4 Results ................................................................................................................................ 68 4.4.1 Time synchronization between PIXEL/STRIP data taking and IBA control system 68 4.4.2 Homogeneity of the dose distribution ........................................................................ 71 4.4.3 Dimension of the irradiated field ............................................................................... 74 4.4.4 Detailed spatial behavior of the scanning beam......................................................... 75 4.4.5 Scanning speed ........................................................................................................... 78 4.4.6 Beam shape ................................................................................................................ 80 Conclusions .................................................................................................................................... 82 5 The CATANA test: experimental setup and results................................................................... 83 5.1 Development of a dedicated kind of strip detector as monitor of passive scanning beam 83 5.2 The CATANA facility ....................................................................................................... 85 5.3 The experimental setup ...................................................................................................... 86 5.3.1 The beam line setup ................................................................................................... 86 5.3.2 Data acquisition structure........................................................................................... 87 5.4 Experimental Results ......................................................................................................... 88 5.4.1 Calibration .................................................................................................................. 88 5.4.2 Fluence measurements and homogeneity of dose distribution .................................. 91 5.4.3 A mathematical method for the evaluation of the beam structure ............................. 93 Conclusions .................................................................................................................................. 102 Bibliography................................................................................................................................. 103 Table of figure Figure 1.1: Time scale of the effects of radiation exposure on biological systems. ............................ 8 Figure 1.2: Four types of chart leading to the construction of an isoeffect plot. (A) Time-course of radiation damage in a normal tissue. (B) The cumulative response. (C) A dose-response relationship, constructed by measuring the response (R) for varius radiation doses (D). (D) Isoeffect plot for a fixed level of normal tissue damage. ........................................................... 10 Figure 1.3: (A) Computer-simulated tracks of 1 KeV electrons. Note the scale in relation to the 2.3 nm diameter of DNA double helix. (B) Illustrating the concept of a local multiply damaged site produced by a cluster of ionizations impinging on DNA .......................................................... 12 Figure 1.4 The structure of DNA, in which the four bases (G,C,T,A) are linked through sugar groups to the sugar-phospate backbone. .................................................................................... 13 Figure 1.5 Types of damge to DNA produced by radiation and chemical agents. ............................ 14 Figure 1.6 Survival curves for culturaed mammalian cells exposed to x-rays under oxic or hypoxic conditions, illustrated the radiation dose-modifying effect of oxygen. Note that the broken line extrapolate back to the same point on the survival axis (n=5.5). ............................................... 16 Figure 1.7 Variation of oxygen enhancement ratio (OER) with oxygen tension. The horiziontal arrows indicate the range of physiological blood oxygen tensions on the lower scale ............. 17 2 Figure 1.8 The oxygen fixation hypothesis. Free radical produced in DNA either by direct or indirect action of radiation can be rapired under hypoxia but fixed in the presence of oxygen.17 Figure 1.9 The structure of particle tracks for low-LET radiation (above) and particle (below). The cricles indicate the typical syze of mammalian cell nuclei. Note the tortuos tracks of lowenergy secondary electrons. ....................................................................................................... 18 Figure 1.10 Survival of human kidney cells exposed in vitro to radiations of different LET ........... 19 Figure 1.11 Dependece of RBE on LET and the phenomenon of overkill by very high-let radations. .................................................................................................................................................... 20 Figure 1.12: The oxygen enhancement ratio (OER) decreases with increasing LET. Closed circles refer to monoenergetic a-paritcles and deuterons; the open triangle to 250 kVp x-rays. .......... 20 Figure 1.13: response of 20 human tumour cell lines to (A) 4MVp photons, or (B) p(62.5)-Be neutrons. The vertical lines show the photon (2Gy) and the neutrons (0.68 Gy) doses that give the same mediam cell survival; the average RBE is therefore 2/0.68=2.94. ............................. 21 Figure 2.1: Range-energ y relationship according to ICRU 49 and fit to relation (2.7) for water, air and gadolinium using the parameters in table 2.2 ...................................................................... 25 Figure 2.2: The Vavilov distribution function as a function of the scaled energy loss for 200 MeV protons. ............................................................................................................................. 26 Figure 2.3: Number of electrons produced per incident proton per cm2 H2O calculated using equation (2.12) ........................................................................................................................... 28 Figure 2.4: Solid line (scale on right axis): ratio between total elastic cross section of 180 MeV protons incident on 16O and Coulomb contribution (using equation (2.17)) as a function of deflection angle. Dotted line: ratio = 1. Dashed line (scale on left axis): Coulomb contribution. .................................................................................................................................................... 30 Figure 2.5: Total nonelastic nuclear cross section for proton incident 16O. The line represent a fit to the experimental data ................................................................................................................. 31 Figure 2.6: Proton flux reduction due to inelastic nuclear reaction for a 80 and a 180 MeV beam in a water medium calculated using equation (2.21) and the cross sections in figure 2.5. ............... 32 Figure 2.7: The dose per fluence as a function of depth in water calculated with (2.28) for 80 and 180 MeV proton beams, with an initial energy spread s increasing from 0% to 1.5% in steps of 0.25%. Entrance dose per fluence is for a 180 MeV beam 5.78 MeV cm2/g and for a 80 MeV beam 9.32 MeV cm2/g. R0 is shown as dashed line. .................................................................. 34 Figure 2.8: Contour plot of dose of a 160 MeV, 2 cm radius parallel proton beam in water. The contour lines go from 1% to 99% of the maximum dose. At the zero depth and on the beam axis the dose is 21% ofht edose in the Bragg peak. ................................................................... 35 Figure 2.9: Bragg curve of 270 MeV/u carbon beam in water illustrating the effects of the beam fragmentation. The colour-code lines are calculated dose distribution: red=total dose, black=primary particles, blue=secondary particles, gree=fragemnts of the secondary particles. Circles indicate measured data. .................................................................................................. 36 Figure 2.10: IBA 235 MeV room temperature cyclotron. ................................................................. 38 Figure 2.11: Layout of the TERA synchrotron .................................................................................. 39 Figure 2.12: schematic layout of a passive beam spreading system consisting of a double scatterer, a range modulator, and a snout housing the patient-specific collimator and bolus. When used with a fixed energy beam, the first scatterer also acts as energy absorber. The range modulator is not required when full energy is achieved from the accelerator. .......................................... 40 Figure 2.13: Schematic irradiation procedure for a tumour conform treatment. The tumour is dessected in slices of equidistant particle ranges which are covered using a magnetic scanning system......................................................................................................................................... 41 Figure 2.14: Pixel (right) and raster (left) scan pattern for a rectangular area and a model of a tumour slice. Additional scan lines that shortcut the standard (rectangular) scan pattern have to be introduced. ............................................................................................................................. 42 Figure 2.15: An exploded view of the pixel chamber ........................................................................ 44 3 Figure 2.16: Logic diagram of the TERA chip .................................................................................. 47 Figure 3.1: The GSI accelerator facility............................................................................................. 48 Figure 3.2: Sketc of the beam line set-up........................................................................................... 49 Figure 3.3: A sketch of the data acquisition....................................................................................... 51 Figure 3.4: (a) Chamer response and (b) the relative deviation from linearity as a function of the beam intensity ............................................................................................................................ 53 Figure 3.5: (a) Lego plot of the raw counts as a function of the pixel position. (b) Normalized distribution of the gains.............................................................................................................. 54 Figure 3.6:(a) Lego plot of the corrected counts as a function of the pixel position. (b) Distribution of the gains. ................................................................................................................................ 55 Figure 3.7: Measured position resolution as a function of the integrated number of ions: (a) x coordinate; (b) y coordinate. ...................................................................................................... 56 Figure 3.8: Beam position along x (a) and y (b) as measured in time slices of 100 ms during the spill. ............................................................................................................................................ 56 Figure 3.9: Corrected second moment vs. beam width (mm). ........................................................... 57 Figure 3.10: (a) Pictorial map of the scanning points; (b) normalized ratio between the number of ions as measured by the pixel chamber and GSI system; (c) x and (d) y resolution. ................. 58 Figure 3.11: X-Y scatter plot of four slice at different depths. ........................................................... 59 Figure 3.12: Coordinate measurements for a small volume tumour treatment: resolution along x (a) and y (b). .................................................................................................................................... 60 Figure 3.13: (a) Pixel chamber vs. GSI flux measurement and (b) the normalized ratio distribution. .................................................................................................................................................... 60 Figure 4.1: Typical ISEU response observed durign the experimentation at the CRC and the error between the expected reference and the measured beam current response. .............................. 64 Figure 4.2: Sketc of the beam line setup ............................................................................................ 65 Figure 4.3: Picture of the IBA nozzle and strip and pixel chambers ................................................. 66 Figure 4.4: Y position of Center of gravity for strip chambers (red) and data expected from trajectories files (black) .............................................................................................................. 68 Figure 4.5: ISEU intensity (red curve), expected intensity (black curve), strip counting (green curve, threshold of 80 counts). .............................................................................................................. 69 Figure 4.6: Y position of center of gravity for strip chambers (black) and data expected from trajectories files (red), with a correct time synchronization,at the beginning (top plot) and at the end (bottom plot) of the run ....................................................................................................... 70 Figure 4.7: Y position of center of gravity for strip chambers (black) and data expected from trajectories files (red), with a correct time synchronization and the correction of the trajectories data period. ................................................................................................................................. 70 Figure 4.8: Horizontal - vertical strip counting ratio ......................................................................... 71 Figure 4.9: Ratio of horizontal to vertical strip count for four different scans .................................. 72 Figure 4.10: Sum of the counts over all the strips for a single acquisition versus the acquisition number. ...................................................................................................................................... 72 Figure 4.11: A single scan line. Ibeam expected (black curve), ISEU intensity (red curve), strip counts (green curve). .................................................................................................................. 73 Figure 4.12: Vertical strip detector: mean and sigma of the counts of each acquisition when only the central part of each scanning line is considered. Each plot contains one point per each one of the 30 vertical lines of the scan. ................................................................................................. 74 Figure 4.13: Center of gravity measured with horizontal strip chamber for the “edge” of the field. 74 Figure 4.14: Distribution of X components of the center of gravity on the mean value for a single scanning line .............................................................................................................................. 76 Figure 4.15: Mean X values per scanning line for strip (blue), trajectories expected (gree) and measured (red) on the detector surface. ..................................................................................... 77 4 Figure 4.16: Distribution of Y components of the center of gravity on the expected values for a single scanning line. ................................................................................................................... 78 Figure 4.17: Computed scanning speed for a scan............................................................................. 79 Figure 4.18: mean FWHM (top plot) and sigma FWHM (bottom plot) for all the runs................... 80 Figure 4.19: FWHM X and Y for all the acquired points (top plot) and in a particular scanning line (bottom plot). ............................................................................................................................. 81 Figure 5.1: An exploded view of the strip chamber ........................................................................... 83 Figure 5.2: Sketch of the CATANA facility ...................................................................................... 85 Figure 5.3: Sketch of the CATANA beam line .................................................................................. 86 Figure 5.4: Picture of the CATANA beam line with the MOPI strip detector .................................. 87 Figure 5.5: Mean background per second of the two strip chambers ................................................ 88 Figure 5.6: Comparision of the left tail of beam profiles with and without background for the strip chamber and the film.................................................................................................................. 89 Figure 5.7: (a) Single film profile with (red points) and without (black points) saturation effects correction. (b) Integrated film profile with (red points) and without (black points) correction. 89 Figure 5.8: (a) Horizontal strip and film profile for a modulated field. (b) Distribution of the point to point differences between the two profiles ................................................................................ 90 Figure 5.9: (a) Horizontal strip and film profile for a not modulated field. (b) Distribution of the point to point differences between the two profiles ................................................................... 90 Figure 5.10: Distribution of the point to point differences between the calibration coefficients computed with modulated and not modulated beam, for horizontal (a) and vertical (b) strip detector. ...................................................................................................................................... 91 Figure 5.11: Integrated counts over all the strips with 5 Hz of acquisition rate: (a) first run (b) eleventh run. ............................................................................................................................... 91 Figure 5.12: Distribution of the vertical and horizontal strip detectors gains ratios; data are referred to the two curves of Figure 5.11 (b). .......................................................................................... 92 Figure 5.13: Distribution of the integrated counts over all the strips (horizontal chamber) with a 5 Hz acquisition rate. .................................................................................................................... 93 Figure 5.14: Comparison of the fluence integrated over the acquisition run for the vertical and horizontal strip chambers and for the CATANA ionization chambers, for 8 different runs. .... 93 Figure 5.15: Comparison between the skewness and the collected charge. ...................................... 95 Figure 5.16: Skewness vs acquisition number and distribution of the values. .................................. 96 Figure 5.17: Skewness and c.o.g. vs acquisition number in a beam line configuration with the modulator wheel ......................................................................................................................... 96 Figure 5.18: Vertical and horizontal beam profiles, for four different magnet settings, in a not modulated beam condition ......................................................................................................... 97 Figure 5.19: Vertical and horizontal beam profiles, for four different magnet settings, in a modulated beam condition. ........................................................................................................ 97 Figure 5.20: Beam profile at the isocenter, in air, detected with a silicon diod................................. 98 Figure 5.21: RT (from vertical diode profile) vs skewness (from horizontal strip chamber profile) computed for seven different magnet configurations. ............................................................... 99 Figure 5.22: ST (from vertical diode profile) vs skewness (from horizontal strip chamber profile) computed for seven different magnet configurations. ............................................................. 100 Figure 5.23: Comparison between skewness computed for modulated and not modulated beam, for four different magnet setting. ................................................................................................... 100 5 Introduction The work I have done during the three years of Ph.D. was mainly aimed at the development and test of different kind of ionization chambers to be used as beam monitor during hadrontherapy treatments; it has been done within a collaboration between the University and the INFN of Torino. The cure of tumours with hadrons (protons and carbon ions) presents, with respect to the conventional x-rays radiotherapy, some advantages, from both the biological and physical point of view. In the first chapter the main effects caused by the radiation in the interaction with the biological matter are explained, discussing in particular way the damages leaded by the use of the high LET (linear energy transfer) particles to the mammalian cells. The physical description of the interaction between radiation and matter are treated in the first part of the second chapter; in particular the energy loss by hadrons in function of depth is presented. There is then a brief overview of the accelerators used in hadrontherapy, followed by the description of the passive and active beam spreading system. The last part of this chapters presents the description of the mechanical structure and the electronic read-out of the ionization chambers designed and built by the group of Torino. My personal contribution given to the development of these detectors is explained in the following chapters in detail. The chapters 3, 4 and 5 present the experimental setup and the results of three different tests. The first (chapter 3) has been done in April 2001 at the GSI centre of Darmstadt, in Germany, where patients are treated with carbon ions. The accelerator is a synchrotron and the beam spread is obtained with the voxel scanning technique. In particular I have described the acquisition system dedicated to this scanning method, and the analysis aimed to the characterisation of the detector. In October 2002 a test within a collaboration with the IBA (Ion Beam Application) was done at Louvain-la-Neuve. This test was aimed to check and characterise the raster scanning system developed by the IBA. In chapter 4 the experimental setup and the analysis done on the data collected are reported. The last chapter describes the development of a new strip ionization chamber designed for the beam line of CATANA, at INFN LNS Catania. Since four years in fact in Catania patients are treated for ocular pathologies with protons. Until the end of the 2004 no detectors for the on-line check of the beam structure were present on that beam line. The last part of my doctorate work was aimed to design, build and test a detector dedicated to CATANA, for the on-line verification of the beam shape. In chapter five also the mathematical functions used to study the beam stability, optimizing the computation speed, are reported. For the end of February 2005 the first treatment session with the detector placed on the beam line are planned. 6 1 The cure of cancer with radiation 1.1 The significance of radiobiology in radiotherapy 1.1.1 The role of radiotherapy in the management of cancer Radiotherapy is one of the two most effective treatments for cancer. Surgery, which of course has the longer history, is in many tumour types the primary form of treatment and it leads to good therapeutic results in a range of early non-metastatic tumours. The combination between radiotherapy and surgery often achieve a reasonable probability of control for many tumours, and in case of tumour of head and neck, cervix, bladder, prostate and skin the only radiotherapy gives good results. In addition to these examples of the curative role of the radiation therapy, many patients gain valuable palliation by radiation. Chemotherapy is the third most important treatment modality at present time. Many patients receive chemotherapy at some point in their management and useful symptom relief and diseases arrest are often obtained [1.1]. The following is a brief outline of the role of radiotherapy in six disease sites: Bladder: the success of surgery or radiotherapy varies widely with the stage of disease; both approaches give 5-year survival in excess of 50%. Breast: early breast cancer, not known to have metastasized, are usually treated by surgery and this have a tumour control rate in the region of 50-70%. Radiotherapy given to the chest wall and regional lymph nodes increases control by up to 20%. Hormonal therapy and chemotherapy also have significant impact on patient survival. In patients who have evidence of metastatic spread at the time of diagnosis, the outlook is poor. Cervix: disease that has developed beyond the in situ stage is often treated by combination of intracavitary and external-beam radiotherapy. The control rate varies widely with stage of the disease, from around 70% in stage I to perhaps 7% in stage IV. Lung: most lung tumours are inoperable and for them the 5-year survival rate for radiotherapy combined with chemotherapy is in the region of 5%. Lymphoma: in Hodgkin’s disease radiotherapy alone achieves a control rate of around 50% and when combined with chemotherapy this may rise to 80%. Prostate: where there is evidence of local invasion, surgery and radiotherapy have similar level of effectiveness, with 10-years control rates in the region of 50%. Chemotherapy makes a limited contribution to local control. Very substantial number of patients with common cancers achieves long-term tumour control largely by the use of radiation therapy. There are three main ways in which an improvement in radiotherapy might be obtained: by rising the standard of radiation dose prescription and delivery with respect to those currently in use. by improving radiation dose distributions beyond those that are conventionally achieved, either using techniques of conformal radiotherapy with photons, or by use of hadronic beams. by exploring radiobiological initiatives. 