Download invited review-image registration in veterinary radiation oncology

1
IMAGE REGISTRATION IN VETERINARY RADIATION ONCOLOGY: INDICATIONS,
2
IMPLICATIONS AND FUTURE ADVANCES
3
4
Yang Feng1, 2, Jessica Lawrence2, Kun Cheng1, Dean Montgomery1, Lisa Forrest3, Duncan B.
5
McLaren1, Stephen McLaughlin4, David J. Argyle2, William H. Nailon1
6
7
1
8
Western General Hospital, Edinburgh, UK. 2The University of Edinburgh, Royal (Dick) School
9
of Veterinary Studies and Roslin Institute, UK. 3The University of Wisconsin-Madison,
10
Department of Surgical Sciences, 2015 Linden Drive, Madison WI, USA. 4Heriot-Watt
11
University, School of Engineering and Physical Sciences, Edinburgh, EH14 4AS, UK.
The University of Edinburgh, Department of Oncology Physics, Edinburgh Cancer Centre,
12
13
Keywords: Oncology, Image Registration, Computed tomography, Radiation therapy, Magnetic
14
resonance imaging
15
Running Head: Image Registration in Radiation Oncology
16
Funding Sources: This work was generously supported by NHS Lothian, Edinburgh and
17
Lothians Health Foundation (charity number SC007342), the James Clerk Maxwell Foundation,
18
the Jamie King Uro-Oncology Endowment Fund, a University of Edinburgh Darwin Award and
19
a University of Edinburgh Individual Stipend.
20
Previous Abstracts: Material has not previously been presented.
1
21
Abstract
22
The field of veterinary radiation therapy (RT) has gained substantial momentum in recent
23
decades with significant advances in conformal treatment planning, image-guided radiation
24
therapy (IGRT) and intensity-modulated (IMRT) techniques. At the root of these advancements
25
lie improvements in tumor imaging, image alignment (registration), target volume delineation,
26
and identification of critical structures. Image registration has been widely used to combine
27
information from multi-modality images such as computerized tomography (CT), magnetic
28
resonance imaging (MRI) and positron emission tomography (PET) to improve the accuracy of
29
radiation delivery and reliably identify tumor-bearing areas. Many different techniques have
30
been applied in image registration. This review provides an overview of medical image
31
registration in RT and its applications in veterinary oncology. A summary of the most commonly
32
used approaches in human and veterinary medicine is presented along with their current use in
33
IGRT and adaptive radiation therapy (ART). It is important to realize that registration does not
34
guarantee that target volumes, such as the gross tumor volume (GTV), are correctly identified on
35
the image being registered, as limitations unique to registration algorithms exist. Research
36
involving novel registration-frameworks for automatic segmentation of tumor volumes is
37
ongoing and comparative oncology programs offer a unique opportunity to test the efficacy of
38
proposed algorithms.
39
40
2
41
Introduction
42
Imaging data from multiple anatomical and functional imaging studies is becoming a routine
43
component of veterinary patient management for a variety of medical and surgical conditions.
44
Spanning from initial diagnosis to determining therapeutic options to assessing response or
45
recrudescence of disease, these data help direct decisions regarding disease management,
46
efficacy of treatment and patient outcome. While computed tomography (CT) and magnetic
47
resonance (MR) imaging are readily available in veterinary medicine, novel imaging techniques
48
such as molecular imaging offer complementary information to aid in disease recognition, extent
49
and treatment planning. In order to optimally use information gathered from these various
50
techniques, the data must be easily compared despite differences in image acquisition and
51
presentation. Image registration is defined as the process of aligning two or more images from
52
the same or different imaging modalities to allow for data mapping (data fusion) and
53
interpretation. While registration carries importance across imaging disciplines, the purpose of
54
this review is to provide a broad overview of image registration and data fusion techniques used
55
in radiation therapy (RT), given the high frequency with which it is utilized and radiation
56
oncologists’ reliance on registered images. While RT is the focus of the review, the principles
57
outlined for registration techniques can be broadly applied to non-oncologic applications.
