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
Eur Radiol (2008) 18: 1375–1384 DOI 10.1007/s00330-008-0903-3 Soenke H. Bartling Julien Dinkel Wolfram Stiller Michael Grasruck Ijad Madisch Hans-Ulrich Kauczor Wolfhard Semmler Rajiv Gupta Fabian Kiessling Received: 21 September 2007 Revised: 1 December 2007 Accepted: 19 January 2008 Published online: 23 April 2008 # European Society of Radiology 2008 Soenke H. Bartling and Julien Dinkel contributed equally. S. H. Bartling . F. Kiessling Junior Group Molecular Imaging, German Cancer Research Center (DKFZ), Heidelberg, Germany S. H. Bartling . W. Stiller . W. Semmler . F. Kiessling Department of Medical Physics in Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany J. Dinkel . H.-U. Kauczor Department of Radiology, German Cancer Research Center (DKFZ), Heidelberg, Germany M. Grasruck Siemens Medical Solutions, Forchheim, Germany EXPERIME NTAL Intrinsic respiratory gating in small-animal CT I. Madisch . R. Gupta Department of Radiology, Massachusetts General Hospital, Boston, MA, USA S. H. Bartling (*) Junior Group Molecular Imaging, Department Medical Physics in Radiology, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany e-mail: [email protected] Tel.: +49-622-1422686 Fax: +49-622-1422572 Abstract Gating in small-animal CT imaging can compensate artefacts caused by physiological motion during scanning. However, all published gating approaches for small animals rely on additional hardware to derive the gating signals. In contrast, in this study a novel method of intrinsic respiratory gating of rodents was developed and tested for mice (n=5), rats (n=5) and rabbits (n=2) in a flat-panel cone-beam CT system. In a consensus read image quality was compared with that of nongated and retrospective extrinsically Introduction With the increasing utilization of small-animal CT in preclinical and basic research, effective methods to compensate for physiological motion during scanning have been sought. Prospective [1–5] as well as retrospective [6, 7] methods for respiratory and cardiac gating have been described, and their benefit for structural lung imaging has been shown [2, 6, 7]. Furthermore, gating gated scans performed using a pneumatic cushion. In comparison to nongated images, image quality improved significantly using intrinsic and extrinsic gating. Delineation of diaphragm and lung structure improved in all animals. Image quality of intrinsically gated CT was judged to be equivalent to extrinsically gated ones. Additionally 4D datasets were calculated using both gating methods. Values for expiratory, inspiratory and tidal lung volumes determined with the two gating methods were comparable and correlated well with values known from the literature. We could show that intrinsic respiratory gating in rodents makes additional gating hardware and preparatory efforts superfluous. This method improves image quality and allows derivation of functional data. Therefore it bears the potential to find wide applications in small-animal CT imaging. Keywords CT . Small-animal imaging . Flat-panel detector . Intrinsic gating enables reconstruction of a 4D time series that can be used to calculate functional parameters such as respiratory tidal volume and cardiac ejection fraction [6–8]. Regardless of whether retrospective or prospective gating is used, all gating methods that are currently in use depend on an extrinsic sensor to derive a gating reference signal. Such a sensor could be a respiratory cushion that is placed below the chest of the small animal [1, 6, 7] or an optical system to deduce the breathing 1376 movements [9, 10]. For cardiac gating, EKG electrodes are routinely used. Intrinsic methods extract the gating information from the acquired projection data directly, obviating the need for additional hardware or preparatory efforts. Intrinsic methods have been described for human cardiac [11, 12] as well as respiratory gated CT imaging [13, 14]. However, intrinsic gating methods for small animals have not yet been described. Human and small-animal CT instrument characteristics and data acquisition parameters vary substantially. To the best of our knowledge, this paper describes the first intrinsic respiratory gating method for small-animal imaging that takes into account the characteristics of smallanimal cone-beam CT instruments such as a relatively large z-coverage, relatively slow data acquisition, continuous volume acquisition and non-spiral scanning. In this research we successfully demonstrated that, using intrinsic respiratory gating in small animals, it is possible to achieve the same image quality as that from the established extrinsic gating using external gating hardware. It is