1.1.2 The role of radiation biology Experimental and theoretical studies in radiation biology contribute to the development of radiotherapy at three different levels, moving in turn from the most general to the most specific: 7 Ideas: providing a conceptual base for radiotherapy, identifying the mechanisms and processes that underline the response of tumour and normal tissues to irradiation and which help to explain the observed phenomena. Examples are: hypoxia, reoxigenation, tumour cell repopulation or mechanism of repair of DNA damage. Treatment strategy: development of specific new approaches in radiotherapy. Examples are hypoxic cell sensitizers, high LET-radiotherapy, hyperfractionation. Protocols: advice on the choice of schedules for clinical radiotherapy, for instance conversion formulae for changes in fractionation or dose rate, or advice on whether to use chemotherapy concurrently or sequentially with radiation. We may also include under this heading methods for predicting the best treatment for the individual patient (individualized radiotherapy). There is no doubt that radiobiology has been very fruitful in the generation of new ideas and in the identification of potentially exploitable mechanisms. A variety of new treatment strategies have been produced, but few of these have so far led to demonstrable clinical gains. 1.1.3 The time-scale of effects in radiation biology Irradiation of any biological system generates a succession of processes that differ enormously in time scale. This is illustrated in Figure 1.1 where these processes are divided into three phases. Figure 1.1: Time scale of the effects of radiation exposure on biological systems. The physical phase consists of the interaction between charged particles and the atoms of which the tissue is composed. An high speed electron takes about 10-18 seconds to traverse the DNA molecules and about 10-14 seconds to pass across a mammalian cell. As it does so, it interacts mainly with the orbital electrons, ejecting some of them from atoms (ionization) and raising other to higher energy levels within an atoms or molecule (excitation). If sufficiently energetic, these secondary electrons may excite or ionize other atoms near which they pass, giving rise to cascade of ionizing events. For 1 Gy of absorbed radiation dose there are in excess of 105 ionizations within the volume of any cell of diameter of 10m. The chemical phase describes the period in which the damaged atoms and molecules react with other cellular components in rapid chemical reactions. Ionization and excitation lead to the breakage of chemical bonds and the formation of broken molecules, known as ‘free radical’. These are highly reactive and engage in a succession of reactions that lead eventually to the restoration of electronic 8 charge equilibrium. Free-radical reactions take place within approximately 1ms of radiation exposure. An important characteristic of the chemical phase is the compensation between scavenging reactions, for instance with sulphydryl compounds that inactivate the free radical, and fixation reaction that lead to stable chemical changes in important biological molecules. The biological phase include all subsequent processes. These begin with enzymatic reactions that act on the residual chemical damage. The vast majority of lesions, for instance in DNA, are successfully repaired. Some rare lesions fail to repair and this is what lead eventually to cell death. Cells take time to die; indeed, after small dose of radiation they may undergo a number of mitotic divisions before dying. It is the killing of stem cells and the subsequent loss of the cells that they would have given rise to, that causes the early manifestations of normal-tissue damage during the first week and month after irradiation exposure. Examples are: breakdown of the skin or mucosa, denudation of the intestine and haemopoietic damage. A secondary effect of cell killing is compensatory cell proliferation, which occurs both in normal tissue and tumours. At later times after the irradiation of normal tissue the so called ‘late reactions’ appear. These include fibrosis and telangiectasia of the skin, spinal-cord damage and blood vessel damage. An even later manifestation of radiation damage is the appearance of secondary tumours (i.e. radiation carcinogesis). The timescale of the observable effects of ionizing radiation may thus extend up to many years after exposure. 1.1.4 Response of normal and malignant tissues to radiation exposure The effects of radiation exposure become apparent during the weeks, month and years after radiotherapy. These effects are seen both in tumour tissues and normal tissues that surround a tumour and which are unavoidably exposed to radiation. The response of a tumour is seen by regression, often followed by regrowth, but perhaps with failure to regrow during the normal lifespan of the patient (which is termed cure or local control). The response of normal tissue to therapeutic radiation exposure range from those that cause mild discomfort to other that are life threatening. The speed at which a response develops varies widely from one tissue to another and often depends on the dose of radiation that the tissue receives. Generally speaking the haemopoietic and epithelial tissues manifest radiation damage within weeks of radiation exposure, whereas damage to connective tissue becomes important at later times. 1.1.5 Response curve, dose-response curves and isoeffect relationship The damage that is observed in an irradiated tissue increases, reaches a peak, and then may decline (Figure 1.2A). It could be possible use the measured response at some chosen time after irradiation, such at the time of maximum response, but the timing of the peak may change with radiation dose and this would lead to some uncertainty in the interpretation of the results. A common method is to calculate the cumulative response by integrating this curve from left to right (Figure 1.2B). The response for some normal tissue gives a cumulative curve that rises to a plateau, and the height of the plateau is a good measure of the total effect of that radiation dose on the tissue. Other normal tissue response, in particular the late responses seen in connective and vascular tissues, are progressive and the cumulative response continues to rise. 9 Figure 1.2: Four types of chart leading to the construction of an isoeffect plot. (A) Time-course of radiation damage in a normal tissue. (B) The cumulative response. (C) A dose-response relationship, constructed by measuring the response (R) for varius radiation doses (D). (D) Isoeffect plot for a fixed level of normal tissue damage. The next stage in a study of the radiation response of a tissue consist in varying the radiation dose and thus investigating the dose-response relationship (Figure 1.2C). Radiation dose-response curves have a sigmoid shape, with the incidence of radiation tending to zero as dose goes to zero and tending to 100% at very large doses. Many mathematical functions could be used with these properties, but the most standard formulation used is the Poisson distribution. Munro and Gilbert published a landmark paper in 1961 in which they formulated the target cells hypothesis of tumour control: ‘The object of treating a tumour by radiotherapy is to damage every single potentially cell to such an extent that it cannot continue to proliferate’. From this idea and the random nature of cell killing by radiation they derived a mathematical formula for the probability of the tumour cure after irradiation ‘of a number of tumours each composed of N identical cells’. More precisely, they showed that this probability depends only on the average number of clonogens surviving per tumour. When describing tumour cure probability (TCP), it is the probability of zero surviving clonogens in a tumour that is of interest. This is the zero-order term of the Poisson distribution and if denotes the average number of clonogens after irradiation this is simply: TPC = e- (1.1) The simple exponential was later replaced by the linear-quadratic model and thus we arrived at what could be called the standard model of tumour control: TCP=exp[-N0*exp(-D-dD)] (1.2) Here N0 is the number of clonogens per tumour before irradiation and the second exponential is simply the surviving fraction after dose D given with dose per fraction d, according to the linear quadratic model. Thus when we multiply these two quantities we obtain the average number of surviving clonogens. Diagrams similar to Figure 1.2A, B, C can also be constructed for fractionated radiation treatment, although the results are easiest to interpret when the fraction are given over a time that is short compared with the time scale of development of the response. If we change the schedules of dose fractionation, for instance by giving a different number of fractions, changing the fraction size 10 or radiation dose rate, we can then investigate the therapeutic effects in term of isoeffect plot (Figure 1.2D). Experimentally this is done by performing multiple studies at different doses for each chosen schedule and calculating a dose-response curve. We then select some particular level of effect (T in Figure 1.2C) and read off the total radiation dose that gives this effect. For effects on normal tissues the isoeffect will often be some upper limit of tolerance of the tissue, perhaps expressed as a probability of tissue failure. The isoeffect plot show how the total radiation dose for the chosen level of effect varies with dose schedule. The dashed line in Figure 1.2D illustrates how therapeutic conclusion may be drawn from isoeffect curves. 1.2 DNA damage and cell killing 1.2.1 Initial process of radiation damage As mentioned above the irradiation of a biological system initiates a series of processes that can be classified in term of time scale over which they act. The physical, chemical and biological phases of this processes have been described in section 1.1.3. An electron with an energy of 1 MeV has a range in soft tissue of a few millimeters [1.2]. In the early part of its track the particle moves very quickly and its rate of energy deposited is low; the result is a relative straight track in which the ionizations may be separated by distance of around 0.1mm on average. We describe this as radiation with a low linear energy transfer (LET). As the electron slows down, it interacts more strongly with the orbital electrons in the medium. Its rate of energy loss increase, the track becomes more tortuous due to the stronger collision, and the ionization density increase. Figure 1.3A shows a computer simulation of the tracks of 1 KeV electrons, representing a very small part of the tracks of 1 MeV electrons. The important feature is the tendency towards clustering of the ionization events at the end of the track, each cluster having the size of a few nanometers. Within each electron track there is opportunity of interaction between the products of separate ionization events and it may be, particularly at low dose rate or following acute radiation doses up to few Gy, that the main biological effects of radiation (i.e. cell killing and mutation) are predominantly due to damage that is produced by these ‘hot spots’. Within perhaps 10-10 seconds of exposure to either photon or particles beam, the irradiated volume will contain atoms that have been ionized and a corresponding number of free electrons, all produced by cascade of atomic reaction just described and with a rather non-uniform spatial distribution. The number of ionization produced at therapeutic dose levels is very large – approximately 105 ionizations per cell per Gy – but the vast majority of these produce no toxic damage. The biological effect is influenced by three factors: free radical scavenging processes, the number of ionizations that are closed to DNA to damage it, and the cellular repair process. Free-radical process Since biological systems consist largely of water, the bulk of the ionization produced by irradiation occur in water molecules. Negatively charged free electron that are produced by ionization will rapidly become associated with polar water molecules, greatly reducing their mobility. The configuration of an electron surrounded by water molecules (a ‘hydrated electron e -aq) has a degree of stability and lifetime under physiological condition of few microseconds. The water molecule that has lost an electron is a highly reactive positive charged ion. It quickly breaks down to produce a hydrogen ion (H+) and an (uncharged) OH radical. OH is a molecule that normally doesn’t exit in water, indeed the stable configuration is H2O. The uncharged OH radical has an unpaired electron (‘unattached valence’) that makes it highly reactive. We designate it as a free radical thus: OH. Free radicals are simply fragment of broken molecules. OH is different from 11 OH+ which is positive charged ion: the OH radical has equal number of protons and orbital electrons but because of unpaired electron is chemical reactive (some ions may also be radical, for example a water molecule that has lost an electron is actually H20+, a radical cation). Similarly, H+ is a bare proton, positively charged, whilst H is a proton plus an electron (neutral charge) but again highly reactive because the stable form of hydrogen is H2. Figure 1.3: (A) Computer-simulated tracks of 1 KeV electrons. Note the scale in relation to the 2.3 nm diameter of DNA double helix (adapted from Chapman and Gillespie, 1981). (B) Illustrating the concept of a local multiply damaged site produced by a cluster of ionizations impinging on DNA Around 10-10 seconds after irradiation there will be three principal radiolysis products of water: e-aq, OH and H. These highly reactive species will go on to take part in further reactions. An important one is: OH + OH H2O2 the production of hydrogen peroxide. Oxygen, if present, plays an important part in the free-radical reactions following the irradiation. Molecular oxygen has a high affinity for free-radical (R): R + O2 RO2 giving rise to further reactive products and acting to fix the free-radical damage. The oxygen effect in radiation cell killing has often been explained in term of this type of process. In biological system the free radicals produced in water may react with essential macromolecules. A vast range of reaction takes place, most of which are unimportant for the survival and functioning of the cell. The most important reactions are those with DNA, because of the uniqueness of many parts of this molecule. Damage of DNA by free radicals produced in water is called the indirect effect of radiation; ionization of atoms that are part of the DNA molecule is the direct effect. 1.2.2 Radiation damage to DNA The structure of DNA Deoxyribonucleic acid (DNA) is a large molecule that has a characteristic double helix structure consisting of two strands, each made up of a sequence of nucleotides (Figure 1.4). 12 Figure 1.4 The structure of DNA, in which the four bases (G,C,T,A) are linked through sugar groups to the sugarphospate backbone. A nucleotide is a subunit in which a ‘base’ is linked through a sugar group to a phosphate group. The sugar is deoxyribose, which have five-atom ring: four carbons and one oxygen. The ‘backbone’ of the molecules consists of alternating sugar-phosphate group. There are four different bases. Two are single-ring group (pyrimidines): thymine and cytosine, and two are double-ring group (purines): adenine and guanine. It is the order of these bases along the molecule that specifies the genetic code. The two strands of the double helix are held together by hydrogen boning between the bases. These bonds are made between thymine and adenine, and between cytosine and guanine; the bases are paired in this way along the length of DNA molecule. During the S phase of the cell cycle, DNA synthesis takes place (the process of replication) in which every base pair is accurately duplicated. The first stage in the manufacture of proteins is the construction by the process of transcription of a messenger RNA (i.e. mRNA) that has a similar to a single strand of DNA expect that the sugar groups are ribose in place of deoxyribose, and the thymine is replaced by uracil. The decoding is based on the pairing of bases: A-U, C-G, G-C, U-A. Transcription is performed by RNA polymerases, which bind to DNA and generate the corresponding mRNA. Radiation damage to DNA Early experiments showed that irradiation leads to a loss of viscosity in DNA solutions. Subsequently this has been shown to result from DNA strand breaks. There are two categories of DNA strand breaks: single-strand (SSB) and double-strand (DSB). The detection of these depends on a study of the size distribution of fragments of DNA after extraction from irradiated cells. As shown in Figure 1.5, there is a variety of other types of DNA lesion that may have a role in cellular responses to radiation or chemical damage. 13 Figure 1.5 Types of damge to DNA produced by radiation and chemical agents. There are many sources of evidence to suggest that DNA damage is the critical event in radiation cell killing and mutation, including the following: Micro-irradiation studies show that to kill cells by irradiation of only the cytoplasm requires far higher radiation dose than irradiation of the nucleus. Isotopes with short-range emission (such as 3H, 125I) when incorporated into cellular DNA efficiently produce radiation cell killing and DNA damage. The incidence of chromosomal aberrations following irradiation is closely linked to cell killing. The number of lesions induced in DNA by radiation is far greater than those that eventually lead to cell killing. A dose of radiation that induces on average one lethal event per cell will kill 63% and leave 37% still viable (this result from Poisson statistic) and we call this the D0 dose. D0 values for oxic mammalian cells are usually in region of 1-2 Gy. The numbers of DNA lesions per cell that are detected immediately after such a dose have been estimated to be approximately: Events per D0 Base damage Single-strand breaks Double-strand breaks >1000 ~1000 ~40 In addition, cross-links between DNA strands and between DNA and nuclear proteins are formed (Figure 1.5). Irradiation at clinical used doses thus induces a vast amount of DNA damage, most of which is successfully repaired by the cell. In a variety of experimental situations it has been found that the incidence of cell killing fails to correlate with the number of SSB induced, but relates better to the incidence of DSB. Significantly, a dose of hydrogen peroxide that induces many DDB produces little cell killing and few DSB unless the number of SSD is so large that they are close enough to form DSB. On this basis it is generally believed that DSB are the critical lesions for radiation cell killing in most cell types, although experimental evidences indicate that only some DSB are important. 14 Modifier High-let rad Hypoxia Thiols Hypertermia Hydorgen per. Cell kill 0 DSB 0 SSB 0 Base 0 0 0 DNA-protein cross-link 0 - , Increased; , Decreased; 0 little or no effect; -, not know. See Frankenburg Schwager (1989) for further information Table 1-1 Double-strand DNA breaks correlate best with cell killng 1.2.3 Cell death in mammalian tissues The definition of effective cell death could be described as the loss by the cell of its ability to produce progeny. This end point is defined as clonogenic cell death. But for organized tissue and tumours the definition of cell death is much more complicated. We recognized that clonogenic potential is the essential element for the maintenance of a cell line, either in vitro or in organized tissue, but there are other important issues in the behavior of complex tissue system. Normal senescence of cells is one of these important issues. Another important issue is the removing of the cells that are in the wrong place at the wrong time. Example of this would be the metastatic arrival of tumour cells transported from a primary tumour elsewhere or the resolution of inflammatory processes. One can define at least two types of cell death that go beyond the end point of clonogenic potential and involve the actual disappearance of the cells. Acute pathological cell death, which generally results from cell injury or from lack of oxygen or essential metabolites, is called necrosis. Necrosis is characterized by a tendency for cells to swell and ultimately to lyse, which allows the cell’s contents to flow into cellular space. Necrosis is usually accompanied by an inflammatory response. In the case of neoplasm, necrosis is most often seen in rapidly growing tumours, where the tumour mass outgrows its blood supply and regions of the tumour became undernourished in oxygen and energy sources. In this case inflammation is not a characteristic of the necrotic process. For cell death that results from senescence or cell population control, by contrast, the characteristic process, called apoptosis, involves shrinkage of the nucleus and cytoplasm, followed by fragmentation and phagocytosis of these fragments by neighbouring cells or macrophages. The contents of cells do not usually leak into extracellular space, so there is no inflammation. Since there is no inflammation accompanying apoptosis, the process is histologically quite inconspicuous [1.3]. 1.3 The oxygen effect The response of cells to ionizing radiation is strongly dependent upon oxygen (Gray et al., 1953; Wright and Howard-Flanders, 1957). This is illustrated in Figure 1.6 for mammalian cells irradiated in culture. The cell surviving fraction is shown as a function of radiation dose administrated either under normal aerated condition or under hypoxia, generally achieved by flowing nitrogen gas over the surface of the cells suspensions for a period of 30 minutes or more. The enhancement of radiation damage by oxygen is dose-modifying, i.e. the radiation dose that gives a particular level of 15 survival is reduced by the same factor at levels of survival. This allows us to calculate an oxygen enhancement ratio (OER): OER = dose in N2 for surviving fraction, S/S0 dose in O2 for surviving fraction, S/S0 for the same level of biological effect. For most cells the OER for x-rays is around 3.0. However, some studies suggest that at radiation dose of 3 Gy or less the OER is actually reduced (Palcic and Skarsgard, 1984). This is an important finding because this is the dose range for clinical fractionation treatments [1.4]. Figure 1.6 Survival curves for culturaed mammalian cells exposed to x-rays under oxic or hypoxic conditions, illustrated the radiation dose-modifying effect of oxygen. Note that the broken line extrapolate back to the same point on the survival axis (n=5.5). It has been demonstrated from rapid-mix studies that the oxygen effect only occurs if oxygen is present either during irradiation or within a few millisecond thereafter (Howard-Flanders and Moore, 1958). The dependence of the degree of sensitization on oxygen tension is shown in Figure 1.7. By definition, the OER under anoxic condition is 1.0. As the oxygen level increases, there is a steep increase in radiosensitivity (and thus in OER). The greatest change occurs from 0 to about 20 mmHg; further increase in oxygen concentration, up to that seen in air (155 mmHg) or even to 100% oxygen (760 mmHg), produces a small though definite increase in radiosensitivity. Also shown in Figure 1.7 is the oxygen partial pressure range typically found in arterial and venous blood. Thus, from a radiobiological standpoint most normal tissues can be considered to be well oxygenated, although it is now recognized that moderate hypoxia is a feature of some normal tissue such as cartilage and skin. 16 Figure 1.7 Variation of oxygen enhancement ratio (OER) with oxygen tension. The horiziontal arrows indicate the range of physiological blood oxygen tensions on the lower scale. Adapted from Denekamp (1989). The mechanism responsible for the enhancement of radiation damage by oxygen is generally referred to as the oxygen-fixation hypothesis and is illustrated in Figure 1.8. When radiation is absorbed in a biological material free radicals are produced. These are highly reactive molecules and can thus break chemical bonds, produce chemical changes, and initiate the chain of events that results in biological damage. They can be produced either directly in target molecule (usually DNA) or indirectly in other cellular molecules and defuse far enough to reach and damage critical targets. Most of the indirect effects occur by free radical produced in water, since this makes up to 70-80% of mammalian cells. It is the fate of the free radicals ultimately produced in critical target, designated in R in Figure 1.8, that is important. If oxygen is present, then it can react with R to produced RO2 which then undergoes further reaction ultimately to yield ROOH in target molecule. Thus we have a change in the chemical composition of the target and the damage is chemically fixed. Subsequently this damage can be processed enzimatically and perhaps repaired. In the absence of oxygen, or in the presence of reduced species, R can react with H+, thus restoring its original form. Figure 1.8 The oxygen fixation hypothesis. Free radical produced in DNA either by direct or indirect action of radiation can be rapired under hypoxia but fixed in the presence of oxygen. Adapted from Hall (1989). 