58
It is estimated that greater than 60% of human patients with cancer will receive RT and
59
although there is no similar strong data in dogs and cats, RT has become a common therapeutic
60
modality in veterinary oncology.1 There is a large body of evidence supporting the use of RT for
3
61
optimal local tumor control, therefore identification of the tumor volume is of utmost importance
62
when determining a treatment plan or evaluating changes in tumor volume. In practice, medical
63
images acquired from different imaging modalities are used to guide the entire RT process from
64
the initial treatment plan to fractionated radiation delivery through to dose verification. While
65
radiation oncologists have always been guided by images in some form, the advent of image-
66
guided radiation therapy (IGRT) has revolutionized how integral images are to modern radiation
67
oncology. In a broad sense, IGRT may reflect any aspect of RT that utilizes imaging to improve
68
treatment, such as weekly or twice weekly portal imaging. In a stricter sense, IGRT refers to
69
contemporaneous functional and structural imaging to improve target delineation, adjust for
70
target motion and/or uncertainties in patient positioning, and potentially adapt treatment to the
71
response of the tumor during adaptive radiation therapy (ART).2 While radiation oncologists are
72
highly dependent on image registration to ensure that IGRT is successful at targeting tumor and
73
limiting dose to adjacent normal tissue, the underlying process is complex and often overlooked.
74
A typical IGRT process, using veterinary images as an example, where image registration is vital
75
in order to merge information from multi-modality images and therefore provide an accurate
76
guide for radiation delivery is illustrated in Fig. 1.
77
Image registration provides a geometric transform that makes it possible to map
78
information between the images, often with sub-pixel resolution. As the use of multimodality
79
image data in RT increases, medical image registration is essential to combine the information
80
from each modality. As a result, it has become a very active area of research.3 Images utilized in
4
81
RT can be registered to obtain comprehensive information regardless of patient positioning, time
82
point with respect to therapy, or type of imaging acquisition.. Registration may also be used to
83
combine multiple images from the same imaging modality,4 or to combine information from
84
multiple modalities such as CT, MR imaging, positron emission tomography (PET) and single-
85
photon emission computed tomography (SPECT).5 For the purpose of this review, areference
86
image is defined as the source image containing the reference information, while a target image
87
is defined as the movable image spatially matched to the reference. After registration,
88
information from the reference image, such as contoured structures, can be used on the target
89
image.
90
Image registration methods have been widely used both in human medicine6-11 and in
91
preclinical studies12-18 to improve matching between images acquired on different modalities
92
such as PET andMR,19 SPECT andCT,20 and PET andCT.21 Registration approaches applied to
93
humans are also suitable for companion animals.22-31 and may prove useful in comparative
94
oncology, which refers to the study of cancer etiology, biology and treatment in companion
95
animals that develop spontaneously-occurring cancer.1, 32-34 Comparative oncology presents a
96
unique opportunity to improve RT for both human and veterinary patient groups given the
97
number of similarities in tumor biology, imaging techniques and therapeutic modalities. Image
98
registration methodologies, regardless of the species imaged, will help facilitate this by finding
99
the optimal geometric transformation between corresponding image data.35
5
100
In general, an image registration algorithm can be divided into four parts: image
101
conditioning, geometric transformation, similarity function and optimization (Fig. 2). Any step
102
that alters the original data to make it more suitable for applying image registration is known as
103
image conditioning. Due to the diversity of the methods used for this, a comprehensive review of
104
image conditioning falls outside the scope of this manuscript but may be found elsewhere.5, 36-37
105
106
Image Registration
107
Table 1 provides a comprehensive list of the most commonly used terminology in image
108
registration.5 Key technical details of the transformation function, similarity function and
109
optimization process (Fig. 2) are introduced in this section.
110
111
Transformation
112
As it is common for patient orientation and immobilization to vary between imaging studies, the
113
first essential component of registration is transformation. This describes the geometrical shift
114
that is required for the target image to match it to the reference image. Rigid transformation
115
methods are most commonly used and are usually applied on images that have no distortion,
116
such as CT to CT. Generally, a rigid geometric transformation can be achieved by translation and
117
rotation and is commonly used to match between bones on medical images, such as the skull.38-39
118
An affine transformation, which is an extension of the rigid transformation, allows for rotations,
6
119
translations, scaling and shearing.40 Rigid transformations can also be applied before a non-rigid
120
registration as an image conditioning step.41-42
121
The assumption that rigid movement of anatomy occurs globally is incorrect in any
122
number of situations, therefore limiting widespread use of rigid registrations for sites other than
123
the head.40 Most non-rigid registration methods are based on deformable models. These range in
124
complexity and may vary from simple with relatively few parameters tocomplicated where each
125
point or voxel moves independently.40 There are two directions in deformable registration: free
126
form deformable registration (often abbreviated FFD) and guided deformable registration, which
127
are controlled by models based on prior knowledge of the registered objects or organs. The
128
fundamental difference between them is that the free form deformable registration allows any
129
deformations by moving the positions of its control grid. The most commonly used guided
130
deformable registration methods are elastic-based and flow-based. Elastic-based methods treat
131
organs as elastic solids and define two forces: internal forces that oppose the deformation and
132
external forces that try to deform the images. The best transform can be obtained by finding
133
equilibrium between both internal and external forces. Flow-based methods such as fluid flow
134
and optical flow treat registration as a motion problem and therefore achieves the best match by
135
meeting a pre-set constraint in a physical model. The application of the popular B-spline based
136
FFD method registering pre- and post-treatment CT images of a dog with a sinonasal tumor is
137
demonstrated in Fig 3.43 A similar method called thin-plate spline (often abbreviated TPS) can
7
138
also generate FFD registration. However, compared to the B-spine method this approach is
139
limited by the fact that the deformation is applied globally. 44
140
141
Similarity Function
142
The transformation, or alignment, of the image datasets is assessed by measuring the similarity
143
between them.40 There are two predominant similarity measurements: intensity-based and
144
feature-based measurements. The most popular intensity-based similarity function is based on
145
intensity difference or intensity correlation. Feature-based similarity functions depend on the
146
feature structure extracted, such as anatomic structures (bones) or artificial landmarks (fiducials).