also possible to reconstruct 4D data sets, thus enabling derivation of functional parameters of respiration. tion translates into a minimal detectable feature size of 200 μm. Each projection image acquired during gantry rotation is time stamped and is labelled with the angle of acquisition. Gantry rotation times can be varied from 2 s to 19 s in steps of 1 s, the maximum total CT data acquisition time being 80 s. The tomographic image reconstruction is based on a modified Feldkamp algorithm [16]. For each animal examined, a total data acquisition time of 80 s was used with a gantry rotation time of 5 s. This resulted in a projection data set over 16 full rotations. A tube voltage of 80 kV and a tube current of 50 mA with continuous radiation were selected. Both extrinsic as well as intrinsic motion-gated reconstruction was performed for each animal. The reconstruction field of view was 4.5 cm transaxially with a reconstruction matrix of 512×512 pixels and an axial slice spacing 0.2 mm resulting in a voxel size of 0.08×0.08×0.2 mm3. A sharp reconstruction kernel (H80s) was used for image reconstruction. All gating algorithms contained a procedure to obtain a gating reference signal. Essentially, both gating methods compared herein differ only in the way the reference signal is derived. Materials and methods Derivation of a gating reference signal for extrinsic gating Flat-panel-based volume CT instruments and data acquisition A prototype CT instrument (Siemens Medical Solutions, Forchheim, Germany) was used for gating experiments. The technical CT setup is identical to that described in [7, 15] where extrinsic gating was originally tested [7]. Its main features are a flat-panel detector and a modified X-ray tube, both mounted on a multi-slice CT gantry. Taking the geometry of the CT system setup into consideration, the instrument’s total field of view is 25×25×18 cm3. For rat and mouse CT imaging, the active detector area was limited to 192 lines in z-direction and 1,024 rows in x-y-direction to increase the frame rate. The detector was read out in a 2×2 binning mode, meaning that four neighbouring pixels were averaged. This resulted in a decreased field of view of 25×25×4.5 cm3, which was still big enough to cover the entire thorax and diaphragm of a rat. The resulting frame rate was 100 frames per second (fps), which translates into an exposure time of 10 ms per projection. Rabbits were examined without reduction of the active CT detector area, because the lung did not fit in the reduced field of view that was employed for rats and mice. The detector read out frame rate was therefore reduced to 30 fps. All other parameters were the same for rabbit CT imaging. The spatial resolution of the CT system, as computed by examining a tungsten wire phantom, is 24 lp/cm at 10% modulation transfer function. This isotropic spatial resolu- The aim of this research is to show that the proposed intrinsic gating method leads to the same results as an established extrinsic gating method. Therefore the extrinsic respiratory gating was performed as described previously [6, 7], including the extrinsic method of gating reference signal derivation: A commercial small-animal monitoring unit (1025L and Signal Breakout Module, SA Instruments, Stony Brook, NY) was used to track the respiration movements using a pneumatic cushion. The pneumatic cushion was placed beneath the thorax of the small animal in prone position. Commercially available software was used to derive a respiratory gating reference signal. Derivation of a gating reference signal for intrinsic gating In comparison to extrinsic gating, the gating reference signal in intrinsic gating was derived from the image data alone. The key innovation described in this paper is the method used for deriving a gating reference signal from the raw projection data. A region of interest (ROI) that covers the diaphragm and adjacent structures on all projections was defined for each individual CT raw data set (Fig. 1). Its position was the same in all projections and was defined in absolute detector coordinates. The ROI extension in x-y direction was broad 1377 enough so that the ROI covered the lateral extension of the animal fully on all projections. Within this ROI, the center of mass (COM) in z-direction (P) was calculated from the raw projection data. All raw projection values were gain calibrated, and the value of defective detector elements was estimated using interpolation [15]. For calculating the COM, each line sum of projection values (mz) was multiplied with a weighting factor (z) that depends on z-position of the particular line. Weighted projection values from all lines were summed and divided by the total sum of