17 1.4 Particle beams in radiotherapy Figure 1.9 shows examples of microdosimetric calculation of ionization tracks from -rays or particles passing through a cell nucleus (Goodhead, 1988). At the scale of cell nucleus, the -rays deposit much of their energy as single isolated ionizations or excitations and much of resulting of DNA damage is efficiently repaired by enzymes within the nucleus. About 1000 of these sparse tracks are produced per Gy of absorbed dose. The -particles produce fewer tracks but the intense ionization within each tracks leads to more severe damage where the track intersects vital structure such as DNA. The resulting DNA damage may involve several adjacent base pairs and will be much more difficult or even impossible to repair; this is probably the reason why these radiations produce steeper cell survival curves and allow less cellular recovery than x-rays. At the low doses that are encountered in environmental exposure, only some cells will be traversed by a particle and many cells will be unexposed [1.5]. Figure 1.9 The structure of particle tracks for low-LET radiation (above) and particle (below). The cricles indicate the typical syze of mammalian cell nuclei. Note the tortuos tracks of low-energy secondary electrons. From Goodhhead (1988). Linear energy transfer (LET) is the physical quantity used to describe the density of ionization in particle tracks. LET is the average energy given up by a charged particle traversing a unitary distance, normally expressed in KeVm-1. In Figure 1.9 the -rays have a LET about 0.3 KeVm-1 and are described as low-LET radiation. The -particles have a LET of about 100 KeVm-1 and are an example of high-LET radiation. Also neutrons are describes as high-LET radiation, even if they are uncharged. In fact they do not interact with the orbital electrons in the tissue through which they pass and they do not directly produce ionization. They do, however, interact with atomic nuclei from which they eject slow, densely ionization protons. It is this secondary production of knock-on protons that confers high LET. 18 1.4.1 Biological effects depend upon LET As LET increases, radiation produces more cell killing per Gy. Figure 1.10 shows the survival of human T1G cells plotted against dose for high different radiations, with LET varying from 2 keVm-1 (250 kVp x-rays) to 165 keVm-1 (2.5 MeV particles). As LET increases, the survival curves became steeper; they also become straighter with less shoulder, which indicates either a higher ratio of lethal to potentially lethal lesion (in lesion-interaction models) or that high-LET radiation damage is less likely to be repaired correctly. For particles of identical atomic composition, LET generally increases with decreasing particles energy. However, notice that 2.5 MeV particles are less efficient compared with 4 MeV particles even though they have a higher LET; this is due to the phenomenon of overkill indicated in Figure 1.11. The relative biological effectiveness (RBE) of a radiation under test (e. g. a high LET radiation) is defined as: dose of reference radiation RBE= dose of test radiation to give the same biological effect. Figure 1.10 Survival of human kidney cells exposed in vitro to radiations of different LET. From Barendsen (1968). The reference low-LET radiation is usually 250 kVp x-rays. Figure 1.11 shows RBE values for T1g cells featured in Figure 1.10. Curves have been calculated at cell survival levels of 0.8, 0.1 and 0.01, illustrating the fact that RBE is not constant but depends on the level of biological damage and hence on the dose level. RBE rises to a maximum at a LET of about 100keVm-1, then falls for higher values of LET due to overkill. For cells to be killed, energy must be deposited in a number of critical sites in the cell. 19 Figure 1.11 Dependece of RBE on LET and the phenomenon of overkill by very high-let radations. From Barendsen (1968). Sparsely ionizing low-LET radiation is inefficient because more than one particle may have to pass through the cell to kill it. Densely ionizing very high-LET radiation is also inefficient because it deposits more energy than necessary in critical sites. These cells are overkilled and per Gy there is then less likelihood that other cells will be killed, leading to a reduced biological effect. Radiation of optimal LET deposits just enough energy per cell to inactivate the critical targets. This optimum LET is usually around 100mm-1 but it does vary between different cell types and depends on the spectrum of LET values in the radiation beam as well as the mean LET. As LET increases, the oxygen enhancement ratio decreases. The measurements shown as an example in Figure 1.12 were also made with T1g cells of human origin. The sharp reduction of OER occurs over the same range of LET as the sharp increase in RBE (Figure 1.11). Figure 1.12: The oxygen enhancement ratio (OER) decreases with increasing LET. Closed circles refer to monoenergetic -paritcles and deuterons; the open triangle to 250 kVp x-rays. From Barendsen (1968). 20 1.4.2 The biological basis for high-LET radiotherapy We have seen in Figure 1.12 that the differential radiosensitivity between more oxygenated (more resistant) cells is reduced with high-LET radiation. Therefore, tumour sites in which hypoxia is a problem in radiotherapy (some head and neck tumour, for example) might benefit from high-LET radiotherapy in the same way as from chemical hypoxic-cell sensitizers. The effect of low-LET radiation on cells is strongly influenced by their position in the cell cycle, wit cells in S-phase being more radioresistant than cells in G2 or mitosis. Cells in stationary phase also tend to be more radioresistant than cells in active proliferation. Both these factors act to increase the effect of fractionated radiotherapy on more rapidly cycling cells comparing with those cycling slowly or not at all, because the rapid cycling cells that survive the first few fractionation are statistically more likely to be caught in a sensitive phase and so killed by a subsequent dose, a process termed ‘cell-cycle resensitization’. This differential radiosensitivity due to cell cycle position is considerably reduced with high-LET radiation and is a reason why we might expect high-LET radiotherapy to be beneficial in some slowly growing, x-ray resistant tomours. A third biological rationale for high-LET therapy is based on the observation that the range of radiation response of different cells types is reduced with high-LET radiation compared with x-rays. This is shown in Figure 1.13, which summarized the in vitro response of 20 human cells lines to photon and neutron irradiation ( Britten et al, 1992). This reduced range of response affects the benefit expected, which is the balance between the tumour and the normal-tissue response. Thus, if tumour cells are already more radiosensitive to x-rays than the critical normal-cells population, high-LET radiation should not be used since this would reduce an already favourable differential. Possible examples are seminomas, lymphomas and Hodgking’s disease. However, if the cells are more resistant to x-rays than the critical normal cells, high-LET radiation might reduce this difference in radiosensitivity and thus would effectively ‘sensitize’ the tumour cells population relative to a fixed level of normal-tissue damage. High-LET radiation would be advantageous in this case. Figure 1.13: response of 20 human tumour cell lines to (A) 4MVp photons, or (B) p(62.5)-Be neutrons. The vertical lines show the photon (2Gy) and the neutrons (0.68 Gy) doses that give the same median cell survival; the average RBE is therefore 2/0.68=2.94. 21 2 Radiotherapy with hadrons 2.1 Proton interactions with matter 2.1.1 Introduction For an understanding of the dose distribution produced by protons, a knowledge of their energy loss and scattering is needed. Protons traversing matter lose energy through successive collision with atoms and molecules of the material. With respect to energy loss, the most important interaction is between the protons and the atomic or molecular electrons. The interactions between the protons and the atomic nucleus effect the protons flux (nuclear reactions), and the proton trajectory ((in)elastic scattering) [2.1]. The most important parameter characterizing the energy loss of an incident proton is the stopping power, which is the mean energy loss per unit path length in material. A full description of the proton energy loss process, however, requires more detailed information that is provided by the stopping power alone. The amount of energy transferred from a proton to an atomic electron, as well as the number of interactions that occur per unit path length has a probabilistic distribution. Moreover, there is a certain probability that very energetic electrons are produced ( -electrons or rays) which can travel a considerable distance before their energy is deposited. The most important contribution to proton scattering comes from the electromagnetic interaction with the nucleus. This gives rise to a small scattering angles, but since there are a large number of collisions, the effect can be considerable. If the impact parameter is small also the hadronic interaction contributes to elastic scattering. In addition inelastic interaction can occur: these can be either an inelastic scattering process during which the proton transfer energy to the nucleus (which will then be in an excitated state and decay by -emission) or a nuclear reaction process (such as (p,n), (p,d), (p,2p) or (p,3p)) where the incident proton will disappear. In case of scattering of protons by very light nuclei, such as protons in hydrogen, also the recoil nucleus can travel a considerable length before its energy is fully deposited. 2.1.2 Proton interactions with electrons: energy loss Within the energy range of importance in proton therapy (from stopping proton to about 250 MeV) it is convenient to consider two energy intervals separately: Low energy: below ≈ 0.5 MeV protons can pick up orbital electrons and form hydrogen. Also energy can be lost to atomic nuclei due to electromagnetic interactions (nuclear stopping power). These are complicated process, but fortunately they only play a role at the very last microns of a proton track. It is important for the subject of microdosimetry, which deals with the energy loss process on a microscopic scale (for example the study of the effect of ionizing radiation on DNA). High energy: for proton energies between ≈ 0.5 MeV and 250 MeV the atoms in the stopping medium can be excited or ionized. The collision process is well understood and in principle the stopping power can be calculated theoretically. The mean energy loss per proton S can be described by the Bethe theory [2.2] which leads to the following expression: 22 1 S 1 dE Z 1 K L dz A 2 (2.1) with: K = 2πr2e mc2Nav ≈ 0.135 MeV cm2g-1 (2.2) where re = e2/40mc2 is the classical electron radius, 0 is the permittivity of the vacuum (which is introduced by the use of SI units), mc2 is the electron rest mass energy, Nav Avogadro’s number, is the particle velocity in unit of velocity of light, Mc2 is the proton rest mass ≈ 938.3 MeV, E the proton kinetic energy, Z and A are the atomic number and relative atomic mass of the target atom. The quantity L() takes into account the fine details of the energy loss process and is written as the sum of three terms: L() = L0() + L1() + L2() (2.3) The first term is given by: 2mc 2 2Tmax L0 ln 2 1 C 2 2 2 ln I 2 Z (2.4) where I is the average excitation potential of the atoms of the medium, C/Z the shell correction and the density-effect correction. The shell correction plays a role at low velocities and deals with effects due to the finite speed of the proton compared to the velocity of bound electrons. It can be as high as 10% but it is usually implicitly taken into account by the choice of the excitation potential I. The density correction takes into account the polarization of the medium to the passage of the projectile proton. It can be neglected ( « 0.1%)for proton energies below 500 MeV. Tmax is largest possible energy loss in a single collision with a free electron, given by: Tmax 2mc2 2 1 2 2 2m m 1 M 1 2 M 1 (2.5) where c is the velocity of light, m the electronic mass and M the proton mass. The error obtained by setting the factor between the square bracket to 1 is smaller than 0.2% for proton energies below 250MeV. The L1 term in equation (2.3) is known as the Barkas correction and is responsible for the slightly different stopping power for positively and negatively charged particles. It does play a role at low energies, but is usually taken together with the shell correction and implicitly included into the average excitation potential I. The L2 term is kwon as the Bloch correction and only important for relativistic energies. When one deals with a compound medium, which is usually the case in protontherapy, it is possible to use the Bragg’s rule: the S value for compound can be found by averaging the S over each atomic element in the compound weighted by the fraction of electrons belonging to each element. In this way we can define effective values for Z, A and I. The average excitation potential I is the average of the excitation energies over all atomic states weighted by their transfer probability. These probabilities, which are called optical dipole oscillator strengths, are unknown for most material other than hydrogen. It is therefore easy to determine values for the average excitation potential I by fitting the stopping power formula to the experimental range-energy data. In this way also the shell correction C/Z and Barkas correction L1 can be determined simultaneously. Unfortunately there is still disagreement on which data to use. 23 For instance, the data of H.H. Andersen et al. obtained in 1967 [2.3] differ by 2% from the ones obtained in 1981 [2.4]. Although 2% is a small number, it has the consequence that the uncertainty in the dosimetry chain already begins with 2%. ICRU has therefore decided to tabulated the stopping power for the most frequently used compounds and element [2.5] in order to achieve at least consistency. A small summary is given in Table 2-1 for the material water, air (of interest for ionizing chamber dosimetry) and gadolinium (which is the most important element in scintillator screen system). In this table also the maximum energy loss in a single collision Tmax is tabulated. Table 2-1: Proton stopping power and ranges according to ICRU 49. H2O=1.00 g/cm3, air=1.20 mg/cm3 and Gd=7.90 g/cm3. IH2O=75.0 eV, Iair=85.7 eV and IGd=591 eV The range of protons R in a medium can be determined by integrating the stopping power from 0 to E: 1 E dE R dE 0 dx (2.6) This is called continuous slowing down approximation (CSDA). In practice however not all protons that start with the same energy will have the same range. This is caused by the statistical fluctuation in the energy loss process. It is possible to fit the relation between CSDA range (in g/cm2) and initial energy (in MeV) by: R=AEp (2.7) The factor A is approximately proportional to the square root of the effective atomic mass of medium. Relation (2.7) is known as the Bragg-Kleeman rule [2.6]. In Figure 2.1 a comparison between the fit using the parameters from Table 2-2 and ICRU ranges (Table 2-1) can be seen. 24 The largest difference between the fit and the ICRU ranges is 0.1 g/cm 2, that is in the order of magnitude of the uncertainty of the underlying theory. Figure 2.1: Range-energ y relationship according to ICRU 49 and fit to relation (2.7) for water, air and gadolinium using the parameters in table 2.2 A(g/cm2 MeV-p) p (dimensionless) water 2.56·10-3 1.74 air 2.73·10-3 1.75 Gadolinium 5.96·10-3 1.70 Table 2-2: Parameters for the fit of relation (2.7) to the ICRU data (Table 2-1). R is expressed in g/cm2, E inMeV. 2.1.3 Proton interaction with electrons: energy loss distribution The amount of energy loss of a proton in a medium is subject to two sources of fluctuations. The number of proton-electron collisions can fluctuate, and at the same time the energy lost in each collision varied statistically. Both distribution are characterized by a Poisson-like behavior. In most of proton-electron collisions only a small amount of energy is transferred from the proton to the electron, due to the large ratio of the proton to electron mass. There is however a small, but finite probability that a collision occurs where the energy transfer approaches T max and the atomic electron is dissociated from the atom. This extracted electron is called a -electron or -ray. Depending on the application it is possible to use a macroscopic, statistical description of the energy loss or a microscopic description, in which the -electrons are treated separately. With the statistical description, the probability of occurrence of a certain energy loss in a medium layer with thickness t as a function of mean energy loss S and proton velocity can be calculated. The macroscopic energy loss distribution function depends on the scale one is considering. In a thick layer a large number of collisions occur and the energy loss is expected to be distributed according to a Gaussian. For a small layer of material, however, the probability of a collision with an energy transfer close to Tmax occurs remains constant, but its relative contribution to the total 25 energy loss becomes much larger. This means that larger fluctuation can occur for small layers. The parameter that describes the collision regime is called the skewness parameter k, which relates the energy loss in a medium with thickness t to the maximum energy transfer Tmax (see equation (2.5)) in a single collision: k Tmax (2.8) with Z 1 Kt A 2 (2.9) The symbols are the same as in equations (2.1), (2.2). As will be explained later in equation (2.12), in the case k < 1, has the following physical meaning: when a proton passes through a medium with thickness t, there will occur, on the average, one collision with an atomic electron in which the proton loses an amount of energy greater than . For k « 1the energy distribution is described by the Landau theory [2.7].where distribution has a large tail towards high energy loss. In case k > 1 ( becomes larger than Tmax for thick media) the energy distribution approaches to a Gaussian. Vavilov has treated the energy loss distribution in a more general way, which includes the Gaussian and Landau distribution as a limiting case [2.8]. The Vavilov energy distribution function v is a function of , k and the scaled energy loss : 2 1 0.577.... ln k (2.10) where is the actual energy loss and, is the mean energy loss (St), 0.577… is Euler’s constant and the other symbols are as previous. The actual expression for v (,k) is very complicated and is easiest to evaluate numerically. Figure 2.2: The Vavilov distribution function as a function of the scaled energy loss for 200 MeV protons. 26 In the case of thick absorbers the energy loss distribution approaches a Gaussian with a given by [2.2]: E2 Tmax 1 2 2 (2.11) where is given by equation (2.9), Tmax is given by equation (2.5). The Gaussian approximation is valid if the proton energy can be assumed to be constant during the passage through the absorber. This does not hold for very thick absorber. In that case one has to divide the absorber into several smaller slabs and sum the 2E of the individual slabs. A comparison of the Vavilov distribution function for different skewness values k is shown in Figure 2.2. This calculation has been performed for 200 MeV protons (2 = 0.32) but because of the small dependence on it is also valid for low energy. To use the distribution at a different energy one has to adapt the scaled energy loss . It is interesting to apply the results of this section to a cell model. The mean energy loss for a 200 MeV proton passing through a cell is 4.5 keV, is 0.26 keV and the skewness value k for a cell is 510-4, which is clearly in the Landau region. The result is that the most probable energy loss is 2.7keV (=-0.22) but 5% of the protons deposits 8.4 keV or more. In case the energy transfer from the proton to the electron T is much larger than the mean excitation potential I, it is also possible to use a microscopic description and explicitly consider the -electrons. The number of -electrons N as a function of electron energy T per unit thickness can be calculated using the Bhabha cross section [2.9] which can be written as the product of the classical Rutherford cross section and a quantum mechanical correction for spin -½ particles (factor between square brackets): d 2N Z 1 1 T T2 K 2 2 1 2 dTdZ A T Tmax 2 E Mc 2 2 (2.12) where the symbol are the same in equations (2.1), (2.2). The difference between the classical Rutherford cross section and equation (2.12) increases for increasing proton energy and energy transfer, up to 40% for 250 MeV protons and an energy transfer of Tmax. The meaning of can also be illustrated using the integral of equation (2.12) without the quantum mechanical correction factor: this is the probability that one -electron with an energy is produced: N , t K Z 1 t A 2 1 (2.13) In Figure 2.3 the -electron spectrum per incident proton according to equation (2.12) can be seen for a number of proton energies. 27 Figure 2.3: Number of electrons produced per incident proton per cm2 H2O calculated using equation (2.12) For a number of topics such as the biophysical effect of radiation, ionization conversion factor or scintillator efficiencies it is useful to establish a threshold energy Tcut below which the energy loss is assumed to be local and above which -electrons are produced. The component of the stopping power which originates from the energy transfer smaller than Tcut is usually called the restricted stopping power. It can be calculated using the restricted Bethe-Bloch equation which has only a contribution from energy transfer below Tcut. It can be written in the form of equation (2.1) but with a different L0()-factor [2.5]: 2mc 2 2Tcut L0 ln 2 1 T 2 1 cut Tmax C 2 ln I 2 Z (2.14) The integral from Tcut to the kinematical limit Tmax of the product of the number of -electrons (equation (2.12)) and the -electrons energy T corresponds with the contribution of -electrons to the total stopping power. The choice of Tcut depends on the scale of the problem, but a usual choice is 10 keV (10 keV electrons have a CSDA range of 2.5m in water and 2.4 mm in air). This implies that for 200 MeV protons the contribution of -electrons is 22% of the total stopping power, for 100 MeV 20% and for 50 MeV 17%. The energy loss fluctuation are caused by fluctuation in the number of -electrons and by fluctuation in the -electrons energy. The energy loss fluctuation below Tcut can be described by the Vavilov-Landau theory, although the condition T » I has to be fulfilled also in that case. The energy straggling caused by the energy distribution described in this section causes also a distribution in the proton range, which is called range straggling. The range is in first approximation also distributed according to a Gaussian with a width R. The mean square fluctuation in the range R2 at the depth R depends in the following way on the mean square fluctuation in the energy R2 [2.2]: 2 d 2 dE R E dz dz dz 0 R 2 R (2.15) 28 For R2 we can use equation (2.11) which yields a complex integral. In [2.2] it is shown that the error we make by using the Bohr approximation: R2 = Tmax ( is given by equation (2.9), Tmax is given by equation (2.5)) is small for proton energies above 10 MeV. Together with the empirical range-energy relation (2.7) this results in a simple expression for R2 : Z p 2 A2 / p 32 / p R A 3 2 / p R2 K 2mc2 (2.16) where the symbols are the same in equations (2.1), (2.2), (2.7). 2.1.4 Proton interactions with nuclei: scattering A proton will experience a deflection as it passes in the neighborhood of a nucleus. This deflection is the result of the combined interaction with the Coulomb and hadronic field of the nucleus (the deflection caused by collisions with electrons can be neglected because of the mass ratio). The Coulomb interaction has been derived by Rutherford [2.2]: 2 e2 Z 2 1 2 d 2 sin d 4 16 E sin 0 2 (2.17) where E is the proton kinetic energy (in MeV) and d() the differential cross section in mb (≡ 10-31 m2), the other symbols are in equations (2.1), (2.2). This cross section decreases very rapidly with increasing scattering angle, and with increasing energy. The consequence is that most particles are only slightly deflected. Hadronic interactions will only play a role when the distance between the proton and the nucleus becomes very small (≈ diameter of the nucleus ≈ 1.3 10-13 A1/3 m). Although the mechanism of hadronic interactions is very complicated, it manifest itself experimentally with a deviation from a Rutherford scattering cross section, which can be parametrized using optical models [2.10]. In Figure 2.4 the ratio between the Rutherford cross section and the observed elastic cross section (so including hadronic interaction) can be seen as a function of angle in case of 180 MeV protons on 16O. The elastic cross section has been calculated using the DWUCK1 code with parameter for oxygen from [2.10]. In addition to this elastic hadronic interaction, also inelastic interactions occur, which are described in the next section. 1 DWUCK4 version 10/10/81 by P.D.Kunz et al. 29 Figure 2.4: Solid line (scale on right axis): ratio between total elastic cross section of 180 MeV protons incident on 16O and Coulomb contribution (using equation (2.17)) as a function of deflection angle. Dotted line: ratio = 1. Dashed line (scale on left axis): Coulomb contribution. Only for large scattering angles (> 4°) the difference becomes significant. The probability that a scattering occurs with an angle more than 4° in 1 cm H2O is less then 0.05% (this can be calculated with equation (2.17)). So in practice only the small angle deflections, which are caused by distant collisions, contribute. However because of the large number of interactions the total effect is considerable. A multiple scattering theory has been derived by Molière [2.11] and lateral on improved by Bethe [2.2] which is valid for scattering angle 30. Analog to the energy distribution function it is possible to derive macroscopically an angular deflection distribution function. According to Molière this distribution function can be expressed as a series expansion which involves complicated functions. The limiting case for many collisions again is a Gaussian distribution: plane 2 d f d exp 2 02 2 0 1 (2.18) rms For the calculation of the width 0 (the mean squared angle projected on a plane plane , which is rms ) of the Gaussian several approximations exist. In [2.12] is shown experimentally that 1/ 2 space the Highland formula gives the best results for protons: 0 14.1MeV 2 E Mc 2 t t 1 1 log 10 LR 9 LR rad (2.19) where t is the thickness of the medium, E the proton kinetic energy , Mc2 the proton rest mass (expressed in MeV) and LR the radiation length of the material, which is the distance over which the electron energy is reduced to a factor 1/e due to bremsstrahlung only (tabulated in [2.12], for water LR = 36.1 g/cm2). Although the radiation length is a material property derived for electrons, this formula turns out to fit the experimental data with protons well [2.13]. For small angles we can approximate 2space ≈ (2plane,x + 2plane,y). The deflection in x and y direction are independent and identical distributed. The Highland formula works under the assumption that the proton kinetic energy remains constant during the passage, which means that the thickness t has to be small. In the 30 case of thick absorber, it is possible to apply the equation (2.19) to a thin slab and in an analog way as for the energy straggling by taking the sum of the 20 of the individual slabs. Because of the artificial use of the radiation length however is it necessary to remove the factor [1+1/9log10(t/LR)] from the integral, thus to treating it as a correction factor which depends on the entire target thickness [2.13]. 2.1.5 Proton interactions with nuclei: nuclear reactions Because the energy of the proton in radiotherapy applications is much higher than the Coulomb barrier, protons have a probability to react with the nucleus other than by elastic or inelastic scattering. This causes a decrease of the proton flux with depth, already long before the end of the proton range. When we assume the CSDA approximation (see equation (2.6)), this can be described by: N av E0 dE ' 0 exp E ' inelas A S E ' eff E (2.20) where is the flux of protons with energy E, 0 the initial flux, E0 the starting energy, S(E) the stopping power and inelas(E) the inelastic nuclear reaction cross section. For radiotherapy applications (where water is the reference material) we are mainly interested in the cross section with oxygen because the reaction cross section with hydrogen is negligible [2.14]. The 16O cross section has been measured by Carlsson et al. [2.15]and by Renberg et al. [2.16]. Figure 2.5 shows the experimental points together with a fit that has been made by Seltzer [2.17] using the GNASHcode [2.18]. Figure 2.5: Total nonelastic nuclear cross section for proton incident data 16 O. The line represent a fit to the experimental With respect to the reaction products the situation is more complicated. The secondary particles can be neutrons, protons and recoiled fragments. The energy transferred to the recoil fragments will be deposited locally, but secondary protons can travel a considerable distance before stopping. The secondary neutron will either escape from the medium or produce another reaction, in which tertiary particles can be produced. In general is not possible to make an analytical calculation of the contribution of nuclear reactions to the energy deposition as a function of depth. Berger [2.17] uses 31 the estimate that 60% of the initial proton energy (i.e. before the reaction) is deposited locally while other 40% escapes from the medium in the form of neutrons and ’s. Monte Carlo calculations, however, in which the secondary particles are separately followed are not in agreement with this and show that a depth dependent contribution is deposited. The result for the flux reduction of the primary protons using equation (2.20) can be seen in Figure 2.6. The steepness of the flux reduction is slightly larger for the 80 MeV proton beam, due to the slightly higher cross section. Figure 2.6: Proton flux reduction due to inelastic nuclear reaction for a 80 and a 180 MeV beam in a water medium calculated using equation (2.21) and the cross sections in figure 2.5. 2.1.6 Proton dose distribution When we summarize the results of the previous sections, it is possible to calculate the proton dose (energy lost in the medium) as a function of position. We start with the dose dependence in the direction parallel to the beam. We define the energy fluence at depth z by: (z) = (z)E(z) (2.21) where is the proton flux, and E the proton kinetic energy as a function of the depth z. The energy loss by protons as a function of depth is then given by: D z 1 d 1 dE z d z z C E z dz dz dz (2.22) The first term describes the energy lost to the atomic electrons (see section 2.1.2, 2.1.3). In first approximation we assume that this energy is deposited locally (this applies when the medium is water, for air this is not the case). The second term describes the energy lost by the flux reduction due to the nuclear interactions. Here the assumption of local deposition is no longer valid. The best thing we can do analytically is to use the approximation that a fraction C of 0.6 is deposited locally, while the remaining energy escape in the form of neutrons and ’s. Using this assumption, the energy loss D(z) in equation (2.22) is equal to the deposited dose. 32 Using the empirical relation between E and R given in relation (2.7), it is possible to simplify the calculation considerably. The energy E(z) at a depth z is determinate by the residual range R0 - z which the protons traverse before stopping: 1 ( R0 z )1 / p 1/ p A (2.23) dE 1 1 / p ( R0 z )1 / p 1 dz pA (2.24) E( z) From this we can determinate dE/dz: With respect to the flux reduction due to nuclear reactions, we can use equation (2.20), but in [2.12] it is shown that the error results from ignoring the energy dependence of the cross section inelas and using a straight line fit: 0 1 B( R0 z ) 1 BR0 (2.25) is small compared to the uncertainty on the position of energy deposition of secondary particles (see also Figure 2.6). Therefore: d B 0 dz 1 BR 0 (2.26) where B is a medium dependent parameter and 0 the initial flux. In order to incorporate the range straggling caused by the energy straggling in the medium and the initially spread of the proton beam , the relevant distribution functions have to be folded into equation (2.22). Bortfeld has shown [2.19] that is possible to do so, with a limited number of simplifications to the proton energy spectra: for the range straggling he uses a Bohr approximation (2.16) and for the initial beam spread the sum of a Gaussian and a small ‘tail’ extended towards low energies. The result is a rather complicated expression: 1 B e / 4 1 / p (1 / p) D1 / p ( ) C D1 / p 1 ( ) 1/ p R0 2 pA (1 BR0 ) p 2 D( z ) 0 (2.27) where D is the parabolic cylinder function, is the gamma function (both tabulated in [2.20]) and is (R0-z)/ where is the total width of the range straggling. The parameter describes the low energy tail of the initial proton spectrum. The tail contains a small fraction of the initial flux, but is influenced by the way a proton beam is produced. In the calculation has to be set to zero. The total width consist of a term 2mono caused by the energy straggling in the medium and a term 2E0 caused by the initial beam spread. The total can be calculated with: 2 2 2 mono dR 2 0 mono E20 A 2 p 2 E0( 2 p 2 ) dE 0 2 Eo (2.28) 33 mono is the range straggling that is present for a monoenergetic beam and can be calculated using equation (2.16). The other, medium depended parameters are determined by equation (2.7), (2.22), (2.26). A summary of the parameters is given in Table 2-3 together with the values in case the medium is water. p A R0 B C mono E0 description exponent of range-energy relation oroportionality factor range slope parameter of flux reduction fraction of energy released locally width of range straggling initial beam width → beam dependent contribution to tail → beam dependent value for water 1.74 2.56·10-3 AEp0 0.012 0.6 0.012R00.935 ≈ 0.01E0 ≈ 0.0-0.2 Unit 1 cm MeV-p cm cm-1 1 cm MeV 1 Table 2-3: The parameter which determine equation (2.27) togheter with their numerical values in case of water medium. The last two parameters are beam dependent Although expression (2.27) it is easy to evaluate numerically. In Figure 2.7 we see dependence of the height of the Bragg peak as a function of the initial energy spread E0 both for a 80 and a 175 MeV proton beam. Also the position of the Bragg peak changes. The definition for the term ‘Bragg peak’ is the position where the maximal dose occurs. In Figure 2.7 and subsequent figure we plot the dose per flux instead of dose, which means that to obtain a dose this value has to be multiplied with the number of protons per cm2. The range R0 for 180 MeV protons is 21.64 cm. It can be seen in Figure 2.7 that this equals the depth at which the dose has dropped at 80±1% of this maximum value behind the Bragg pick, independent of the initial energy spread. Figure 2.7: The dose per fluence as a function of depth in water calculated with (2.28) for 80 and 180 MeV proton beams, with an initial energy spread s increasing from 0% to 1.5% in steps of 0.25%. Entrance dose per fluence is for a 180 MeV beam 5.78 MeV cm2/g and for a 80 MeV beam 9.32 MeV cm2/g. R0 is shown as dashed line. With respect to the dose distribution in the lateral direction we can make use of the Gaussian approximation of the multiple scattering theory as described in section 2.1.4. The total width of the proton beam is given by the quadratic sum of the width caused by the initial divergence of the proton beam and the (depth dependent) broadening caused by the multiple scattering. The result for a parallel beam can see in Figure 2.8. The multiple scattering increases in the Bragg peak, due to the 34 1/E dependence of the scattering angle. For two or more dimensional problems it is in general not possible to use analytical tools. To calculate the distribution in Figure 2.8 it has therefore used a numerical method: the Monte Carlo code PTRAN [2.17]. Figure 2.8: Contour plot of dose of a 160 MeV, 2 cm radius parallel proton beam in water. The contour lines go from 1% to 99% of the maximum dose. At the zero depth and on the beam axis the dose is 21% ofht edose in the Bragg peak. 2.1.7 Fragmentation In the previous sections of this chapter we have described the interactions of protons with matter, while in the first chapter we have presented the medical advantages of light ions for radiotherapy treatments (i.e. an enhancement of the Relative Biological Effectiveness and a reduction of the Oxygen Enhancement Ratio). However, with respect to the protons, the light ions present the unfavourable effect of fragmentation. This diminishes the number of primary ions delivered to the volume under treatment or investigation, and produces a tail of damaging ionisation beyond the Bragg peak. Carbon ions play a special role in therapy. They produce the best physical dose distribution because of the decrease of longitudinal and lateral scattering compared to the protons, and the relatively small fragmentation compared to heavier ions as neon. In addition, a small fragmentation rate into +-emitting carbon isotopes, allows one to detect the stopping carbon beam inside the target via Positron Emission Tomography technique. Fragmentation reactions have been studied theoretically and experimentally since many years. However there is a still paucity of experimental data, especially for light system and for energy ranges below ~ 100MeV/u, and their difference with theoretical models is sometimes quite large [2.21]. 35 Figure 2.9: Bragg curve of 270 MeV/u carbon beam in water illustrating the effects of the beam fragmentation. The colour-code lines are calculated dose distribution: red=total dose, black=primary particles, blue=secondary particles, gree=fragemnts of the secondary particles. Circles indicate measured data. 36 2.2 Accelerators for hadrontherapy 2.2.1 Introduction Accelerators are present in the field of hadrontherapy since their early history and most of the development has taken place at accelerator facilities for physics research. Modern accelerator technology exists today to meet all of the clinical requirements within a reasonable budget for hospital-based hadronterapy facilities. We will describe the fundamental characteristics of the two kinds of accelerator commonly used in hadrontherapy: the cyclotron and the synchrotron [2.22]. 2.2.2 Cyclotron Since its early history in Berkeley, the cyclotron has been use for medical purpose. Cyclotrons are good candidates for at least four aspects. Energy The choice of a fixed energy makes the design simple. The magnet can be optimized and trimming coils are not necessary. The accelerator is at fixed radio frequency and all the setting of the beam lines are fixed. Magnet In order to reduce the dimension of the cyclotron, a high magnetic field should be chosen. This choice has two important consequences: 1. It is impossible to accelerate negatively charged ions at high energies in a high magnetic field due to electromagnetic stripping (the MEDICYC [2.23], high field compact cyclotron in Nice accelerates H- ions up to 65 MeV, very close to the limit of stripping). 2. Superconducting magnet are attractive because they are lighter than room temperature magnets (about half of the weight) and the running cost are lower. Intensities An isochronous cyclotron provides a continuous beam (CW) with ample intensity. This is a key factor for reliability. The beam intensity is easily controlled. The beam can be very stable and this is an important advantage for a dynamic beam spreading system. To fully benefit from this advantages, an external injection system makes the cyclotron very flexible. For example, on the MEDICYC cyclotron which uses an axial injection at 33 keV, a feedback control of the extracted beam intensity is realized using a few volts on the bunching cavity. A beam stability higher than 1% could be easily be achieved. Operation No sophisticated controls are needed. A simple programmable logic controller is sufficient. The operation requires less complexity and this reduces the manpower costs. 37 2.2.3 Synchrotron Energy A synchrotron produces pulsed beams at variable energies. The energy of the extracted beam can be varied from one cycle to the next in steps of few MeV. Hence, the necessary modulation of the Bragg pick to scan the target volume in depth can be achieved without absorber. Magnet The weight of the different magnets (dipoles, quadrupoles) is low. This facilitates the transport and the beam assembly. No particular basements are needed in the building. Intensity The required intensity could be achieved by a synchrotron fed by an adequate injector. Nevertheless, the traditional resonant extraction system is subject to small perturbations due to small variations in the excitation of all magnets which induce a time modulation of the extracted beam intensity. Operation Rapid magnetic field variations are requested, and the operation is more complex. Moreover, undesired intensity variations require careful controls. Nevertheless, many sophisticated synchrotrons are running smoothly for high-energy physics reaching a high reliability level. Figure 2.10: IBA 235 MeV room temperature cyclotron. 2.2.4 Example of dedicated designs Cyclotron There is now an increasing number of project for high-energy protontherapy aiming at treating many tumour types at any depth. These machines could use either a room temperature or a superconducting magnet. The characteristics of a few designs are presented below. 38 A room temperature design A design by the Ion Beam Application company has chosen a high field (2,15 T at extraction) produced by conventional coils (Figure 2.10). The cyclotron and the associated gantry has been installed in 1997 in a protontherapy facility at the Massachusetts General Hospital in Boston, USA. An other important proton facility which uses a cyclotron is that one at Loma Linda (California): 9282 patients have been treated since 1990 to February 2004 [2.32]. Synchrotron H-synchrotron. The TERA collaboration in Italy proposed a multitask synchrotron [2.24] for its hadrontherapy Centre which would be able to provide: 250 MeV proton beams by accelerating H-, hence simplifying the extraction but requiring a low magnetic field and a very good vacuum in order to avoid electromagnetic stripping of the H- ions. The drawback is a relative large diameter. The concept was proposed by ITEP in Moscow [2.25], and by R. L. Martin [2.26] in the USA. Light ion beam up to 16O8+ with energy up to 400 MeV/nucleon (Figure 2.11). Figure 2.11: Layout of the TERA synchrotron At HIMAC facility at Chiba/Japan since 1994, until February 2004, 1976 patients have been treated, using a synchrotron to accelerate different kind of ions up to a maximum clinical energy of 800 MeV per nucleon [2.32]. An other facility is under construction since February 2004 at Hildelberg. 39 2.3 Beam spreading The method used to spread the beam, transversely and longitudinally, defines the relationship between clinical specifications and accelerator performance. We consider two extremes: passive spreading by scatters and active spreading or “scanning” by magnets. For a given field size, depth and dose rate, passive spreading requires more energy and somewhat more current. On the other hand it is relative simple. Scanning requires considerably less energy and slightly less current. It permit better conformation of the dose to the target shape and always provides the sharpest possible distal falloff. On the other hand it is significantly more complicated, requires far more sophisticated beam diagnostic, and places serious demands on the time structure of the beam and the accelerator control system [2.27]. 2.3.1 Passive spreading The reference passive system, using double scattering, is shown in Figure 2.12. The fixed absorber and the modulator taken together comprise the first scatterer. Each consists of high-Z and low-Z (e.g. lead and plastic) sandwich. Such a sandwich allows one to obtain any desired energy loss combined with any desired scattering angle (over a wide range). Thus the fixed absorber can be used to tune the first scattering angle and at the same time to make fine adjustments in the range. The same principle apply to the modulator allows scattering to be independent of modulator step. Figure 2.12: schematic layout of a passive beam spreading system consisting of a double scatterer, a range modulator, and a snout housing the patient-specific collimator and bolus. When used with a fixed energy beam, the first scatterer also acts as energy absorber. The range modulator is not required when full energy is achieved from the accelerator. The position of the absorber and modulator in the beam line differs from the traditional one where they are much nearer to the patient. By coupling scattering with modulation this makes the system more difficult to design, but offers two important advantages. The modulator can be far smaller, making it easy to incorporate to the gantry. Even more important, by avoiding the angular confusion by downstream degraders, it gives a sharper dose distribution (less penumbra) after the patientspecific aperture. The penumbra can be nearly as good as the physical limit due to scattering in the patient. The second scatterer is a contoured lead foil compensated with plastic so that the total energy loss is independent of radius. It is a good idea to provide several basic field size, small, medium and 40 large, each with its own nozzle, for each beam line or gantry. This avoids to compromise the dose rate for small targets by using an unnecessarily large beam. It also minimize neutron production and, by reducing mechanical interface, allows one to bring the aperture and the bolus closer to the patient. Changing field sizes requires to change second scatterers, but each modulator works for each field size, since the first scatterer strength can be adjusted by means of the fixed absorber sandwich. A passive system similar to the one just described is in use at the Harvard Cyclotron Laboratory. 2.3.2 Scanning Figure 2.13: Schematic irradiation procedure for a tumour conform treatment. The tumour is dessected in slices of equidistant particle ranges which are covered using a magnetic scanning system. Two strategies of active magnetic beam scanning have been proposed, the raster and the pixel scan [2.28]. Both techniques are based on the virtual dissection of the tumour in slices of equidistant particle ranges (Figure 2.13) but the modes to cover these slices are different [2.29]: Pixel scan In the pixel scan mode the complete scanning area is covered by a mesh of distinct points which are irradiated separately. For each point the magnet are adjusted while the beam is turned off. Then the beam is turned on and the desired number of particles for that point is delivered. Then the beam is switched off again and the magnets are adjusted for the next spot. As the treatment for a slice is intermittent, a fast beam switch is needed to turn the beam on and of very frequently (≈ 103-104 beam positions in each slice). Therefore the treatment time is mainly limited by the time to switch the beam on and off and the time needed to reach a stable magnetic field. However, this technique can be used without any modifications for the treatment of rectangular as well as irregular shaped areas. Raster scan In the raster scan mode, the beam is moved continuously in a preselected pattern over the target area and a well defined number of particles is delivered in each line element. Because the beam will not turned on and off the writing velocity of the beam has to be controlled by the intensity of the incoming particle beam. At high beam intensity and for low covering rates, the beam has to be moved faster in order to spread the incoming intensity over a larger area. For low intensity and higher covering rates, the beam will be scanned slowly to achieve the required number of particles 41 for each line element. In order to guarantee a precise control of the scan velocity, beam fluctuations must be detected by the intensity monitor and transmitted to the control system. In the raster scan technique, the shape of the treatment areas influences the pattern of the pathway of the beam. Irradiating rectangular areas the beam has to be swept in a zig-zag mode. The rectangular area is the ideal (and the basic) shape of the target area because the beam can be scanned along a pattern of parallel guide lines each yielding a homogeneous covering. A frequently found prejudice is that a zig-zag scan mode would yield a inhomogeneous dose distribution at the turning point of the irradiated field. This is not correct because using the intensity controlled raster scan technique there is no difference in proceeding from one beam position to another with or without a change in the scanning direction. In order to irradiate irregular shaped areas the path of the beam in this area has to be optimized. At the edge of an irregular field additional scan lines have to be introduced (see Figure 2.14). On the technical level there is no fundamental difference between the two scan strategies. Raster scan is conceptually more difficult because a proper scan path has to be selected and optimized, but, in contrast to the pixel scan mode, the intensity-controlled raster scan method does not need to switch the current on and off during the exposure of a slice. Therefore the treatment time of each slice is limited only by the beam intensity and by the maximum speed of the scanned beam that depend on the ramp rate of the power supplies and the distance between the magnets and the target area. A fast beam turn off is necessary when the irradiation of a slice is finished but also in case of malfunction of the scanning system. Figure 2.14: Pixel (right) and raster (left) scan pattern for a rectangular area and a model of a tumour slice. Additional scan lines that shortcut the standard (rectangular) scan pattern have to be introduced. 42 2.4 Monitoring of hadrontherapy beam The basic issue of the safety system is to protect the patient against any possible failure. In such a complex machinery as hadron accelerators many components have to be controlled for a correct beam delivery. For the safety system only components that could cause an irradiation at a false position or a false intensity of the beam are important. Malfunctions that cut off the beam (like vacuum problems) are not risky but may appear as warning signal. The inhibitive signals of the safety system are mainly created by the beam diagnostic at the treatment area. There the beam is monitored in a non-destructive way downstream of the spreading system (passive spreading or scanning) and its localization, shape and intensity are measured just before it enters the patient [2.27]. 2.4.1 Passive beam spreading systems Using a passive beam spreading system the dimensions of the field could be up to 20x20 cm2. The main difference between the active and passive scanning system is that in the first case the slices of the tumour situated at different depth are painted with a few millimeters wide pencil beam. In the other case the whole axial section of the tumour has to be covered by the beam during the irradiation. This leads the beam used with passing scanning system to be wide enough to match the whole target. Moreover the use of the bolus is possible under the condition of a flat beam fluence distribution. The main aims of a monitor are thus the measurement of the delivered dose using the fluence measurement and, at the same time, the check of the flatness of the beam fluence distribution. In order to guarantee a correct dose delivery to the patient, is in fact necessary that the shape of the beam does not change during the irradiation time. 2.4.2 Active scanning systems Active scanning systems are more flexible than the passive one because they allow a very good conformation of the dose distribution to the target volume and because no individual beam shaping elements have to be manufactured for each patient. On the other side, because of their complexity they require faster on-line monitoring instruments than those used in a passive beam spreading. As for the passive scanning systems one of the parameters that has to be checked during patient irradiation is the dose delivered to the patient. The detector has thus to be used like a beam intensity monitor. We have just seen that with beam spreading active techniques the target volume is dissected in slices of equal particle range and each slice is painted using a small pencil beam having a diameter of few millimeters only. Once the dose distribution inside the target is decided, the beam movement with time and its energy are computed using a Treatment Planning System (TPS). The on-line verification of the TPS with the beam measured position is the other aim of the control system. The detector moreover has to be used as a position sensitive monitor that covers the full swept beam. The detector, for on-line dose measurement, needs a large dynamic range and linearity for highintensity radiation. On the other hand, it needs a good spatial resolution, comparable with the voxel dimension as from typical computer tomography (CT) scanners. 