projection values from the ROI (M) as shown in: P¼ X 1X mz Z; with M ¼ mz M z z (1) An example of a curve that shows the z-position of the COM as a function of angular projection position is shown in (Fig. 2a). The z-position of the COM depends mainly on two factors: (a) the angular position of the imaging chain and (b) the position of the diaphragm in the selected ROI reflecting the phase of respiration. The variations due to the angular position of the gantry have a fixed periodicity of 500 projections reflecting the number of projections in one gantry rotation around the animal. The phase of respiration has a more irregular period and reflects breathing excursion of the diaphragm along z-axis. In order to derive a gating reference signal, the influence of angular position on the resulting curve should be minimized while maximizing the effect of breathing excursions. This is accomplished by baseline correction described below. The COM of each projection position is normalized to the mean of the zposition values at this projection position for all acquired rotations to decrease the influence of (a). The resulting curve is shown in Fig. 2b. As can be seen, by explicitly reducing the angular position dependent oscillations, the dependence on the effect position of the diaphragm can be increased. Local maxima of this curve were used as gating reference points as marked in Fig. 2b. Fig. 1 An example of a manually selected region of interest (ROI) placed over the diaphragm and adjacent structures. Note that the ROI box has been made deliberately larger in x-y direction and extends beyond the margins of the animal. This is to ensure that the Motion-gated rebinning and CT reconstruction As described above, gating reference points were derived for every respiratory cycle for extrinsic gating by analysing the compression of the respiratory cushion and for intrinsic gating by analysing the parameter P derived from raw projections. All other steps, such as retrospective binning of projections from several rotations according to their phase, and volumetric reconstruction using these projection sets, were identical in both gating methods. In essence, this leaves the derivation of the gating signal the only difference between the new intrinsic and established extrinsic gating method. This assures that potential differences in retrospective binning, interpolation and reconstruction algorithms cannot cause a difference in the two methods compared. This is also the reason why all other steps of the gating algorithm beside the derivation of the gating reference signal are nearly identical to the already published procedure [7] and are therefore only briefly summarized below. The starting point of each respiratory cycle (0% point) was defined to correspond to the gating reference point of every motion cycle. In order to reconstruct a given phase of the respiratory cycle, the projections that were acquired within a certain time frame around that respiratory phase were selected for image reconstruction. Time frames were defined by start and end points, given as the percentage of the cycle length. The selected projections from the re-binning step, representing projections pertaining to a given phase of the respiratory cycle, were then interpolated to yield a new 360° projection data set consisting of 600 evenly distributed projections. If two or more selected projections were found to be at identical positions-recall that the angular position of each projection is recorded during different rotations-they were averaged to improve the signal-to-noise ratio of the interpolated projection. If no projections were found for a selected angular position, interpolation from the closest neighbouring projections was performed. Interpolation was weighted with respect to angular distance. whole body is within the ROI in all projections. The selection of ROI is the only manual interaction necessary for the proposed intrinsic gating algorithm 1378 Fig. 2 The z-position of the center of mass (COM), calculated from an ROI that included the diaphragm of a mouse, is plotted against the projection number. Each rotation consisted of 500 projections (i.e., the marks along the X axis constitute two consecutive rotations). The z-position of the COM is influenced by the projection number-with a periodicity of 500-and the respiratory motion of the diaphragm. The latter is visible as long spikes with less regular periodicity. Normalization with respect to projection number can accentuate the influence of the respiratory motion so that a gating signal can be reliably derived using the local maxima, shown as circles in the above example. For the purpose of illustration, the z-position is given