43 2.4.3 Development of segmented anode ionization chambers For the purposes just described, within a collaboration between INFN-section of Torino and the University of Torino, different kinds of segmented anodes ionization chambers have been developed. We have developed and build two different kind of anode segmentations, pixel and strip, to have the possibility of use similar chambers in different conditions and with different aims [2.30]. The pixel chamber and its applications are described in this section, while the strip chamber will be presented in the next. The pixel chamber The design of the prototype has been focused on the following aims: use of a completely established technology rather than trying further solutions; keep the amount traversed by the beam as small as possible; ensure a large degree of modularity during the protopying phase to have all possibilities open with minor modifications. In Figure 2.15, which shows an exploded view of the chamber, one can see the two planar electrodes which are mounted on square frames made of Vetronite [3.6]. The cathode is a 25 m thick aluminized mylar foil. The thickness of the Aluminum layer is about 0.2 m. The anode is made of a 100m thick Vetronite foil, sandwiched between two layers of copper, 35 m each. By using the standard printed circuited board (PCB) technique we obtained 32 x 32 conductive pixels on one side and tracks (one for each pixel) on the other side. The connection between a pixel and the appropriate track has been obtained through a connective hole. The tracks bring the signals from the pixels to the connectors located at the edge of the foil. There are 256 connections one each of the four sides of the foil. The pixel dimension is (7.45 x 7.45) mm2, with a pitch in both directions of 7.5 mm. Thus the electrically isolated interspace between adjacent pixels is 0.1 mm, which is only determined by the technique used to engrave the copper layer. The total covered area is (240 x 240)mm2, which matches the typical maximum treatment field dimensions. While increasing the pixel dimension is technically feasible, going to smaller dimension presents a limit determined by the number of tracks one can fit between adjacent pixels. Figure 2.15: An exploded view of the pixel chamber 44 To avoid charge build up and leakage currents, the anode foil around the (240 x 240) mm 2 active area is covered by a layer of copper and polarized to same voltage of the pixels (anode potential). The mounting of the electrode foils on the Vetronite frames, which assure rigidity and mechanical strength, is rather delicate. The foils are stretched by using a larger frame. The goal is to obtain a planar surface capable of supporting a limited tension, as on a drum. The stretching operation is performed by applying an increasing tension to the foil as one moves around the frame in small steps of ≈ 5 cm at a time. Once the tension of the foil is satisfactory, the foil is glued to the chamber frame. The flatness and, even more, the stability of the electrode surfaces are of paramount importance for the performance of the chamber. In fact, the gain associated with each pixel depends on the thickness of the gas gap. Ideally one would require a pixel-to-pixel variation of the gas gap in the few percent range. Such a variation can be accounted for by means of an appropriate calibration, provided it remains stable. To check the stability of the gap, we measured the flatness of the electrode planes during a one and half year period. The average variation was 50 m. As stated above, 1024 signals have to be carried out of the anode foil and this is done in four groups of 256 signals. Four 68-pine half pitch connectors are located on each side of the anode plane. To avoid loss of charge collection efficiency do to ion recombination (relevant for irradiation at very high local ionization), it might be necessary to fill the chamber with a gas like nitrogen or carbon dioxide, but replacing the air and keeping the gas flowing requires an overpressure. Though the pressure can be limited to a few centimeters equivalent of water, nevertheless the combination of the rather large area and thin foils can result a considerable change of the gas gap. To avoid this effect, we added two blank chambers, one on each side of the active chamber, where the gas flows at the same pressure. The blank chamber has been made with a frame supporting a 25 mm thick aluminized mylar foil. In order to reduce the induced electrical noise the foils can be set to ground. A gas fitting on each of the blank chambers provides the in and the out gaslet. Finally the gas path through the active chamber was constrained via grooves and holes machined on the frames. The sandwich of frames, three for the active chamber and five when the two extra blank cambers are added, is geometrically constrained by four reference pins and bolted together. The anode frame holds a PCB card on each side, while the four cards house the front-end electronics. The overall dimensions perpendicular to the beam direction are ≈ 650 x 650 mm2, whilst the actual dimension along the beam direction is ≈ 20 mm. The water equivalent thickness of the material seen by the beam is about 0.6 mm (0.06 g/cm2). The chamber is mounted on an Aluminum frame to keep it in the upright or horizontal position according to the beam direction. The Aluminum frame is also used to install plexiglass sheets in front and behind the chamber, where most the elements described can be recognized. With the present prototype design there are several degrees of freedom, which can be summarized as follows: the gas gap can easily be increased by inserting a frame of the appropriate thickness; the choice of pixel dimensions can easily be changed; this requires making the anode foil with the PCB standard technique; by adding two blank chambers, one can run with gases other than air, which may be necessary in a very high ionization environment. The electric field is generated by polarizing the cathode mylar foil. Typically the high voltage value has been set to -500 V. The pixel voltage, as discussed in the next section, has been set to the same voltage as the front-end input circuit, around 2 V. To filter high frequencies picked up from the high voltage cable, we added a passive low-pass filter with a double-pole at ≈ 2 Hz. 45 The strip chamber The mechanical structure of the strip chambers is approximately the same already described for the pixel chamber: the electrodes (anode and cathode) are glued on a side of a machined fiberglass frame, and are placed one in front of the other. The dimension and number of the strips that compose the anode change in function of the use of the detector. Until now the number of strips varied in a range between 32 and 256, and their width between 500 m and 4 mm. Also the thickness of the anode foil and the material with which it is done depend on the application of the detectors. In the chapter five we will present in detail the description of a strip detector developed within a collaboration between the INFN-Torino and LNS-Catania. 2.4.4 Electronic read-out: the VLSI chip The signal from each pixel or strip is brought to a connector which is plugged into a card servicing the front-end electronics. Four connectors are hosted by the same card. In the pixel chamber there is one card on each side of the anode foil, while in the strip chamber only one card per anode is needed. The design of the front-end is based on the recycling integrator architecture [2.31]. Briefly, the input current from each pixel or strip is integrated and a number of pulses proportional to the charge collected by an individual pixel or strip is sent to a synchronous 16-bit counter. The maximum counter frequency is 5 MHz. An ensemble of 64 channels have been integrated in a single VLSI chip. At any given time all the counters can be latched and the results stored. This is like taking a snapshot of the charges collected by all the channel (pixel or strip). In what follows the relevant features of the front-end are summarized. The charge corresponding to a single pulse (charge quantum) can be adjusted between 100 and 800 fC via an externally regulated voltage, Vp. In fact the charge is generated with loading a 200fF capacitor, which is then discharged on the integrator itself. The maximum pulse frequency is 5 MHz, which results in a limit for the pixel input current of 4 A for a charge quantum set at 800 fC. The linearity from 10 pA up to the maximum current is compatible with the charge quantum better than 1%. The pedestal is at the level of a few Hz. The charge quantum spread is 1% at 600 fC and 1.5% at 100 fC. The polarization voltage of the chip is +5V, with the integrator input (and consequently the pixels or strips) at ≈ 2 V. The current from a given strip is integrated on a capacitor C I. When the voltage across the capacitor exceeds a give threshold a charge quantum is sent to the integrator input and subtracted from the capacitor CI. The full cycle can continue without any interruption, and this leads the read out system to have no dead time. 46 Figure 2.16: Logic diagram of the TERA chip The bus from the front-end cards to the data acquisition includes, besides the 16 data lines, 10 address lines to multiplex all 1024 pixels. Three more lines are furthermore necessary: an analog reset to discharge the integrator capacitor at the beginning of the operations, a digital reset to clear all the counters, and a latch to store them. We used two flat 34-lead cables to carry the digital signals and standard differential RS-422 (differential TTL) as the transmission protocol. 47 3 The GSI test: experimental setup and results 3.1 The GSI facility in Darmstadt In 1994 a joint project was started by the Radiologische Universitätsklink Heidelberg, the Deutsche Krebsforschungszentrum Heidelberg (DKFZ), the Gesellschaft für Schwerionenforschung Darmasdt (GSI) and the Forschungszernrum Rossendorf to build an experimental therapy unit for ion beams at the accelerator complex of GSI. Major achievement of this Pilot Project were the world-wide first realization of a 3D tumour-conformal beam delivery system based on the scanning technique (which we have described in the previous section) as well as the in situ verification of the beam by application of positron emission tomography (PET). A further highlight of the project has been the development of a biological optimized treatment planning which explicitly takes into account the higher biological effectiveness of the applied therapy beams. Following four years of construction and preparation, in December 1997, two patients were treated with carbon beams. These were the first patient treatments with ion beams in Europe ever. Until the end of 2002 160 patients were treated with carbon beams. The tumours treated were located mainly in the base of the skull and in the spinal region [3.1]. Figure 3.1: The GSI accelerator facility 48 Figure 3.1 shows a global view of the GSI accelerator facility with the grey-coded path representing the accelerator chain involved in radiotherapy. 3.2 The experimental setup In April 2002 we have carried out a test at the GSI carbon ion beam with a prototype of the pixel chamber, in order to study its performances using the detector as on-line beam monitor with the voxel scanning technique. In this section we will present the experimental setup of the test, while in the followings we will describe the dedicated data acquisition system and, finally, the discussion of the results [3.6]. The beam line and the instrumentations, used for cancer treatments, are located in the Medical Cave. We shall first briefly describe the operation principles of a treatment that have been developed by the GSI group [3.2]. Physical properties of the volume to be irradiated are known from a Computed Tomography scan, which measures the density of each voxel corresponding to a volume of (1.12 x 1.12 x 3.02(z)) mm3. The beam is aimed to each voxel (or group of voxels) to deliver the required biological dose, sparing the healthy surrounding tissue as much as possible. The beam line is normally equipped with a set of detectors that are used to guarantee both the beam fluence and position [3.3]. The beam line setup is sketched in Figure 3.2 and it includes: two multi-wire proportional chambers (MW1 and MW2), made of two planes each, measure the beam position in the plane transverse to the beam direction; three planar ionization chambers (IC1, IC2 and IC3) monitor the beam fluence. These chambers are sandwiched between MW1 and MW2. Figure 3.2: Sketc of the beam line set-up This ensemble of detectors gives the appropriate feedback to the accelerator control system to move the beam to the next voxel once the fluence required by TPS has been achieved. Furthermore, MW1 and MW2 check the consistency of the beam position with respect to the required ones. 49 For this test the pixel chamber was installed ≈ 36 cm downstream of MW2. We used N2 gas at a flow of 5 l/h with the cathode voltage set to -500 V, corresponding to an electric field of 1670 V/cm. The pedestals were regularly checked during the data taking. Pedestal measurements were made with the beam off by reading the whole chamber at a frequency of a few Hz for several seconds. The average counting rate for most of the pixels was below 1 Hz. For a limited number of pixels the rate went up to 10 Hz. Pedestals have been taken into account by subtracting the appropriate amount to the readout according to the readout frequency. 3.3 The data acquisition system The data acquisition system, which is sketched in Figure 3.3, is based on the VME bus with Motorola 2301 CPU, running VxWorks [3.4]. 3.3.1 The operating system VxWorks and the integrated environment TornadoII Any operating system where interrupts are guaranteed to be handled within a certain specified maximum time, thereby making it suitable for control of hardware in embedded systems and other time-critical applications, is defined real-time operating system. This means that, for example, UNIX and Windows hosts are system for program development and for many interactive applications, but they are not appropriate for real-time applications, where the operation-time of a task has to be shorter than a maximum allowed delay. VxWorks is the WindRiver® real-time operating system that we have used to write the data acquisition program for this specific test. With this application two complementary and cooperating operating systems are utilized: VxWorks and UNIX or VxWorks and Windows. VxWorks handles the critical real-time chores, while the host machine is used for program development and for applications that are not time-critical. Tornado is an integrated environment for software cross-development. It provides a way to develop real-time and embedded applications with minimal intrusion in the target system. Tornado comprise the following elements: VxWorks, a real time operating system; application-building tools (compilers and associated programs); a development environment that facilitates management and building projects, establishing and managing host-target communication, and running, debugging and monitoring VxWorks applications The Tornado interactive development includes: the launcher, an integrated target management utility; a project management facility; integrated C and C++ compilers and make; the browser, a collection of visualization aids to monitor the target system; CrossWind, a graphically enhanced source level debugger; WindSh, a C-language command shell that controls the target; an integrated version of the VxWorks target simulator, VxSim; an integrated version of the WindView logic analyser for the target simulator. Tornado facilities execute primarily on a host system, with shared access to a host-based dynamic linker and symbol table for a remote target system. 50 With Tornado it is possible to use the cross-development host to manage project files, to edit, compile, link and store real-time code, to configure the VxWorks operating system, as well as to run and debug code on the target while under host-system control. 3.3.2 The data acquisition system: hardware In Figure 3.3 the sketch of the data acquisition system is shown, based on the VME bus with Motorola MVME2031 CPU and running VxWorks. Figure 3.3: A sketch of the data acquisition Address Generator and Controller The connection between the CPU and the front-end cards is achieved via a custom developed Address Generator and Controller (AGC) and a commercial Firs-In-First-Out (FIFO) module. Indeed, for many applications where the transverse dimension of the beam is a small fraction (a few square centimeters) of the chamber area, one can limit the read out to a box confined around the beam. Given the coordinate of a specific box corner and the box dimensions, the AGC generates the addresses of the pixels to be read out. The results are stored in the FIFO and then transmitted with the DMA (Direct Memory Access) protocol to the CPU to determine the position and width, and the flux delivered. As said in section 2.4.4 we use the 42-lead cables to carry the digital signals and standard differential RS-422 (differential TTL) as the transmission protocol for the connection between the pixel chamber and the control modules. The frequency with which the AGC sends the addresses to the pixel chamber is 10 MHz. We can see in Figure 3.3 that two more modules are connected with the CPU through the VME bus to receive the signals from the accelerator control system. 51 Input Output Register The first is an Input Output Register, module that is able to read ECL and NIM levels, and NIM pulses. We used it to monitor the acquisition stop signal (manually generated with an external module) and spill on-off signal (given by the SIS control system). Both signals have to be NIM pulses and have to be sent in input to the Input Output Register. Scaler The second module is a Scaler. In section 2.3.2 we have presented the active scanning technique: the tumour is dissected in slices, each one at different depth and so reached by a different energy beam. The minimal element volume that composes the target is the voxel, and the painting of a single slice is obtained by moving the beam from one voxel to another. Moreover, the SIS cycle was set to ≈ 4 seconds, with a spill length of ≈ 2 seconds. We have thus used the Scaler to count the number of voxels, slices, spills and the irradiation time during a treatment. For the time measurement we sent pulses, generated from a Dual Timer module and set with a period of 1 s, to the Scaler. 3.3.3 The detector read-out As previously stated, the goal of this test was to use the detector as an on-line beam monitor, verify the Treatment Planning System (TPS) and controll the beam during the dose delivery. To reach this aim we have decided to proceed via two methods, according to the application. Passive acquisition method With the first method one can extract the position sequence from the TPS and generate the information necessary to the AGC. The system needs to be fully integrated in the accelerator control system: every time the beam spot moves from one voxel to the next, the pixel address list needs to be updated according to the TPS. This is done by translating the expected beam position to the chamber coordinate system and selecting the appropriate pixel matrix to read. The first part of the data acquisition program is thus dedicated to the reading of the TPS file, in order to store the coordinate of the voxels shot by the beam. Subsequently the program enters in a stop or spill waiting condition. With the stop signal set true the program execution ends, creating a data file, while, with the spill signal true, it enters is an acquisition loop. Each of these is composed of two different parts: the first is the generation and transfer to the detector of the addresses around the struck voxel, by the AGC module. The second part is the acquired data transfer, via DMA, from the FIFO to the CPU for analysis: the computing of the center of gravity of the beam and its comparison with the expected position, as given by the TPS. During the spill the data acquired are temporarily stored on the CPU RAM, but in the interspill period are then transferred to the host PC where the environment TornadoII is active. The data is conveyed by an ethernet connection with a frequency of approximately 10 Mbit per second. As a result of an interspill ≈ 2 seconds the amount of data that can be acquired during the spill is 2 Mbyte. Active acquisition method With the second method the read out system acts in a standalone mode: the beam, which is sweeping across the voxels, is followed automatically by the read out. This operation mode requires the following phases: 1. fast online computation of the actual beam position 2. determination of the pixel map which fully contains the beam. 52 In contrast to the first method in this case the addresses of the pixels to be acquired are obtained by computing the last acquisition data, and are not read from the TPS file. Regardless of the read out method used, the speed of the operation is of paramount importance. A full read out cycle (1024 pixels) takes 500 s, of which ≈ 300 s are due to the overhead of the start-up phase. The readout speed per channel is 200 ns. On the other hand, the time budget necessary to read a (7 x 7) matrix is reduced to 20 s. 3.4 Results 3.4.1 Charge collection efficiency We checked the charge collection efficiency at the chosen operating conditions to be independent of the beam intensity. In Figure 3.4(a) we show the counts summed over a (5 x 5) box as a function of the average number of ions per spill. The intensity spans over two orders of magnitude from 106 to 1.5 x 108 ions per spill which is the typical beam intensity range used at GSI for treatments. The beam intensity was cross-checked using the GSI system ionization chamber IC3. The beam energy was set to 200.3 MeV/u. In Figure 3.4(b) we plot the relative deviation of each measurement with respect to a fitted straight line. There is no evidence of any loss of efficiency with the beam intensity. Furthermore, the apparently large deviations of the first two points are in fact compatible with the rounding errors. Figure 3.4: (a) Chamer response and (b) relative deviation from linearity as a function of the beam intensity 53 3.4.2 Calibration The gain of a given pixel is proportional to the gas gap seen by the pixel and to the inverse of the charge quantum of the integrator channel serving it. From laboratory tests we found that the electronics gain of each channel had a spread (r.m.s.) less than 1%. To measure the relative gain it is necessary to irradiate the chamber with a field as uniform as possible. This was part of the standard procedure at the Medical Cave, and it was achieved by steering the beam over a (14 x 14) cm2 area in 1.7 mm steps. The profiles of the beam had a similar shape in both X and Y directions, and the beam full-width half-maximum (FWHM) was set to 7.1 mm. For each point of the scan, a fixed amount of carbon ions, 9 x 106, was delivered. This corresponds to a uniform flux of 108 ions/cm2. The average beam intensity was kept at ≈ 2 x 108 ions/spill and the beam energy was set to 200.3 MeV/u. With this procedure, the uniformity has been independently measured to be better than 1%. The maximum field size achievable covered approximately half of the chamber surface; furthermore, the uniformity on the last two rows/columns irradiated was found to be poor. Ultimately, we calibrated only ≈ 30% of the total chamber area used in the test. Figure 3.5: (a) Lego plot of the raw counts as a function of the pixel position. (b) Normalized distribution of the gains. In Figure 3.5(a), we show a lego plot of the raw count per pixel as a function of the pixel position. The fraction of the chamber which has been irradiated is visible as a prominent flat top. The surrounding part, characterized by very low counts, represents quite well the typical electronic noise (pedestal). The distribution of the raw counts is shown in Figure 3.5(b) where the average over the ≈300 central pixels has been normalized to unity. The r.m.s. of the distribution is found to be ≈3.2%, due to the convolution of two contributions: the spread of the beam and the variation of the pixel gains. The calibration constant of a pixel (i,j), C, is given by C(i,j) = 1/G(i,j), where G(i,j), is the normalized gain. Raw counts are thus multiplied by the calibration constants before being used in the analysis. By applying to subsequent runs the calibration constants, derived from a previous reference data taking, we could monitor the gain shift during the data taking period. In Figure 3.6, we show a lego plot of the corrected counts as a function of the pixel position (a) for a typical run. The distribution of the corrected gains is reported in Figure 3.6(b). The r.m.s. of the distribution of the corrected 54 gains is 0.8%, which includes the spread due to the beam reproducibility. This result has been confirmed by several measurements during the data-taking period. Figure 3.6:(a) Lego plot of the corrected counts as a function of the pixel position. (b) Distribution of the gains. 3.4.3 Space coordinates resolution with a steady beam The reconstruction of the beam position is of primary importance to monitor the agreement between the TPS and its actual implementation. Let us first consider the position resolution for a fixed beam. The beam was aimed at the center of the chamber which, for this purpose, was read out asynchronously every 100 ms. The space coordinates were measured at each read out by computing the center of gravity (c.o.g.) of the pixel counts over a 7 x 7 pixel box. By comparing the space coordinates resulting from several measurements it is straightforward to estimate the resolution. Due to short-time beam intensity fluctuations, one can determine the measured position accuracy as a function of the delivered number of ions. Results are shown in Figure 3.7, where we plot x and y coordinate resolution as a function of number of ions. As expected, the resolution improves as the collected charge used for the measurement increases. The difference between x and y resolutions is mainly due to the drifting of the beam during the spill which was much larger along y than along x. The beam position behavior during the spill is shown in Figure 3.8, where the center of gravity (c.o.g.) is plotted for a set of 10 spills. Each dot represents a measurement in a 100 ms interval. The curves, in this case, show the precision of the pixel chamber system. We discuss now the measurement of the beam profile. This is a parameter of the beam which needs to be controlled continuously during the treatment. The beam profile can be assumed to have a Gaussian profile along both x and y directions. The dimensions were altered in a range between 4.3 and 10.1 mm FWHM. Data were taken for five beam dimensions. The second moment of the signals is strongly correlated to the beam width. In Figure 3.9, we show the second moments as a function of the beam width FWHM. The beam width depends on the accelerator set up, and the values have been checked with the GSI set up. The values have been corrected to account for the beam position with respect to a pixel. If the beam is centered within a pixel, the second moment is expected to be smaller than when the beam impinges close to a pixel border. The corrections, which are limited to ±20%, have been evaluated with a simulation. 