without (a) and with (b) normalization with respect to the projection number. As can be seen, without this normalization step one respiration cycle would have been missed (arrow), jeopardizing the gating results. (AU = arbitrary unit) A motion-compensated, “still” image was reconstructed to illustrate one motion-free phase of the respiratory cycle. Optimal gating intervals used for computing this still image depended on the animal being imaged; these were empirically derived through experimentation. In rats and mice, a reconstruction phase interval of 20% to 80% yielded the best. This is because in mice and rats the most intense motion occurs either before 20% or after 80% of the respiratory phase. For rabbits, a reconstruction phase interval of 50% to 100% was selected because little perceptible motion was found to occur beyond the midpoint of the respiratory cycle. In general, the reconstruction phase interval should be chosen to cover the duration of least motion in the respiratory cycle. For comparison, nongated reconstruction was also performed. Here all projections from all rotations were used for image reconstruction. In order to visualize different phases of respiration and to compute functional respiratory data, a 4D time series of volumetric data was computed. For deriving these temporally varying images for rats and mice, the respiratory cycle was additionally divided in the following six phases: (70–80%), (80–90%), (90–100%), (0–10%), (10– 20%) and (20–30%). In rabbits the 0% to 50% interval was additionally split in five 10% long phases. Recall that the most intense motion in rabbits occurs during the first half of the respiratory cycle. Animals, contrast media All animal experiments were approved by the German (mice and rats) and American (rabbits) Governmental Review Committees on Animal Care. Six C3H/HeN wild-type mice (20 g), six healthy Copenhagen rats (250 g) and two New Zealand white rabbits (4 kg) were scanned. Four of the rodents (two rats and two mice) were additionally scanned after intravascular contrast administration to improve the conspicuity of lung 1379 vessels. For this purpose, 2.5-ml Fenestra-VC (ART Advanced Research Technologies, Saint-Laurent, CA)-a blood pool contrast agent with 50 mg iodine/ml-was injected into the tail vein of the rats, 5 min prior to the data acquisition [1, 17]. The mice were given 0.5 ml of the same contrast agent. The rodents were anaesthetized by continuous inhalation of 3% Sevoflurane (Sevorane, Abbot, Maidenhead, UK) in oxygen during preparation, the injection of contrast media and data acquisition. Rabbits were anaesthetized by intraperitoneal injection of 1 mg/kg acepromazine, 40 mg/kg ketamine and 6 mg/kg xylazine. The pneumatic cushion was attached to the animals to record the respiratory movements. The animals were free-breathing. Post-processing The reconstructed imaged datasets were supplemented by a DICOM3-header to enable importation into standard post-processing software. InSpace (Siemens Medical Solutions, Forchheim, Germany) was used for analysis and generation of appropriate reformations from both 3D and 4D datasets. The Medical Imaging Interaction Toolkit (German Cancer Research Center, Heidelberg, Germany) [18] was used to semi-automatically segment the lung volumes. Evaluation and data analysis Image quality of non-gated, intrinsically and extrinsically gated datasets was compared with respect to the following score table: A. Delineation of the diaphragm: B. Rib delineation: 0 points no clear delineation, severe motion artefacts 1 point some blurring, contours predominantly assessable 2 points clear delineation, no motion artefacts C. Tracheobronchial tract: 0 points trachea and main bronchi not assessable 1 point trachea and main bronchi assessable, but no segmental bronchi can be delineated 2 points trachea, main bronchi and segmental bronchi clearly visualized D. Delineation of central vessels within the mediastinum: 0 points no clear delineation, severe motion artefacts 1 point some blurring, contours predominantly assessable 2 points clear delineation (n, no motion artefacts Three readers (SHB, FK, JD) decided scores in consensus. Sum scores were calculated for each criterion and each animal. In addition, mean total sum scores (± standard deviation) of the non-gated, extrinsically and intrinsically gated datasets in mice, rats and rabbits were calculated. Differences in the achieved total sum scores were statistically compared among all three groups using a non-parametric Friedmann test. Differences between two groups were analyzed using the non-parametric Wilcoxon signed rank test. A p-value of <0.05 was considered to indicate significant differences. For the determination of respiratory tidal volume the time points of maximum expiration and inspiration were selected. Because it was not known whether the time points of maximum and minimum lung volumes lie within the same reconstruction phase for each animal, and for each gating method, the phases of maximum inspiration and expiration were selected by visual inspection. The lung volumes at these time points were segmented. Respiratory tidal volume was calculated as the difference between both lung volumes. Results Comparison of image quality When comparing the mean sum scores of non-gated and respiratory-gated scans significant differences between groups were found by the Friedmann test (p=0.003). Using extrinsic gating image quality only improved significantly in mice (p=0.05). In rats improvement of image quality after extrinsic gating hardly failed significance as compared to non-gated scans, although the overall sum score was higher. Intrinsic gating resulted in a significantly image quality improvement for mice (p=0.03) and rats (p=0.03). However, statistical testing revealed no significant differences in image quality between extrinsically and intrinsically gated images. Also in the two rabbits examined higher sum scores for image quality were found for each gating method. By visual inspection the diaphragm was blurred and had double contour from motion-induced shadowing or ghosting in non-gated datasets. In contrast, it was sharply delineated in the gated datasets (Figs. 3, 4 and 5). Lung structures such as bronchi and vessels were also sharper and better delineated in the gated datasets (Figs. 3, 4 and 5). Motion artefacts, most pronounced around bony structures of the rib cage, were considerably diminished in the gated datasets (Fig. 4). The difference between gated and nongated datasets was most pronounced in rabbit scans, probably because of greater excursion of the thoracic cage. While there was obviously a perceptible difference between the non-gated and gated scans, there was no significant difference in image quality between the two different types of gating schemes tested when comparing 1380 Intrinsic gating schemes in multi-detector CT where the gating signal is derived from the slice-by-slice projection data have been implemented and shown to work for human subjects [11–14]. MDCT machines, however, offer less favorable conditions for intrinsic gating than small-animal imagers that are based on flat-panel detectors. As a consequence algorithms developed for MDCT machines are specifically adapted for human respiratory physiology that differs significantly from that of small animals. However, these intrinsic gating algorithms tended to be complex and difficult to implement. Perhaps not surprisingly, none of them are being used in routine clinical practice so far. The main result of this research was to show that in small animals an intrinsic gating signal derived solely from the projection data using a relatively simple and easy to implement algorithm can be used to improve image quality to the same extent as extrinsic gating derived from external hardware. We also show the feasibility of high-fidelity 4D datasets computed using intrinsic gating. The functional parameters such as respiratory tidal volume, derived from the intrinsically gated 4D datasets, have the same level of fidelity as those from extrinsic gating. These values also compare favourably with the values quoted in the literature, which were derived using other means. The ratings and statistics performed in this study resulted in image quality improvements between gated and nongated CT acquisitions-regardless which gating method was used. No significant difference in image quality was found between the two gating methods. The obvious similarity of respiratory functional parameters derived from both gating methods further support our conclusion that intrinsic as well as extrinsic gating not only both improve image quality, but also do that to a very similar extent. In the following paragraphs, we briefly outline the main differences between the intrinsic gating schemes described in the literature and that proposed here. Fig. 3 Non-gated (a, d), extrinsically gated (b, e) and intrinsically gated (c, f) coronal images of a mouse thorax enhanced with a blood-pool contrast media (same windowing). Magnified views of a region of interest, as marked in the coronal reformations in the upper row, are given in the lower row. Both extrinsic and intrinsic gating improved the sharpness and definition of vessels (small arrows) and bronchi to a similar extent. The diaphragm, which was blurred and had a ghost contour in the non-gated images (long arrows in d), was sharper and without the ghost contour