55 Figure 3.7: Measured position resolution as a function of the integrated number of ions: (a) x coordinate; (b) y coordinate. Figure 3.8: Beam position along x (a) and y (b) as measured in time slices of 100 ms during the spill. 56 Figure 3.9: Corrected second moment vs. beam width (mm). 3.4.4 Space coordinates and fluence measurements with a scanning beam We determined the position resolution with a scanning beam in the same beam conditions we used for the pixel gain determination. The beam spot, set to 7.1 mm FWHM, scanned the chamber in 1.7 mm steps corresponding to ≈ 25% of the beam size. As stated earlier, all beam conditions were controlled by the independent GSI system. The chamber was read out in synchronous mode at each voxel step, roughly every 10 ms. The operation allowed the determination of coordinates and fluencies of single points. Results are plotted in Figure 3.10 where in (a) we show a pictorial map of the scan, which covers approximately (14x 14) cm2 area. The distribution of the difference between the coordinates as measured by the pixel chamber and by the GSI system are reported in (c) and (d), respectively for x and y. The resolution is ≈ 0.2 mm. We remark that the figure is the convolution of the measurement errors of the two systems, and we did not try to separate the contribution of each component. The result of the fluence measurement is shown in Figure 3.10(b): the number of ions as measured by the pixel chamber is compared to the GSI system measurement. Both measurements are in arbitrary units and are shown after the ratio has been normalized. The resolution is 1.8%, which translates into a fluence of ≈ 1.6 x 105 ions, is completely negligible when compared to the number of ions delivered for a treatment in a single voxel. 57 Figure 3.10: (a) Pictorial map of the scanning points; (b) normalized ratio between the number of ions as measured by the pixel chamber and GSI system; (c) x and (d) y resolution. 3.4.5 Performance with a Treatment Planning System (TPS) In this section we discuss the results obtained with the pixel chamber when used as the beam monitor for the GSI Dose Delivery System (DDS). The DDS follows the directives of the TPS [3.5], which, in the case of the voxel-scan method, can be considered as an ensemble of several thousand elementary manipulations of the beam, each characterized by a predefined set of parameters: direction, energy, dimensions, and fluence. Each set is meant for a specific voxel belonging to the target volume. The TPS is designed to deliver a uniform equivalent dose to the target volume, sparing as much as possible the surrounding tissue. A further task of the TPS is to organize the voxel-scan to minimize the elapsed time required for the treatment. As the GSI DDS is implemented, the beam changes directions while the energy remains constant within a spill. The necessary energy changes are made in between spills. By assuming constant beam intensity during the spill the fluence is proportional only to the time spent aiming at a voxel. The beam walk from one voxel to the next is made without switching off the beam. To check the performance of the pixel chamber under the above conditions we synchronized the read out with the voxel scan (passive method). Specifically, at the end of an elementary treatment, counters were latched and the read out of a 7 x 7 pixel box centered at the beam nominal position as given by the treatment plan was performed. In total the box was read out every 200 s and at the end of an elementary treatment. We remark that for both methods there is no dead time associated with this operation because the counters were free to count continuously. 58 We checked the chamber with several TPS, but we report here the results for only two cases, both concerning patient treatments, referred to a small volume tumor (≈ 10 cm3) and a medium volume tumor (≈100 cm3). Both tumors were seated at a ≈ 5 cm depth. Measurements were performed independently with MW1, MW2 which determined the beam coordinates in the transverse plane at each spot, and I1, I2, and I3, which measured the fluences. In Figure 3.11, we show the scatter plots of the x–y measurements of four typical slices. Each slice is composed of several voxels, the number depending on the area of the slice itself. The shape of the slice matches the tumor at that particular depth. Each slice is a collection of different beam fluencies and dimensions at the same energy. Figure 3.11: X-Y scatter plot of four slice at different depths. The distributions of the difference between the measured coordinate and that prescribed by the treatment plan are reported in Figure 3.12 for the x (a) and y coordinate (b). Results from all 27 slices have been added. There is no evidence of effects due to the beam dimensions, which were allowed to vary between 6.1 and 6.5 mm FWHM. The position resolution is ≈ 0.5 mm both in x and y coordinates. This result was confirmed by a simulation which accounted for the pixel and beam geometry to predict the counts in each pixel. Similar position resolutions were obtained in the case of a medium volume tumor. To study the detector performance for the flux measurement we compare the counts summed over the 7 x 7 box to the measurements obtained by the GSI system with a full area parallel plate ionization chamber. The intermediate read out operations were done in any case to monitor performance in terms of speed. In Figure 3.13(a) we show the correlation between flux measurements of the pixel detector read out and of the GSI system for the small volume treatment. Each entry in the scatter plot is the result of an elementary treatment. The points on the high side of the plot are related to the slice of voxels farthest along the beam line. Those voxels are treated with a larger flux, since the upstream slices receive a consistent fraction of dose during the irradiation of the farthest voxels. 59 The ratio between the two measurements, normalized to unity, is reported in Figure 3.13(b). The r.m.s. of the Gaussian curve fitted to the distribution is ≈ 6:6%; which accounts for the measurement errors of both systems. We remark that the above figure is dominated by the contribution of the small flux component. Figure 3.12: Coordinate measurements for a small volume tumour treatment: resolution along x (a) and y (b). Figure 3.13: (a) Pixel chamber vs. GSI flux measurement and (b) the normalized ratio distribution. 60 Conclusions In this chapter and in the previous one we have described the design, construction and tests of a parallel plate ionization chamber with the anode segmented in (32 x 32) square pixels; the dimension of each pixel is (7.5 x 7.5) mm2, with a covered total area of (24 x 24) cm2. The total thickness traversed by the beam is 0.6 g/mm2. The gas gap was set to 3 mm and the high voltage to –500 V. The resulting electric field was about 1.7 kV/cm. The readout of the 1024 pixels is performed individually with a recycling integrator circuit followed by a 16-bit counter. 64 channels are integrated in a VLSI chip allowing a compact detector. The design of the chamber and the data acquisition system have been finalized to use the detector as a beam monitor for therapeutical hadron treatments. Using this kind of therapy in fact, the beam, which can be kept quite small in the dimension transverse to the beam direction, is continuously moved across the target profile (voxelscan technique). The beam intensity can be modulated to deliver a adequately uniform dose distribution on the target volume as prescribed by the physician. Both the verification of the Treatment Planning System and the beam control during the actual dose delivery are more challenging with respect to conventional radiotherapy. The GSI test was done to verify that the detector (chamber and data acquisition system) satisfy the spatial and timing requirements, in order to be used as beam monitor on line with the treatment. The DAQ was set in two different configurations: passive and active acquisition. In both cases a box of 7 x 7 pixels was read every 200 s and at the end of each voxel treatment. In the first method the position sequence was extracted from the TPS and, with the aim to generate the information necessary to acquire the pixels hit by the beam. The system needs to be fully integrated in the accelerator control system. In the second method the beam, which is sweeping across the voxels, is followed automatically by the read out. With both DAQ configurations we obtained the following results for the beam measurements: a position resolution of ≈ 0.5 mm a flux resolution of 1.8% for fluxes ≈ 107 carbon ions. Using this kind of scanning technique the beam is moved in separate “steps” covering the whole target area; the volume defined by a single step is called voxel, and, at GSI, its dimension is (1.2 x 1.2 x 3.02(z)) mm3. A resolution of 0.5 mm in the beam position measurement and a read out frequency of 5 kHz (200 s of acquisition period) are sufficient to use the pixel chamber and the DAQ system as beam monitor on line with the treatment. A time of 200 s is thus need to compare the TPS and the data acquired, and, in case of error, to stop or correct the irradiation. 61 4 The IBA test: experimental setup and results In this chapter we will discuss the test carried out in October 2002 at the CRC (Centre de Recherches du Cyclotron) in Louvain-la-Neuve. This test was done within a collaboration between the IBA (Ion Beam Application) and INFN-Torino, with the goal to verify the operation of the magnetic scanning system, designed and built by the IBA. In particular the first aim was to check the performances of the low regulation loop of the SMPS (the X-Y beam scanning control system) and of the ISPS (the beam current control system, subsection 4.2.1) and the eventual need of a highlevel regulation loop. A pair of strip chambers and a pixel chamber were thus mounted close to nozzle of the cyclotron, after the scanning system magnets. The collected data were analysed to study the performance of the SMPS and the ISPS. 4.1 The CRC facility at Louvain-la-Neuve The CRC operates three different cyclotrons to be used by national and foreign experimental groups. Its unique character, in Europe and in the world, resides in the production of very intense and pure Radioactive Ion Beams (RIB) in the low-energy region (0.2 to 10 MeV per nucleon). These beams are most suitable for direct nuclear reaction cross section measurements in the field of Nuclear Astrophysics and for exotic nuclei Nuclear Physics experiments. Beams of 6He, 7Be, 10C, 11 C, 13N, 15O, 18F, 18Ne, 19Ne, 35Ar are routinely available and new beams of 14O are under development. Other beams available include light and heavy ions (from gaseous and solid elements) and monokinetic fast neutron beams in the 20 to 80 MeV range. Experimental facilities include a large scattering chamber, a LEvel Mixing Spectrometer (LEMS), solid state detector arrays (LEDA), the 4 neutron detector DEMON, the Leuven Isotope Separator On-Line LISOL and the Astrophysics REcoil Separator (ARES). A well characterised neutron therapy beam for dosimetry and radiobiology experiments requiring high LET is available. Very high intensity neutron, light ion and heavy ion irradiation facilities for radiation effects studies are also available [4.1]. CYCLONE110 CYClotron de Louvain-la-Neuve: was built in the early seventies by Thompson CSF in collaboration with ACEC. It is used for nuclear physics, isotope production and medical and technological applications. CYCLONE is a multiparticle, variable energy, isochronous cyclotron capable of accelerating protons up to 80 MeV, deuterons up to 55 MeV, alpha particles up to 110 MeV and heavier ions up to an energy of 110 Q²/M MeV (where Q is the charge and M the mass of the ion). The energy range for heavy ions extends from 0.6 to 27.5 MeV/AMU depending, among other things, on the ion's charge state. CYCLONE30 It was designed and built by the CRC team in the period 1984 - 1987. It is a fixed frequency cyclotron, capable of accelerating H - particles up to an energy of 30 MeV. After the acceleration process these particles are stripped through a carbon foil, and the resulting protons are extracted and transported to the target. The use of this technique allows production of large intensities (500 µA) at variable energy (from 15 to 30 MeV). CYCLONE 30 has a low power consumption (less than 90 kW for a 15 kW beam) and is fully automated in order to be usable by personnel without long 62 cyclotron experience. This cyclotron, primarily designed for industrial and medical applications, is used for isotope production, for Positron Emission Tomography equipment and as production accelerator for the radioactive beams. CYCLONE44 It has been designed for the acceleration of stable and radioactive isotopes mainly for the study of nuclear reactions of astrophysical interest. The design and construction of Cyclone 44 has been undertaken by CRC team in the frame of a Interuniversity Attraction Pole (PAI) between three of the major universities in Belgium (UCLouvain, ULBruxelles, KULeuven). The first RIB ( 19N3+) was accelerated in December 1999. 4.2 The Pencil Beam Scanning (PBS) The main characteristics of IBA’s PBS can be summarized in the following points: continuous magnetic scan in the x and y direction; modulation of both the scanning speed and of the beam current; multiple painting of the tumour slabs; energy changed at cyclotron exit; variable beam spot size from a minimum of 2.5 – 4 mm RMS (230 – 80 MeV) to a maximum of 1 cm RMS at isocenter in air; on-line monitoring by a system of pixel, integral and strip ionization chambers; fast acquisition with the strip and the integral chambers, at the end of each pixel painting. In the following subsection two main structures of the PBS will be briefly described: the beam spot scanning regulation (SMPS system), and the beam current regulation (ISEU system). 4.2.1 Low Level Regulation The SMPS Low Level regulation The pencil beam is magnetically scanned in X and Y directions by the scanning magnet power system (SMPS) regulation. A fast magnet and a slow magnet perform the scanning independently in the two directions. Two fast IGBT’s PWM inverters power the scanning magnet, with magnet voltage as output. The two measured variables are the magnet voltage and current. The regulation of the magnet system is build up with two nesting loops: the inner loop that retroacts the measured magnet voltage and the outer loop that controls the measured magnet current. The inner loop regulator has a proportional-integrator structure, the outer loop regulator has a proportional action. In this model is possible to assume that the voltage variable is proportional to the beam speed and the magnet current variable is proportional to the beam position. The ion beam Low Level Regulation The Ion System Electronic Unit (ISEU) is in charge of the regulation of the proton beam current intensity. A digital predictive regulator has been designed for that purpose and the performances of this system are well known. 63 The IBA team have done several tests at the CRC in order to verify the performances of this system. The ion source at the CRC does not present the same characteristics to a previous one installed in Boston. In Boston the performances of the ion source are well known, and the ISEU control unit has been designed on the basis of this source. During the tests at the CRC, we have observed the following differences: A lot of noise was present in the measure (see Figure 4.1), mainly due to mass problems and due the instability of the beam source. It was difficult to obtain a stable lookup table for the ion source regulation. The objective of the lookup table is to linearize the process, therefore the regulation of the ion source was not optimal during the tests. This is not the case in the ion source process in Boston. As we can see in Figure 4.1, the response of the beam current is not the one expected. A detailed observation of this response shows that the pure delay observed is around 30s and not 60s as observed in the ion source of Boston. This difference explains the non conform behavior observed during the tests. Nevertheless the IBA team expects that this regulation associated with the magnet regulation will give the necessary performances for the dose deposition. Figure 4.1: Typical ISEU response observed during the experimentation at the CRC and the disagreement between the expected reference and the measured beam current response. 64 4.3 The experimental setup 4.3.1 The beam line setup The setup conditions of this test are very similar to those of the GSI test, which have been described in the previous chapter. With respect to GSI, the scanning technique is not voxel but raster. With the voxel techinque the irradiation of a slice of the target is done by moving the beam step by step from one voxel to an other. With the raster one the same area is “painted” by the beam in a continuous way, without stopping in a single voxel. There is a main scanning direction (vertical in our case), which leads the slice to be covered by parallel irradiation lines. When the beam reaches the end of a single line, it is then moved toward the other scanning direction, to cover another line close to the first. The modulation of the released dose is obtained with a variation of the scanning speed. With respect to the GSI test we used two strip chambers besides a pixel chamber. Moreover, instead of using a VME CPU running VxWorks as operating system we decided to use a digital 32 bit input/output National Instrument PCI board mounted on a common PC. In Figure 4.2 the sketch of the beam line setup is shown. The first detector crossed by the beam is the pixel chamber; there is then the integral plane (an ionization chamber with a non segmented anode) followed by one of the strip chambers mounted with the strips vertical (respect to the ground), while in the last one the strips are horizontal. There is no free space between the three detectors, so the distance between any two anodes is given by the thickness of the frames (6 mm each). Figure 4.2: Sketc of the beam line setup We ran dry air gas inside the sensitive volume with the cathode voltage of each detector set to -500 V, corresponding to an electric field of 1670 V/cm. The pedestals were regularly checked during the data taking. As described in the previous chapter the pedestal measurements were made with the beam off by reading the whole chamber at a frequency of a few Hz for several seconds. The average counting rate for most of the pixels – strips was below 1 Hz. For a limited number of counters the rate went up to 10 Hz. Pedestals have been taken into account by subtracting the appropriate amount to the readout according to the readout frequency. The strip chambers used for this test had the anode segmented in 64 strips, each one is 3.86 mm wide. The anode foil is composed by a 100 m kapton thickness layer covered by 35 m of cooper. 65 As said before, the machine used to accelerate protons is a cyclotron, with fixed beam energy of 75 MeV. The complete PBS test bench is composed of quadripoles, scanning magnets and a vacuum chamber. The beam RMS is variable from 3 to 10 mm at isocenter, and the scanning speed is variable up to 20 m/s in y and 2 m/s in x. Figure 4.3: Picture of the IBA nozzle and strip and pixel chambers 4.3.2 The structure of the data acquisition The DAQ system that we used for the beam tests is based on a PCI DAQ card National Instruments DIO 32 HS and LabVIEW software. DAQ is performed with a PC. The DAQ card is a digital input/output card with 32 lines that can be configured independently in groups of 8 lines. We actually use 16 input lines and 16 output lines, which fits our requirements (14 control signals and 16 data signals). To perform the DAQ of the chamber we first have to configure the DAQ card using the appropriate LabVIEW functions. These are the steps followed: Configure the digital input output group (group 0 and 1 are set for input, 2 and 3 for output); the VI used are WriteGroupConfig.vi for the output and ReadGroupConfig.vi for the input. A memory (RAM or FIFO) is located on board. In our configuration we use this memory both to store the control to send to the detector and to receive the data coming from the detector. We have thus to configure the input and output buffers following these steps: o Allocate the memory for the input/output buffers o Configure the internal clock to time the output buffer operations o Fill the output buffer with the sequence of control signals The buffer we used is a circular buffer (the sequence of control signals is created, loaded and then sent in a infinite loop). The VI used are WriteBuffConf.vi for the output and ReadBufferConfig.vi for the input. 66 After the definition of the space on this memory for the output-input operations, the sequence of controls has to be sent to the output buffer. The VI used is WriteFillBuffer.vi. Start the input-output operation enabling first the input buffer to receive data, and then the output buffer to send data in circular configuration. The VI used are StartRead.vi and StartWrite.vi. Finally, the input buffer has to be read in a circular way. From this memory the data has to be moved on the PC RAM where can be computed on-line with the acquisition, and at the end stored on a file on the hard disk. The VI used to read data is ReadBuffer.vi. Theoretically, the DAQ card can acquire data at a maximum rate of 10 MHz. This means that the time interval between the sending of two consecutive words from the output buffer is 100 ns, and, in case of a write buffer of 2000 words, the acquisition rate (the time between two consecutive latch signals) is 200 s. In practice we have never used such a high rate, because at this frequency the system used is not able to read, compute and store data. The following acquisition periods were used in this test: 10 ms with both strip and pixel detectors 1 ms with pixel detector alone 90 s with strip detectors (both vertical and horizontal) 67 4.4 Results During this test three different kinds of scanning beam were used: 1. Scans with fixed beam, to check the stability of the beam shape at different beam widths. 2. Scans with irradiation of a homogeneous rectangle, with constant scanning speed, to obtain scans that can simply give information about the raster scanning. 3. Scans obtained from a real treatment plan: both the scanning speed and beam current are variable. The main analysis that we will present is done using data acquired with the two strip detectors alone (90 s of acquisition period) during the irradiation of homogenous rectangles and in case of fixed beam. We computed the spatial resolution of the scanning beam positioning to check the regulation loop of the SMPS, while the analysis of the collected charge gives information about the regulation of the proton beam current (ISEU, see subsections 4.2.1). The spatial analysis is done comparing the data acquired with the treatment plan input data (called trajectories file). 4.4.1 Time synchronization between PIXEL/STRIP data taking and IBA control system To be able to compare data acquired with the detectors and data set on the scanning control system the first step is to synchronize the start time of every data set. In the Figure 4.4 are plotted, both versus time, the center of gravity of the horizontal strip chamber (black points, Y position of the beam) and the same component of the data expected position (red points, from trajectories files). The whole analysis described in this subsection refers to scans of a homogenous rectangle, acquired with the strip chambers. Figure 4.4: Y position of Center of gravity for strip chambers (red) and data expected from trajectories files (black) On the X axis the time is expressed in seconds and is obtained multiplying the sequential number of acquisition by the period (100 microseconds for trajectories and 90 microseconds for strip 68 chambers). On the Y axis the value of horizontal scanning magnet current (red points) is given, while, for the strip data, the center of gravity calculated in mm and then normalized with the value of trajectories are plotted; obviously we have assumed the relation between the physical position of the beam and the magnet current value to be linear. We will discuss in detail this relation in the subsection 4.4.4. To synchronize the two curves we have to find a common start point; this means that we have to find a value of time shift between one curve and the other. At first we tried to calculate this parameter from plots like that shown in Figure 4.4, normalizing to the first zero value of the two curves on Y axis. In this way we have found two different kinds of problems. The first is that the strip chambers are not correctly aligned with the center of the accelerator nozzle. The second is that we are not always sure, and in fact this is a kind of test, that a specific value of expected magnet current, as zero in this case, correspond to the same physical point of the beam on the chambers surface in different runs. The second way used to determinate the time shift consist in considering the switching on of the beam current. The Figure 4.5 shows the strip counting (green curve), the ISEU intensity (red curve) and the expected beam current (black curve), in the time range between 0.003 seconds and 0.01 seconds. In this case we have synchronized our curve with the ISEU, normalizing the point in which the two curves exceed a fixed threshold. For the strip we have set a threshold of 80 counts. The time shift value is thus obtained setting a threshold on the strip counts and on the beam current intensity. Figure 4.5: ISEU intensity (red curve), expected intensity (black curve), strip counting (green curve, threshold of 80 counts). With this parameter it’s possible to make a real comparison between the data in trajectories file and the data acquired with the strip and the pixel chambers. The Figure 4.6 shows the same curves presented in Figure 4.4, but after the computation of the time shift for a correct synchronization between the two kinds of data set. 69 Figure 4.6: Y position of center of gravity for strip chambers (black) and data expected from trajectories files (red), with a correct time synchronization,at the beginning (top plot) and at the end (bottom plot) of the run As we said before the strip acquisition period is fixed at 90 μs, while the declared trajectories period is 100 μs. After a test aimed at verifying the correct functioning of the internal clock of the PCI board used for the detectors data acquisition, we have supposed that the increasing time discrepancy between the two curves is due to a trajectories period smaller than the planned value. To be able to make a real comparison we have calculated a time correction parameter from the analysis of a group of ten scans. At first we measured the time between the first and the last but one scanning line, at the point with ordinate zero, for both the black and the red lines in Figure 4.6. The time correction parameter is simply obtained by the ratio between these two values (strip versus trajectories). The mean value is 0.99082 with a sigma of 0.00172 (uncertainly of 0.17%). The following results are obtained multiplying the trajectories data timescale by this value. Figure 4.7: Y position of center of gravity for strip chambers (black) and data expected from trajectories files (red), with a correct time synchronization and the correction of the trajectories data period. 