in the gated datasets (long arrows in e, f) the mean sum scores in mice, rats and rabbits. The scoring details are listed for each animal and criterion in Table 1. Functional lung imaging The rodents exhibited a gasping type of respiration with minimal changes in respiratory motion and ventilation frequency once a steady-state of narcosis was reached. Respiration rate varied from 20 to 35 min−1 in mice and 20 to 39 min−1 in rats. In rabbits, the expiratory phase was almost as long as the inspiratory phase. Their respiratory rate ranged from 25–40 min−1. In the 4D time series, the respiratory excursions of both the diaphragm and the rib cage could be visualized. Points of maximum inspiration and expiration could be determined in all cases and for both gating methods. Mean expiratory and inspiratory lung volumes revealed by semi-automated segmentation are shown in Table 2. No outlier was found. The lung volumes obtained from the intrinsic and extrinsic datasets were within standard deviation of each other (Table 2). A literature search for the approximate values of lung volumes and the respiratory tidal volumes was conducted. This revealed that the values quoted in the literature correlated well with values computed using gated images (Table 2). Minor differences in the values may derive from different animal sizes, narcosis states, the physiological conditions under which the animals were tested and measurement error. Discussion 1381 Fig. 4 Non-gated (a), extrinsically gated (b) and intrinsically gated (c) coronal images of a rat thorax after administration of blood-pool contrast media. Extrinsic as well as intrinsic gating showed comparable image quality. Using both methods, motion artefacts around the ribs (black arrow) and double contour along the diaphragm from motion-induced shadowing (long white arrow) were diminished, and the visualization of the basal bronchi (short white arrows) was improved as compared with the non-gated images In multi-slice spiral CT only parts of the volume are examined during one revolution around the patient. Since the detector width in MDCT machines is mostly 2 to 4 cm, the projection data are severely limited in Z extent. However the COM along x-y axis can be used to derive a cardiac gating signal [11]. This is in contrast to the method presented here where the COM is quite sensitive to movements of diaphragm and other structures that predominantly move along the subject’s z-axis. Most methods for human CT were developed for cardiac gating because a normal thorax scan usually does not require respiratory gating. It is possible to adapt the proposed method based on COM for cardiac gating. Cardiac motion-which entails Z shortening, helical screw motion, oblique oscillatory undulations and in-plane (i.e., XY) contraction-is much more complex than diaphragmatic movement. Therefore, for this application more complex post-processing steps will be required, and the algorithmic complexity may rival that in human CT using intrinsic gating [11]. The high scan speed of modern human CT machines ensures that chest scanning can be concluded in one breath hold. Respiratory gating has been most intensively investigated in the domain of CT applications in radiation oncology. One of the proposed schemes in this field is very similar to the intrinsic gating approach proposed in this paper [12]. Sonke et al. used projection data from a large cone-beam flat-panel CT system integrated with a linear accelerator. They used the projections to derive a 1D signal along the z-axis that correlated well with the movement of the diaphragm in the z-direction. In contrast to our approach, this method assumes that respiratory phases correlate with the extent of the 1D signal. Iterative reconstruction algorithms for derivation of intrinsic gating signal have also been described [19]. These techniques, however, tend to be computationally intensive and are not (yet) used routinely. The proposed method can probably be further improved. However, the similarity of images from intrinsic and extrinsic gating suggests that the quality of intrinsic gating signal is already near optimal. Therefore, any algorithmic improvements would most likely result in small, secondorder improvements in the final image quality. There is also a limit to which additional improvements add to the image quality. This is because the positional reproducibility of thoracic structures from one respiratory cycle to the next is of the order of 100 μm [20] in rodents. A spatial resolution exceeding 100 μm is therefore not possible by collecting projections over multiple respiratory cycles. The influence of rotation time, total scan length and dose to the image quality can be assessed and optimised in the future. Since the proposed intrinsic gating method does not rely on any special features of the CT machine used, it can be implemented in any small-animal cone-beam CT systems. Fig. 5 Non-gated (a), extrinsically gated (b) and intrinsically gated (c) coronal images from a rabbit thorax. Without gating, lung structures, e.g., vessels (white arrow), diaphragm (black arrow) as well as lung parenchyma, were blurred. As compared to non-gated scans, the improvement in the image quality was comparable regardless of which gating method was used 1382 Table 1 Results of the multireader analysis of image quality without and with extrinsic and intrinsic gating, respectively Criterion Non-gated Extrinsic gating Intrinsic gating Mice (n=6) with score Sum score Mice (n=6) with score Sum score Mice (n=6) with score Sum score 0 Diaphragm assessability Rib delineation Tracheobronchial tract Delineation of central vessels Mean sum score/ animal ± SD Diaphragm assessability Rib delineation Tracheobronchial tract Delineation of central vessels Mean sum score/ animal ± SD Diaphragm assessability Rib delineation Tracheobronchial tract Delineation of central vessels Mean sum score/ animal ± SD 1 2 0 5 1 2 4 0 1 5 0 0 5 1 3.8±1.3 (max. 8) 4 2 0 0 4 2 0 1 5 4 2 0 3.8±1.16 (max. 8) 2 0 0 1 1 0 0 2 0 0 2 0 2.5±0.7 (max. 8) 0 7 4 5 7 2 8 11 2 0 1 2 2 1 2 0 0 0 0 3 0 2 0 4 6.5±1.3 0 1 0 1 0 1 0 0 7.5±0.83 0 0 0 0 0 0 0 0 8.0±0.0 6 3 4 2 12 9 10 8 5 5 5 6 11 11 11 12 2 2 2 2 4 4 4 4 1 2 0 0 0 2 0 2 0 4 6.7±1.2 0 1 0 2 0 1 0 1 7.2±1.17 0 0 0 0 0 0 0 0 8.0±0.0 6 4 4 2 12 10 10 8 5 4 5 5 11 10 11 11 2 2 2 2 4 4 4 4 The number of scans that were assigned to score 0, 1 or 2 is indicated for each criterion. In addition the mean total sum scores of six mice, six rats and two rabbits considering all criteria are given. Extrinsic as well as intrinsic respiratory gating causes a general improvement of image quality in all animals (p<0.05 in mice and rats; p=0.05 in rabbits). The differences between extrinsic and intrinsic gating were not significant (SD: standard deviation) Table 2 Lung volumes of mice (n=6), rats (n=6) and rabbits (n=2) after semi-automated segmentation from extrinsically and intrinsically gated datasets (SD: standard deviation) Mice (mean ± SD) [μl] Extrinsic gating Expiratory volume 583.7±54.1 Inspiratory volume 798.5±52.8 Tidal volume 214.8±21.9 Intrinsic gating Expiratory volume 567.3±71.7 Inspiratory volume 791.8±93.8 Tidal volume 224.5±59.1 Difference: intrinsic vs. extrinsic gating Expiratory volume 16.3±40.4 Inspiratory volume 6.7±65.2 Tidal volume −9.7±40.9 Reference values from literature Expiratory volume 400 (*) 430(7) Inspiratory volume 640 (7) Tidal volume 260 (*) 210 (7) Rat (mean ± SD) [ml] Rabbit (mean ± SD) [ml] 4.33±0.10 5.84±0.14 1.56±0.21 53.2±2.0 67.0±3.8 15.1±1.8 4.28±0.05 5.62±0.13 1.34±0.16 52.6±1.0 69.5±4.6 16.9±5.7 0.04±0.10 0.22±0.23 0.17±0.29 0.6±3.1 −2.5±0.8 −3.1±3.8 3.9 [21] 4.6 (7) 6.17 (7) 0.6–2.0 [21] 1.52 (7) 76.4 [22] 10.8 [23] The differences between intrinsic and extrinsic gating are also tabulated. All differences are within 1 SD of the values being compared. Volumes correlate well with reference values from the literature. (*Mouse Phenome Project, The Jackson Laboratory, Bar Harbor, ME) 1383 A prerequisite would be access to some basic scan parameters such raw projection data. Even existing systems can be easily adapted with this gating algorithm making hardware changes superfluous. Currently, selection of the ROI is the only manual step in this algorithm. An automated method for the ROI selection, which would make the process fully user independent, would facilitate the workflow. In summary, this paper demonstrates a method for intrinsic gating from projection data that is suitable for flatpanel or cone-beam small-animal CT instruments. The proposed algorithm is easy to implement and makes external gating hardware completely superfluous in smallanimal CT imaging. Since the external gating hardware typically has to interface with the CT hardware in order to phase-stamp each projection, a secondary benefit of the proposed scheme is that it makes gating fully independent from the CT hardware. This means that gating can be performed as a post-processing step, without any changes to the existing CT hardware. Our experience to date shows that this method is quite robust and reliable as far as changes in the examination environment are concerned. It also enables acquisition of 4D functional respiratory data. Finally, it may even be possible to derive a gating signal from several animals that are examined by CT simultaneously, allowing a higher throughput of animals. We expect that all these salient features will lead