70 4.4.2 Homogeneity of the dose distribution Detector gain The first test on the behaviour of the proton beam current control system (ISEU) was aimed to check how the two strip chambers responded to the same beam irradiation. Figure 4.8 shows the ratio of the counts integrated over all the strips for the two detectors, plotted versus the acquisition time. Figure 4.8: Horizontal - vertical strip counting ratio The detectors show different gains for different beam intensities (at the end of each scanning line when the intensity decreases the ratio changes), and for different irradiation positions on the detector surface. The greater variations in the gain ratio correspond to the switching off of beam current: in these cases the big discrepancy is in fact due to the very low amount of counts (see Figure 4.8). Because a threshold on the total acquired counts will be set, only the acquisitions done in “beam on” condition will be used to check the beam position and fluence, when the gain variations between the two detectors is less than 1%. It’s nevertheless reasonable to assume that if the beam hits the same strip in two different points, with the same intensity and for the same time, the count might be slightly different. In Figure 4.9 the count ratios of four scans are plotted in a matrix whose cells correspond to the strip intersections. The red areas represent the zones with lower number of counts, as the extremes of the scans and the parts outside the irradiated area. It is possible to see a trend in the gain from left to right, for all the plots. The reasons could be that one of the chambers has a different gain in a particular zone of the sensible area, or that, when the irradiation starts, the beam intensity is lower than in the rest of the scan. 71 Anyway, with this kind of data, we are not able to make a calibration of the detectors, mainly because the instability of the beam intensity is too high. Figure 4.9: Ratio of horizontal to vertical strip count for four different scans Measured intensity Afirst analysis on counts is made on the scans with a beam of constant expected intensity. In Figure 4.10 the counts for each acquisition as a function of the acquisition number are plotted. Also in this case the threshold is set to 80. It’s possible to see two different bands, one between 250 and 350 counts, and the other between 100 and 150 counts. Figure 4.10: Sum of the counts over all the strips for a single acquisition versus the acquisition number. 72 Making the comparison with the expected beam current (see Figure 4.11) it is clear that the lower bands correspond to periods when the current intensity increases and decreases, at the extremes of the scans. To verify the agreement of acquired data with expected data we have performed two different kinds of analysis on each scan. Figure 4.11: A single scan line. Ibeam expected (black curve), ISEU intensity (red curve), strip counts (green curve). The first step to compute the flatness of the dose delivered is to exclude the tails from the single scanning lines. We decided in fact to use only the 60 % of their length. After computing the center of gravity in X and Y directions, for every acquisition, the center of the lines, and thus the part to be analyzed, are determined. We are thus sure to have enough data for the statistical analysis and to be in a zone where the beam intensity is expected to be flat. In Figure 4.12 mean and sigma of the acquisition counts for the scanning lines are reported. Considering all the points acquired in the zone with a flat delivered energy distribution, the counts dispersion around the mean value is ± 6%. This value represents the mean variation in the dose delivered between points (90 s of acquisition). We have also computed the variation among the mean counts calculated on the single scanning lines and found it to be ≈ 0.8%. Because of the large number of acquisitions, it is possible to consider 0.8% as measure of the uniformity of the integral dose released in the whole treatment. The difference between these two values (6% of point to point variation and 0.8% of line to line variation) is in fact due to the regulation loop of the beam current that uses as feedback the information of the integral plane (see subsection 4.3.1). 73 Figure 4.12: Vertical strip detector: mean and sigma of the counts of each acquisition when only the central part of each scanning line is considered. Each plot contains one point per each one of the 30 vertical lines of the scan. 4.4.3 Dimension of the irradiated field In this subsection the dimension of the irradiated field, and the length of the scanning lines in particular, will be discussed. The expected shape of the field is, in these cases, a rectangle. We have set a threshold, on both tails of the scanning lines, fixed at 50% of the mean counts values, and calculated the start and the end points for every scan line. In Figure 4.13 these points, obtained from the last two scans analyzed in Figure 4.9, are plotted. Figure 4.13: Center of gravity measured with horizontal strip chamber for the “edge” of the field. The dimensions of the fields for a group of 9 scans are reported in the Table 4-1. Xstart is the mean value of the centers of gravity, computed on the vertical strip detector, for all the acquisition 74 points of the first scanning line. Xend has been calculated in the same way considering the last scanning line of the run; ΔX is the difference. Ysup and Yinf are the mean values of the end and the start points of the scanning lines and ΔY represents the mean length of the lines. #scan Xstart(mm) Xend(mm) ΔX(mm) Ysup(mm) 0 -42.2±0.3 70.8±0.2 117 45.8±6.1 1 -46.2±0.2 70.2±0.1 117.6 39.9±4.1 2 -47.4±0.2 69.9±0.2 117 38.5±1.9 3 -47.1±0.2 70.6±0.3 117.3 42.5±3.6 4 -46.7±0.3 70.4±0.3 116.1 42.7±2.1 5 -45.7±0.2 71.5±0.3 117.4 45.9±2.6 6 -45.9±0.3 71.1±0.2 117 42.9±4.5 7 -44.7±0.2 71.9±0.3 119.1 42.2±4.8 8 -47.2±0.2 70±0.2 117.2 36.1±9.4 Yinf(mm) ΔY(mm) -31.3±6.6 77.1 -40.9±4.5 80.8 -42.4±2.5 80.9 -38.8±3.7 81.3 -39.9±1.6 82.6 -41.3±3.2 87.2 -49.9±4.5 92.8 -41.7±5.9 83.9 -51.7±8.3 87.8 Table 4-1: Dimension of the irradiated field for 10 runs The data in the table show a large discrepancy among the different runs and give information about the poor reproducibility of the scanning control system. 4.4.4 Detailed spatial behavior of the scanning beam. To obtain the position of the beam during every acquisition (90 s) we computed the center of gravity on the strip counts, both for vertical (X component) and horizontal (Y component) strip detectors. To skip the calculation of this parameter when the current beam is switched off we set a threshold on the counts integrated over all the strips of one chamber (the value was 80 counts). The number of strips used for the computation was nine, four to the right and four to the left of the strip with the maximum count. The choice of using nine strips is due to the dimension of the beam. Moreover 90 s is a fine period large enough to disregard the movement of the beam between two consecutive acquisitions. Horizontal scanning To analyze the stability of the scanning system in the vertical direction we computed the variation of the horizontal component of the center of gravity (vertical strip detector) line per line. In Figure 4.14 it is reported the distribution of the differences between the X center of gravity, calculated for a single scanning line, and the mean value of that line. 75 Figure 4.14: Distribution of X centers of gravity computed for each single scanning line. The next step is the verification of the distance between the scanning lines in comparison with the expected values. The equations used to obtain the positions of the beam on the detector surface from the trajectories files data are the following: Xexp_surf=1182*tg(Xtraj) Yexp_surf=1610*tg(Ytraj) where 1182 and 1610 are the distances, in mm, from the detector to the scanning magnet; these two values are obtained minimizing the residual distribution obtained with experimental data. Xtraj and Ytraj are the values from trajectories files, which contain theoretical (expected) and experimentally measured values. In Figure 4.15 the mean values for each single scanning line of the X positions expected, measured by IBA and measured by the strip detector are plotted. The horizontal components of the computed centre of gravity are shifted of 7 mm for a correct alignment. It is clear that the distance between the scanning lines measured by the strip chamber and IBA is different from the expected values. The irradiated field is thus larger than the set value of about 6 mm, that becomes 8.5 mm at isocenter considering a contraction factor of 1.43. 76 Figure 4.15: Mean X values per scanning line for strip (blue), trajectories expected (gree) and measured (red) on the detector surface. Vertical scanning A similar analysis has been carried out for the vertical direction, using the data collected with the horizontal strip detector. The main problem is that the beam scanning is along the Y axis; this, with a not so high time synchronization precision, makes it difficult to find a correct way to obtain a value for the resolution of the vertical beam coordinate. To make a point to point comparison possible between experimental and set data we made a linear fit of the Y position expected of trajectories data for every scanning line, plotted versus time. In this way we obtain an analytical expression of the movement of the beam in vertical direction versus time. The comparison is thus made between the data obtained from the horizontal strip detector and the value computed from the fit, using the time of the data acquisition. It is important to mention that the fit has been made on the values expected on the detector surface and not on the trajectories data directly. Figure 4.16 shows the distribution of the differences between the collected and expected values. The sigma is about 165 μm. The value of the sigma computed from all the acquisition points in the run is around 190μm. This value for the resolution is not completely correct, because is possible to reduce the sigma for a single scanning line with an optimisation of the time and space parameters. Unfortunately these parameters vary for different scanning lines; so to the same expected position values do not correspond always the same physical points hit by the beam. 77 Figure 4.16: Distribution of Y components of the center of gravity on the expected values for a single scanning line. 4.4.5 Scanning speed We will present now some results about the scanning speed analysis for data taken at constant values of speed. For the computation of this parameter we have used only the centres of gravity calculated with the horizontal strip chamber (vertical positions), and only in the highest intensity part of the scanning lines, excluding in this way the tails. Moreover, it is easy to see that the theoretical scanning speed is constant only in this central part of the line. Using the centre of gravity obtained with 90 μs integration time, we could have an error on the real position due to the space traversed by the beam in 90 μs. This means that in calculating the speed between two consecutive points the error could be of 100%. For this reason we decided to compute the scanning speed using intervals of 10 acquisitions (space covered by the beam in 900 s). In Figure 4.17 are reported the values for a run with a theoretical speed of 10000 mm/s, at isocenter. The measured mean value is 7590±323 mm/s. 78 Figure 4.17: Computed scanning speed for a scan In Table 4-2 the values of calculated speed are reported, with the relative errors and the theoretical values, for a group of 9 runs. #scan Vmeas(mm/s) 0 7590 1 3820 2 7513 3 3806 4 7580 5 8731 6 7586 7 7557 8 7521 sigma 323 257 307 283 293 224 286 377 536 Vth(mm/s) 10000 5000 10000 5000 10000 10000 10000 10000 10000 Table 4-2: mean speed calculation for all the runs 79 4.4.6 Beam shape In this subsection the beam shape and its variation during the irradiation will be presented. Figure 4.18 shows the mean and the sigma of the FWHM for 9 runs. Also in this case the acquisitions are accepted if the sum of the counts over all the strips exceeds than a fixed threshold. This means that we are not able to compute the FWHM of the beam at the extremes of the scanning line, when the scanning speed decreases to zero. In this particular condition in fact, when the speed is changing, also the intensity of the beam is going to zero The ratio R=(FWHM)/FWHM is constant and equals to 4%. It can be seen in Figure 4.18 (top) that the FWHM is constant in the various runs, with the exception of run number 9. Figure 4.18: mean FWHM (top plot) and sigma FWHM (bottom plot) for all the runs. In Figure 4.19 the values of the FWHM X and Y are shown for a single scan. The bottom plot of the figure is an exploded of the top plot, from which a single scanning line has been chosen. It’s possible to see that there is not a trend of the FWHM as a function of time. 80 Figure 4.19: FWHM X and Y for all the acquired points (top plot) and in a particular scanning line (bottom plot). 81 Conclusions In this chapter the experimental setup and the data analysis for a test done in October 2002 at the CRC (Centre de Recherches du Cyclotron) in Louvain-la-Neuve are presented. This work has been done within a collaboration with IBA (Ion Beam Application) with the aim of testing an active scanning magnet system for hadrontherapy treatment (raster scanning technique). Two magnets are placed close to the nozzle of the cyclotron, in order to move the beam in X and Y direction and delivery the dose in the whole target area. For the beam position and current verification four ionization chambers were mounted after the scanning magnet: two strip chambers for the fast beam position check, a pixel chamber to verify the real beam shape, and an integral plan for the fast beam current control and feedback. This measurement system will be implemented in the high level regulation system, which performs the two complementary tasks above described: the beam diagnostic on line with the irradiation and the feedback action given by the comparison between the data collected and the settings. The data analysis is done with the goal of studying the performance of the low regulation system, i.e. the beam spot scanning regulation (SMPS system) and beam current regulation (ISEU system). Data collected with the two strip chambers with a 90 s period are analysed to study the agreement between the real position of the beam and its theoretical value. Moreover the dose distribution on the target surface is expected flat, in order to simply verify its real distribution. The results concerning the beam measurements are the following: the position resolution is about 200 m the uniformity of the integral dose released in the whole treatment is about 0.8%, while the mean point to point variation (90 s period) is about 6%. In this preliminary test the regulation of the beam position with the magnet has shown promising results. Moreover the measurements done with the strip chambers can be used in real-time for supervision of the beam position and to the detect problems in the magnet and / or in the ISEU. The beam current regulation (ISEU) presents probably a too high instability in a short time verification, while the feedback loop has shown good results in a bigger time range. 82 5 The CATANA test: experimental setup and results The CATANA (Centro di AdroTerapia e Applicazioni Nucleari Avanzate) project stems from a collaboration between Laboratori Nazionali del Sud (LNS) of Istituto Nazionale di Fisica Nucleare, Physics Department, Ophtalmology Institute and the Radiology Institute of Catania University. The main goal of CATANA is the study and the application of protontherapy for the treatment of shallow tumours (4 cm max) like uveal melanomas and subfoveal macular degenerations. A dedicated kind of strip detector has been developed to be used on the beam line of the CATANA hadrontherapy facility. In the following we will describe the structure of the chamber, the experimental setup and the first preliminary tests, aimed to use the chamber as on-line beam monitor. 5.1 Development of a dedicated kind of strip detector as monitor of passive scanning beam Figure 5.1: An exploded view of the strip chamber In Figure 5.1, which shows an exploded view of the chamber, called MOPI, it is possible to see the four square frames which compose the detector. On the first and on the last one the two segmented strip planar anodes are mounted: one anode consists of 256 strips, each one 400 m wide. Because the electrically isolated interspace between adjacent strips is 100 m, 500 m is the pitch of the detector. The total sensitive area is (12.8 x 12.8) cm2, which widely matches the dimensions of the beam. The anodes are made of a 35 m thick kapton foil, covered with a 15 m thick Aluminum layer. By using the standard printed circuit board (PCB) technique we have obtained the conductive strips, placed on the aluminized side of the foil. Outside the sensitive area the width of the strips is reduced, and the signals are brought to a connector located on the edge of the anode frame. Through each connector 64 signals are carried out. The electronic boards for the read-out can receive 256 independent signals, therefore 4 connectors are needed to bring all the signals from the strips to the board. 83 The cathode is placed behind the first anode frame. It is made of a 25 m thick mylar foil, aluminized on both sides. This solution allows to have only one cathode foil between the two anodes. This also leads to turn the conductive side of the anode foils one facing each other, because both have to see the cathode. Each frame is 6 mm thick, and 6 mm is also the gap of the sensitive volume between the first strip foil and the cathode. In order to have the same gap between the cathode and the second anode foil, a gap frame has been inserted as shown in the Figure 5.1. Due to the high beam current no gas like nitrogen or carbon dioxide is filled in the chamber instead of air; the charge collected by the anode in this condition is in fact enough to have a readable signal. We have tried to make the total thickness crossed by the beam as thin as possible, in order not to shift the Brag pick position inside the patient. The maximum beam energy is high enough so that the protons reach the tumour, but the particles range does not have to be reduced too much. The total water equivalent thickness of the anodes and cathode is less then 100m. As described above about the pixel chamber, also in this case the electric field is generated by polarizing the cathode mylar foil. Typically the high voltage value has been set to –500 V. The strip voltage has been set to around 2 V, the same voltage at which the front-end input circuit has been set. A high voltage passive double-pole ( 2 Hz) low-pass filter is also the same of the pixel chamber. 84 5.2 The CATANA facility The CATANA (Centro di AdroTerapia e Applicazioni Nucleari Avanzate) project stems from a collaboration between Laboratori Nazionali del Sud (LNS) of Istituto Nazionale di Fisica Nucleare, Physics Department, Ophtalmology Institute and the Radiology Institute of Catania University. The main goal of CATANA is the study and the application of protontherapy for the treatment of shallow tumours (4 cm max) like uveal melanomas and subfoveal macular degenerations. Figure 5.2: Sketch of the CATANA facility Figure 5.2 shows a sketch of the CATANA facility: the machine used to accelerate to protons is a superconducting cyclotron, which leads the particles to an energy of 62 MeV. The distance between the beam extraction point of the cyclotron to the treatment room is 80 meters, while the beam line length inside the treatment room is 3 meters. In the development of the treatment line the main requirements were the maximum beam energy and the dose distribution. The energy of the beam has to be high enough to obtain, in air at isocenter, ranges up to 3 mm of eye tissue, while the requirements of homogeneity, in terms of lateral dose distribution in air, to reach a correct beam irradiation field are: diameter: at least 30 mm; particle fluence homogeneity: ± 2.5% lateral penumbra: 1 mm 85 5.3 The experimental setup 5.3.1 The beam line setup Figure 5.3: Sketch of the CATANA beam line The scheme of the beam line is shown in Figure 5.3. The proton beam exits in air through 50 m Kapton window placed at about 3 meters from isocenter (where the eye of the patient is placed). Before the window, under vacuum, it is placed the first scattering foil made by a 15 mm tantalum slab. The first element of the beam in air is a second tantalum foil 25 m thick provided with a central brass stopper of 4 mm in diameter. The double foil scattering system is optimized to obtain a good homogeneity in terms of lateral dose distribution, minimizing the energy loss. The range shifter and range modulator are placed downstream the scattering system and mounted on two different boxes. Two transmission monitor chambers, used for on-line control of the dose delivered to the patients during treatment, are placed further down along the beamline followed by MOPI, the strip ionization chamber tested in this work. Two diode lasers, placed orthogonally, provide a system for the isocenter identification and for patient centering during the treatment. The emission light of a third laser is spread out to obtain the simulated field. The last element before isocenter is a 37 cm brass patient collimator located at 8.3 cm upstream of the isocenter. Finally, two Philips xray tubes are, with axis defined by crosshair, are placed perpendicular and coaxial to the beam line, to provide the lateral and axial view of the tumour, during simulation and treatment [5.1]. 86 Figure 5.4: Picture of the CATANA beam line with the MOPI strip detector 5.3.2 Data acquisition structure The data acquisition system used in this test was very similar to the one built for the IBA test, described in subsection 4.3.2. Also in this case the system was based on a PCI DAQ card National Instruments DIO 32 HS and LabVIEW software, run on a PC. The structure of the data acquisition program is identical: a buffer with the controls to send to the detector is created and, at the start of the acquisition, it is emptied in a circular configuration. 16 lines are used for data input and 16 for data output. The main difference between the IBA and CATANA setup is the type of the scanning system: in the first case a pencil beam is moved in an active way in X and Y direction, to cover all the target area. In the CATANA setup the beam scanning is passive; the beam itself is wider and its energy is fixed. To obtain the spread out Bragg Peak into the target a modulation wheel is used. The different type of scanning leads the program to have different requirements. First of all a slower acquisition rate is needed in this case. The read out frequency was, in our case, 5 Hz. An advantage of a slow rate is the possibility of using mathematical methods for the evaluation of the beam structure on-line with the acquisition. The use of this kind of algorithms is, normally, highly time consuming, and their execution would not be possible with a high acquisition frequency. Using two strip detectors the output of the read-out is the integration of the charge produced by the beam in the vertical and horizontal direction. The parameters computed on these profiles are the center of gravity, the full-width half-maximum and the skewness; their mathematical formulation will be explained in detail in the next section. The steps needed to configure the DAQ card are identical to that described in section 4.3.2, and also the subroutines (VI) used to send controls to the detector and to receive data have the same structure. 