to widespread adoption of this gating method in small-animal CT. Acknowledgment The work was supported by the trans-regional grant “Vascular Differentiation and Remodelling” of the German Research Foundation (DFG). We thank Karin Leotta for her excellent technical assistance during data acquisition and post processing. References 1. Buliev IG, Badea CT, Kolitsi Z, Pallikarakis N (2003) Estimation of the heart respiratory motion with applications for cone beam computed tomography imaging: a simulation study. IEEE Trans Inf Technol Biomed 7:404–411 2. Cody DD, Nelson CL, Bradley WM, Wislez M, Juroske D, Price RE, Zhou X, Bekele BN, Kurie JM (2004) Murine lung tumor measurement using respiratory-gated micro-computed tomography. Invest Radiol 40:263–269 3. Walters EB, Panda K, Bankson JA, Brown E, Cody DD (2004) Improved method of in vivo respiratory-gated micro-CT imaging. Phys Med Biol 49:4163–4172 4. Ford NL, Nikolov HN, Norley CJ, Thornton MM, Foster PJ, Drangova M, Holdsworth DW (2005) Prospective respiratory-gated micro-CT of free breathing rodents. Med Phys 32:2888–2898 5. Hu J, Haworth ST, Molthen RC, Dawson CA (2004) Dynamic small animal lung imaging via a postacquisition respiratory gating technique using micro-cone beam computed tomography. Acad Radiol 11:961–970 6. Drangova M, Ford NL, Detombe SA, Wheatley AR, Holdsworth DW (2007) Fast retrospectively gated quantitative four-dimensional (4D) cardiac micro computed tomography imaging of freebreathing mice. Invest Radiol 42:85–94 7. Bartling S, Stiller W, Grasruck M, Schmidt B, Peschke P, Semmler W, Kiessling F (2007) Retrospective motion-gating in small animal CT of mice and rats. Invest Radiol 42:704–714 8. Kiessling F, Greschus S, Lichy MP, Bock M, Fink C, Vosseler S, Moll J, Mueller MM, Fusenig NE, Traupe H, Semmler W (2004) Volumetric computed tomography (VCT): a new technology for noninvasive, high-resolution monitoring of tumor angiogenesis. Nat Med 10:1133–1138 9. Zaporozhan J, Ley S, Unterhinninghofen R, Saito Y, Fabel-Schulte M, Keller S, Szabo G, Kauczor HU (2006) Freebreathing. -dimensional computed tomography of the lung using prospective respiratory gating: charge-coupled device camera and laser sensor device in an animal experiment. Invest Radiol 41:468–475 10. Ritchie CJ, Hsieh J, Gard MF, Godwin JD, Kim Y, Crawford CR (1994) Predictive respiratory gating: a new method to reduce motion artifacts on CT scans. Radiology 190:847–852 11. Kachelrieß M, Sennst DA, Maxlmoser W, Kalender WA (2002) Kymogram detection and kymogram-correlated image reconstruction from subsecond spiral computed tomography scans of the heart. Med Phys 29:1489–1503 12. Manzke R, Kohler T, Nielsen T, Hawkes D, Grass M (2004) Automatic phase determination for retrospectively gated cardiac CT. Med Phys 31:3345– 3362 13. Sonke JJ, Zijp L, Remeijer P, van Herk M (2005) Respiratory correlated cone beam CT. Med Phys 32:1176–1186 14. Zeng R, Fessler JA, Balter JM (2005) Respiratory motion estimation from slowly rotating x-ray projections: theory and simulation. Med Phys 32:984– 991 15. Gupta R, Grasruck M, Suess C, Bartling SH, Schmidt B, Stierstorfer K, Popescu S, Brady T, Flohr T (2006) Ultra-high resolution flat-panel volume CT: fundamental principles, design architecture, and system characterization. Eur Radiol 30:337–343 16. Feldkamp LA, Davis LC, Kress JW (1984) Practical cone-beam algorithm. J Opt Soc Am 1:612–619 17. Bartling S, Stiller W, Semmler W, Kiessling F (2007) Small animal Computed Tomography Imaging. CMIR 3:45– 59 18. Wolf I, Vetter M, Wegner I, Bottger T, Nolden M, Schobinger M, Hastenteufel M, Kunert T, Meinzer HP (2004) The medical imaging interaction toolkit (MITK): a toolkit facilitating the creation of interactive software by extending VTK and ITK. Proceedings of SPIE 5367:16–27 19. De Man B, Edic P, Basu S (2005) An iterative algorithm for time-resolved reconstruction of a CT scan of a beating heart. Proceedings of Fully 3D, The 8th international meeting on fully threedimensional image reconstruction in radiology and nuclear medicine: pp 356–359 1384 20. Mai W, Badea CT, Wheeler CT, Hedlund LW, Johnson GA (2005) Effects of breathing and cardiac motion on spatial resolution in the microscopic imaging of rodents. Magn Reson Med 53:858–665 21. Sharp P, La Regina M, LaRegina M (1998) The Laboratory Rat. CRC Press, Boca Raton, FL, USA 22. Krause MF, Hoehn T (1999) Decrease in lung volume depends on endexpiratory pressure in a rabbit model of airway lavage. Respiration 66:259–264 23. Zhu GF, Zhang W, Zong H, Liang Y (2006) Effects of continuous tracheal gas insufflation during pressure limited ventilation on pulmonary surfactant in rabbits with acute lung injury. Chin Med J (Engl) 119:1415–1420