87 5.4 Experimental Results 5.4.1 Calibration As stated in section 3.4.2 regarding the calibration of the pixel chamber, the gain of a given channel is proportional to the thickness of the gas gap seen by the pixel/strip and to the inverse of the charge quantum of the integrator channel serving it. We said also that the electronics gain of each channel has a spread (r.m.s.) less than 1%. To calibrate the pixel chamber, and thus to measure the gain of each pixel, at the GSI test the chamber was irradiated with a uniform field. With the CATANA passive scanning system this kind of procedure is not possible, because the beam cannot be moved on the detector surface, and a uniform field cannot be obtained; moreover, the dimensions of the beam are very different. The FWHM of the profile is around 10 ÷ 13 cm and this allows the calibration of the strips by inserting a film in front of the detector sensitive area. Strips and film are thus crossed by the same beam, facilitating a direct comparison between the strip and the film data (after the correct integration along the vertical and horizontal directions). The radiographic films used were Kodak Extended Dose Range (EDR) type [5.1]. This kind of calibration is possible only with passive scanning systems. The gain is in fact due to the gas gap, but this can vary among different points of the same strip; thus with the active scanning system of a pencil beam it would not be correct to use a single calibration coefficient for a strip. But in this case, because of the fixed beam position and the constant fluence distribution, this calibration method can be used with a good approximation. The simultaneous irradiation of both strips and film has been done in two different beam line configurations, one inserting the modulation wheel (patient treatment configuration) and the other without it (beam check configuration). The average beam current is 2.5 nA and the beam energy at the exit in air, upstream the range shifter, is 62 MeV. For the calibration runs the exposition time was around 10 seconds. The first step in the comparison between strip and film data is the measurement of the background of each detector. Acquiring the strip chambers in a beam off condition a constant count distribution over all the channels is present. Figure 5.5 shows the background average values (counts per second) that have to be subtracted to the counts measured with the beam on. Figure 5.5: Avarage number of counts/s for the two strip chambers 88 As a similar procedure is not possible for the evaluation of the film background, in this case the outmost area of the film was taken as background. Its subtraction from the raw data leads a variation on the area described by the beam profile of 2 %. Figure 5.6 shows the left tail, of the same beam profile, with and without the background for film and strip detector. Figure 5.6: Comparision of the left tail of beam profiles with and without background for the strip chamber and the film. An other important correction which was applied to the film data is due to the saturation of the central part of the irradiated area, the zone with the higher beam intensity. The central profiles of the film (before integration) were in fact modified using a polynomial fit, to correct the overexposure given by an excessive exposition time. In Figure 5.7 (a) a film profile with the saturation zone is plotted; the red curve is obtained computing a fit over the not saturated film points. Figure 5.7 (b) shows the integrated profile (sum of the real profiles along vertical or horizontal direction) with and without the correction for the overexposed points; the area variation is 0.26%. Figure 5.7: (a) Single film profile with (red points) and without (black points) saturation effects correction. (b) Integrated film profile with (red points) and without (black points) correction. In Figure 5.8(a) we show the horizontal profile of the modulated beam measured with the strip detector (black points) and with the film (red points). The two curves are normalized to total area. We want to remark that now we call profile the curve measured with an integration of the collected charge along one axis, but this is not a real profile of the beam (which can be obtained only with a real bidimensional detector, as the film or the pixel chamber). In Figure 5.8(b) the distribution of 89 the point to point differences between the strip and film profiles is reported. The spread of the differences (r.m.s.) is 0.17, and we decided to divide it by the maximum of the film profile in the left plot. Because the treatment concerns only the central part of the profile it’s possible to compute the average discrepancy between the film and the data strip referring to the maximum value of the curve, which is, in this case, 1.9% Figure 5.8: (a) Horizontal strip and film profile for a modulated field. (b) Distribution of the point to point differences between the two profiles The same measurements were made with the unmodulated beam (Figure 5.9 (a) and (b)). The variation between the two profiles is, in this case, ~1. Figure 5.9: (a) Horizontal strip and film profile for a not modulated field. (b) Distribution of the point to point differences between the two profiles The values obtained using the vertical strip detector are nearly the same: the difference between the curves is 1.9% with the modulated field and 1.6% with the not modulated one. The calibration coefficient of a strip i, C, is given by C(i)=F(i)/S(i), where S(i) is the count of the given strip and F(i) is the correspondent count of the film. Raw counts are thus multiplied by the calibration constants before being used. To check the consistency of the calibration curves obtained in the modulated and unmodulated beam configurations the difference between the two data sets was taken. 90 Figure 5.10: Distribution of the point to point differences between the calibration coefficients computed with modulated and not modulated beam, for horizontal (a) and vertical (b) strip detector. This is shown in Figure 5.10 for the horizontal (a) and vertical (b) strip detectors; in both cases the spread of data (r.m.s.) is about 4%. Such a high discrepancy is due to a large range of central strips considered. On the tails in fact, where the beam intensity decreases, a small difference between strip and film leads the calibration coefficients to be very different, because of the low beam intensity in this region. Considering only the central part of the profile, where the dose released fraction is maximum, the consistency between the two calibration curves is better than 1%. 5.4.2 Fluence measurements and homogeneity of dose distribution The first analysis done on the collected data was the comparison of the gains of the two strip detectors as described in section 4.4.2. Figure 5.11: Integrated counts over all the strips with 5 Hz of acquisition rate: (a) first run (b) eleventh run. Data were taken in runs of about 30 seconds, reading out the MOPI strip ionization chambers at a rate of 5 Hz. Figure 5.11 shows the integrated counts over all the strips as a function of the acquisition number (with a 5 Hz rate the acquisition period is 200 ms), for the first run (a) and the 91 eleventh (b). A very clear improvement in the uniformity of the behaviour between the two detectors is shown comparing the plots of this figure. From a maximum value of ~4.5% (first run), the gain discrepancy goes to less then 1%. In Figure 5.12 the distribution of the ratios between the gains of the vertical and horizontal strip detectors (respectively black and red points of Figure 5.11 (b)) is reported. The mean value of the distribution shows that the difference between the two detectors gains is less then 0.5%, with a stability better then 0.1%. The left tail is due to the initial acquisitions, when the discrepancy is lightly higher then in the last part of the run. The main factors that influence the count of a single strip are the gap between the electrodes (since there is no significant variation on the width strip only the gap may lead to a variation of the sensitive volume) and the electronic gain of the channel. Figure 5.12: Distribution of the vertical and horizontal strip detectors gains ratios; data are referred to the two curves of Figure 5.11 (b). In Figure 5.13 the distribution for a run of the counts of a single acquisition added over all the strips is reported. The spread of the data, normalized to the mean value, is 0.8%. This value, because of the good agreement shown in the trend of the black and red curves of Figure 5.11 (b), can be taken as representative of the beam current stability. As seen in subsection 5.3.1 two transmission monitor ionization chambers, to provide the on-line control of the dose delivered to the patient, are placed on the CATANA beam line [5.2]. To compare the integrated fluence measurements of the four detectors (two strip chambers and two CATANA chambers) eight runs were taken, with the same beam condition. Data shown in Figure 5.14 represent the charge collected by each detector in an entire run, normalized to its value averaged over all the runs. The measured fluctuations of the strip chambers are greater then those of the CATANA monitor chambers, but the results of the integrated beam fluence agree within 0.4%. 92 Figure 5.13: Distribution of the integrated counts over all the strips (horizontal chamber) with a 5 Hz acquisition rate. Figure 5.14: Comparison of the fluence integrated over the acquisition run for the vertical and horizontal strip chambers and for the CATANA ionization chambers, for 8 different runs. 5.4.3 A mathematical method for the evaluation of the beam structure In Figure 5.3 the scheme of the CATANA beam line was presented. The optimal position where to insert a detector able to check the beam shape should be as close as possible to the patient, in order to measure the total dose delivered and of its distribution on the target surface. As it can be seen in Figure 5.3 this kind of configuration is not feasible: the brass cylindrical collimator, with the collimator dedicated to each patient, has to be in fact the last element of the beam line. Since the MOPI strip chamber can be inserted upstream of the collimator, the beam structure seen by MOPI is not exactly the same as at isocenter. We have thus to find a mathematical method to apply to the data collected upstream the collimator, and correlate it with the beam shape at the isocenter. 93 Defining N(x,y) the fluence of the beam, the measurement of the two strip chambers may be defined as follows: y0 f x N x, y dy y0 f | y x0 N x, y dx x0 where the integration is performed on a strip width . The detectors thus measure the projections of the fluence N(x,y) along x (vertical strip chamber) and y (horizontal strip chamber). The parameters computed for the profiles are: the Full Width Half Maximum, the center of gravity (c.o.g.), the sigma ( and the skewness ( (respectively the second and third moment of a distribution). Their mathematical description is the following: N N c.o.g . ci xi i 1 N ci 2 ci x i x i 1 N ci i 1 c x N 2 3 i 1 i 1 i x i 3 N c i 1 i and 3 3 where: ci: count of the strip i (N is up to 256, the total number of strips) xi: position of the strip i x : center of gravity (c.o.g.) : skewness In order to minimize the effects of the tails, where the beam intensity is negligible, on the skewness, we decided to compute this parameter using a fixed number of strips around the c.o.g., removing the external parts of the sensitive area. In a perfectly symmetric curve the skewness value is zero, while it becomes positive or negative if the steeper tail is, respectively, at the left or at the right of the center of gravity. Experimental results One of the main aims of the preliminary tests was the evaluation of the sensitivity of the skewness to the beam shape variation. The role of the detector is in fact to check of the beam state during patient irradiation. Fluence and skewness have thus to be computed for each data taking and verified to be constant within a range to be defined as acceptable. The first step is the computation of the skewness in different beam shape conditions, to verify its level of sensitivity; the second one is the definition of the values that can be accepted during the irradiation. The comparison between 94 the values and the results of the CATANA standard procedure for the beam evaluation, at the isocenter point, was made for different beam setting conditions. The modification of the beam structure was obtained varying the current in the last quadrupoles. The following results were obtained from data collected in a beam line configuration without the modulator wheel and range shifters. At first, to verify the sensitivity of the skewness, and thus its variation within a run with constant current magnet conditions, its relation with the beam intensity was checked., During the irradiation a decrease of the beam current, which may lead to a variation of the skewness, is in fact possible. In Figure 5.15, for a ~60 seconds run, the skewness and the intensity are shown separately in the first and second plots, while are one function of the other in the bottom plot. Considerable skewness variations correspond to large beam current variations, but the last graph of the figure shows that there is not a linear correlation between them. Figure 5.15: Comparison between the skewness and the collected charge. In order not to be affected by beam current variations the beam intensity was kept as constant as possible. In Figure 5.16 the distribution of the skewness of the beam horizontal profile, computed for each acquisition, is shown. The average value is –0.1615 with a spread of 0.0008, which leads the sensitivity of the skewness, in this case, to be about 0.5%. Because Figure 5.11 (b) shows a very good agreement in the charge collected by the two chambers, it’s possible to affirm that the variations of the gains of the strips are negligible with respect to the beam intensity and shape variations. The fluctuations of the skewnees shown in Figure 5.16 may be attributed to the beam (real modification of its structure) and not to the measurement system. 95 Figure 5.16: Skewness vs acquisition number and distribution of the values. In Figure 5.17 the skewness and the c.o.g. in function of the acquisition number are shown for a run in which the modulator wheel was used as normally during treatments. A periodic component in the first plot is evident, but the low acquisition rate (5 Hz) doesn’t allow to know the true frequency value. Figure 5.17: Skewness and c.o.g. vs acquisition number in a beam line configuration with the modulator wheel The preliminary results presented here were obtained with different beam magnet settings, in the beam configurations with and without the modulator wheel. They will be discussed later in relation with the CATANA parameters used for the off line evaluation of the beam features. Figure 5.18 shows the vertical and horizontal profiles in four different beam setting conditions, for a non modulated beam configuration. The ranges of the values assumed in these boundary conditions are from –0.247 to –0.160 for the vertical profile, and from –0.122 to +0.021for the horizontal one. In the same beam conditions, but inserting the modulator, because of the spread of the beam these values are closer to zero, and the variations among the different runs are smaller (Figure 5.19). 96 Figure 5.18: Vertical and horizontal beam profiles, for four different magnet settings, in a non modulated beam condition Figure 5.19: Vertical and horizontal beam profiles, for four different magnet settings, in a modulated beam condition. 97 The CATANA procedure for the beam evaluation For the beam evaluation at isocenter a silicon diode is used, with which the central horizontal and vertical beam profiles are measured. Figure 5.20 shows a vertical profile. Before each treatment the beam is checked using this detector, and different parameters computed on the collected data are verified to be in a defined range. The goal of our test is to find a direct correlation among the skewness and the following parameters used at CATANA: ST: beam symmetry defined as Dx 100 ST D x max where D(x) and D(-x) is the dose measured at the point x; x and –x are symmetrical to central profile axis. Only the upper part of the profile is used for the ST computation and the range of the included points is defined as function of the beam width. RT, related to the maximum dose difference, defined as: Dmax Dmin 100 D max Dmin where Dmax and Dmin are the maximum and minimum dose in the region defined above. The dose values are normalized to the dose measured in the point on the central axis (set to one in Figure 5.20). RT Figure 5.20: Beam profile at the isocenter, in air, detected with a silicon diod The tolerance values for the beam acceptance are: 97%≤ST ≤ 103% RT ≤ 3% 98 Comparison between MOPI and CATANA parameters RT [%] The first step in the comparison between the results of the two detectors is their simultaneous data acquisition for runs with different magnet settings. In this first part of the test the magnet current was changed in order to take the beam condition outside the acceptable range for the treatment. Applying the mathematical definitions of RT to the collected data, the numerical values of these parameter may only be greater than 0%. Using this kind of expression it’s not possible to know which tail is steeper than the other, and thus where the maximum of dose released in the target is placed. Because the skewness also provides this type of information, to allow a direct comparison among the results of the two detectors we decided to modify the definition of RT : if the steeper tail is at the right of the central axis RT becomes smaller then 0%. In Figure 5.21 and Figure 5.22 the RT and ST values, computed for the diode vertical profile, are related with the skewness obtained from the horizontal strip chamber profile. In both cases a trendline has been computed, and the χ2 was used to define the agreement between experimental data and the fit. The χ2 computed for RT data is 4.6, while for ST data is 2.2, with six degrees of freedom. The positive result of this test shows for both cases that there is a good agreement in the measurement of the beam shape variation between the two detectors. Moreover, it’s possible to compute an acceptable range of values that may be taken by the skewness, using the range defined for ST and RT parameters and this kind of comparison (better statistics is needed). 20 15 10 5 0 -5 -10 -15 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 Skewness Figure 5.21: RT (measured with vertical diode profile) vs skewness (measured with horizontal strip chamber profile) computed for seven different magnet configurations. 99 ST [%] 130 120 110 100 90 80 70 -0.35 -0.25 -0.15 -0.05 0.05 0.15 Skewness Figure 5.22: ST (from vertical diode profile) vs skewness (from horizontal strip chamber profile) computed for seven different magnet configurations. This comparison has to be performed using the beam in a test condition, without the modulator wheel. The MOPI detector is intended to be used as on-line beam monitor with patient irradiation, when the beam has to pass through the modulator and the range shifter. The interaction of the protons with the matter not only leads to a particles energy downgrading, but also to a spread in their trajectories, with a beam dimension broadening and a decrease in collimation. The last step in the test is thus aimed to verify if a correlation between the skewness values computed with modulated and not modulated beam exists. In Figure 5.23 the comparison between the skewness calculated in these two different beam line configurations, for four different magnet setting, is reported. The two graphs show that a linear relation between the two variables may exist, but more data are needed to compute a linear fit. Figure 5.23: Comparison between skewness computed for modulated and non modulated beam, for four different magnet setting. 100 The final correlation will have to be computed between the ST and RT parameters (without modulator) and the skewness obtained in a modulated beam configuration. The need to use these two different beam line settings is due to the control procedure used until now by the CATANA group. A future development in the use of the monitor strip chambers will require a new kind of beam checking at the isocenter, with modulated beam. 101 Conclusions In the first part of this chapter the design of a new kind of strip chamber has been presented. The sensitive area is (12.8 x 12.8) cm2 and it is segmented in 256 strips, each one 400 m wide. The detector is composed by two anodes: one is mounted with horizontal strips while in the other one they are vertical. The cathode foil (a 25 m thick mylar foil, aluminized on both sides) is inserted between them. The total thickness traversed by the beam is less then 100m water equivalent. The gap of the sensitive volume is 6 mm and the high voltage is normally set to – 400V. The read out of the 512 strips is performed individually using the TERA chip. This chamber has been developed to be used as a real-time beam monitor in the CATANA facility, where different eye pathologies are treated with proton beams. In the second part of the chapter the setup and the experimental results of a preliminary test are presented. The beam line is composed by a modulator wheel and a range shifter, two monitor chambers for fluence measurement and a cylindrical brass collimator. The new detector is placed between the monitor chambers and the last collimator. Because the dose is delivered using a passive method, the beam diagnostic is performed checking that the fluence measurements and a parameter computed using the measured profiles (skeweness, i.e. the third moment of a distribution) are within a range of values defined as acceptable. Runs with different beam setting conditions have been taken with an acquisition frequency of 5 Hz. To calibrate the strip chambers two runs with a radiographic film mounted on the detector surface have been done. The flux resolution was measured comparing the total charge collected with the value given by the monitor chambers, while the evaluation of the sensitivity of skewness to the beam variations was made relating the results with the CATANA standard beam checking procedures. The experimental results obtained are the following: the average variation between the profiles acquired with the strip chambers and those obtained by the films is about 2% the fluctuations of the charge collected by the strip chambers are greater then those of the CATANA monitor chambers, but the results of the integrated beam fluence agree within 0.4%. a linear relation between the skewness, computed with the strip chambers, and two standard parameters used for the beam evaluation (at the isocenter) has been assumed. In both cases a trendline has been computed, and the χ2 was used to define the agreement: in one case is 4.6 and in the other is 2.2, with six degrees of freedom. The results obtained with the first preliminary test are very promising. The use of skewness as parameter for beam evaluation has been shown to be feasible. However further investigations about the definition of the range values for beam acceptance are still needed. 102 Bibliography [1.1] G. Steel, Basic Clinical Radiobiology, Hardcover, p. 1 [1.2] Trevor McMillan and G. Steel, Basic Clinical Radiobiology, Hardcover, p. 71 [1.3] Edward L. Alpen, Radiation Biophysics, Academic Press, p. 222 [1.4] Michel R. Horsman and Jens Overgaard, Basic Clinical Radiobiology, Hardcover, p. 158 [1.5] Michael C. Joiner, Basic Clinical Radiobiology, Hardcover, p. 205 [2.1] Sjirk Boon, Dosimetry and quality control of scanning proton beams, Rijsuniversiteit Gronigen, p. 29 [2.2] H. A. Bethe and J. Askin, Passage of radiations through matter, Experimental Nuclear Physics, vol I, Wiley, New York, 1953, p. 176 [2.3] H.H. Andersen et al, Stopping power of Be, Al, Cu, Ag, Pt and Au for 5-12 MeV protons and deuterons, Phys. Rev. 153 (1967), 338 [2.4] H.H. Andersen et al, The stopping power of gold in Bethe region, Nucl. Instr. Method 191 (1981), 475 [2.5] ICRU, Stopping power and range for protons and alpha particles, Tech. Report 49, ICRU Bethesda MD, 1993 [2.6] W. H. Bragg, On the ionization curves of radium, Phil. Mag. 8 (1904), 726 [2.7] L. Landau, On the energy loss of fast particles by ionization, J. Phys. 31 (1995) 593 [2.8] P. V. Vavilov, Ionization losses of high energy particles, Zh. Ekps. Teor. Fiz. 32 (1957), 920 [2.9] H. J. Bhabha, On the penetrating component of cosmic radiation, Proc. Roy. Soc. A164 (1938), 257 [2.10] G. Rinker, Patient dose measurements in photon fields by means of silicon semiconductor detectors, Med Phys.14 (1987), 870 [2.11] G. Molier, Theorie der Streuung schneller geladener teilchen II. Mehrfach und vielfachsstruung, Z. Naturforschung 3a (1948), 78 [2.12] M. Lee et al, An empirical method to build up a model for proton dose distribution for a radiotherapy treatment planning package, Phys. Med. Biol. 38 (1993), 989 [2.13] B. Gottscalk et al, Multiple Coulomb scattering of 160 MeV protons, Nucl. Instr. and Meth. B 74 (1993), 467 [2.14] N. Kalantar-Nayestanaki, Results of recent Bremsstrahlung measurements on few-body system at KVI, Nucl. Phys. A 631 (1998), 242 [2.15] R. F. Carlson et al, Proton total reaction cross sections for the doubly magic nuclei 16O, 40Cd and 208Pb in the energy range 25-50 MeV, Phys. Rev C 12 (1975), 1167 [2.16] P. U. Perey et al, Optical-model parameters, 1954-1957, Atomic Data and Nuclear Data Tables 17 (1976), 23 [2.17] M. J. Berger, Proton Monte Carlo Transport Program PTRAN, Tech Report NISTIR 5113, NIST, 1993 [2.18] P.G. Young et al, Transport data libraries for incident proton and neutron energies to 100 MeV, Tech. report LA-11753-MS, Los Alamos National Laboratory, 1990 [2.19] T. Bordfeld, Approximation of Bragg curve for therapeutic proton beams, Med. Phys. 24 (1997), 2024 [2.20] M. Abramowitz et al, Handbook of mathematical functions, Dover, New York, 1972 [2.21] Universitätsklinik Heidelberg, Proposal for a dedicated ion beam facility for cancer therapy. [2.22] P. Mandrillon, Accelerator in hadrontherapy, Proceeding of the International Week on Hadrontherapy, Archamps, 20-24 November 1995, Elsevier [2.23] P.Mandrillon et al, Commissioning and implementation of the MEDICYC cyclotron program, 12th Int Conf Cyclotrons and their applications, Berlin, May, 1989 103 [2.24] U. Amaldi et al, The TERA project and the Centre for Oncological Hadrontherapy, vol I and II, Frascati, 1995 [2.25] Khoroshokov VS et al, Soviet Project of PTF, Proc of the NIRS Int Workshop on Heavy Charged Particles Therapy, Chiba, Japan, 1991 [2.26] Martin RL., A proton accelerator for medical application, NIM B25/25, 1987, 1087-1091 [2.27] G. Arduini et al, Physical specifications of clinical proton beams from a synchrotron, Med Phys. 23 (6), June 1996 [2.28] Th. Haberer et al , Magnetic scanning system for heavy ion therapy, NIM A330 (1993), 296-305 [2.29] Biophysics Group et al, GSI Report 91-18, 1991 [2.30] R Cirio et al, Two-dimensional and quasi-three-dimensional dosimetry of hadron and photon beams with the Magic Cube and the Pixel Ionization Chamber, Phys. Med. Biol. 49 (2004) 3113-3724 [2.31] G. Bonazzola et al, Performances of a VLSI wide dinamyc range current-to-frequency converter for strip ionization chambers, Nucl. Instr. And Method. A 405 (1998) 111 [2.32] http://ptcog.web.psi.ch/ptcentres.html [3.1] Universitätsklinik Heidelberg, Proposal for a dedicated ion beam facility for cancer therapy. [3.2] U. Weber, W. Becher, G. Kraft, Phys. Med. and Biol. 45 (2000) 3627 [3.3] C. Brusasco, B. Voss, D. Shardt, M. Kramer, G. Kraft, Nucl. Instr. and Meth. B 168 (2000) 578 [3.4] http://www.windriver.com [3.5] M. Kramer, J. Radiat. Res. 42 (2001) 39 [3.6] R. Bonin et al, A pixel chamber to monitor the beam performance in hadron therapy, Nucl. Instr. and Meth A 519 (2004) [4.1] http://www.cyc.ucl.ac.be [5.1] G.A.P. Cirrone et al, A 62 MeV Proton beam for the treatment of ocular melanoma at Laboratori Nazionali del Sud-INFN, Nuclear Science, IEEE Transactions on, 51 (2004) [5.2] G.A.P. Cirrone el al, A fast monitoring system for radiotherapeutic proton beams based on scintillating screens and a CCD camera, Nuclear Science, IEEE Transactions on, 4 (2003) 104