147
Most feature-based algorithms use points, lines, or surfaces for matching.40, 45-46 A minimum of
148
three or four pairs of points are required in order to compute the rotations and translations for a
149
rigid or affine transformation.40 Unlike point matching, line and surface matching do not require
150
a one-to-one match between images but rather attempt to maximize the overlap between
151
equivalent lines and surfaces such as the skull surface, ribs or pelvis.40, 45-46 The transformation
152
process and the corresponding similarity calculation used in the registration of images from a
153
dog with a nasal tumor is illustrated in Fig 4.
154
155
Optimization
156
Optimization is the process used to search for a numerical value produced by the similarity
157
function, which is indicative of when the best match between the images has been found. The
8
158
goal of the optimization algorithm is to find a maximum or minimum value of the similarity
159
measure accepted; it is common for optimal registration to be accomplished when the similarity
160
measures are defined by their minimal value. It is an important step in a registration method
161
because the similarity function can often produce several local minima, which produces sub-
162
optimal results.
163
164
Assessment
165
The obvious motivation for registering images from different studies is to map clinically useful
166
information from one study onto another, for example in treatment planning prior to radiation
167
delivery. If data can be successfully fused, radiation oncologists can then map tumor volumes
168
such as the GTV, clinical target volume (CTV) and planning target volume (PTV) or organs at
169
risk (OAR) directly onto the CT needed for dose calculations (structure mapping).40 Therefore,
170
relying on the ability of a bespoke software system to adequately register images is of utmost
171
importance in order to adhere to current radiation guidelines – namely to appropriately outline
172
tumor volumes and limit normal tissue toxicity. In this vain, it is important to realize the
173
difficulties in accurately validating the performance of a complex registration method. In some
174
cases, a registration algorithm can be assessed using phantom data,47 but this is not commonly
175
done as performance on phantoms cannot ensure comparable performance on clinical cases. A
176
commonly used approach is to use synthetic images on which manual definition of
177
corresponding points such as fiducial markers provides a straightforward assessment of a
9
178
registration method.48-49 The Dice coefficient, which measures the overlap between clinical and
179
registered contours, places a numerical value on the registration accuracy.50-51 Within the range 0%
180
to 100%, a Dice value of 100% indicates excellent geometric agreement while a Dice value of 0%
181
represents poor geometric agreement (Fig. 5). The Jaccard index and Tanimoto coefficient (both
182
vary from 0% to 100%), which can be calculated directly from the Dice coefficient, are also used
183
to measure the similarity between regions. However, they need prior knowledge (contours of the
184
same region) on both reference and target images.
185
Once the registration process has been completed, there are a number of techniques that
186
can be used to visualize fused data, including the use of overlays, pseudo-coloring, grey scale
187
coloring, and side-by-side display of anatomic planes.40 This allows the radiation oncologist to
188
evaluate the fused images and to map calculated 3D dose distributions from the CT treatment
189
plan to the coordinate system of another imaging study, such as MR. Not only is this helpful to
190
define the tumor and normal tissue dose for tissues that are not optimally imaged with CT (such
191
as canine or feline intracranial tumors), but the fusion of 3D dose distributions to post-treatment
192
imaging scans to improve detection and understanding of radiation changes over time (Fig. 5).
193
As veterinary radiation oncology advances into the realm of IGRT and ART with the use of
194
volumetric imaging systems (on-board kilovoltage CT or megavoltage CT units), the concept of
195
accurate registration becomes increasingly critical.
196
197
Current State of Image Registration in Veterinary Radiation Oncology
10
198
Image registration and data fusion are useful in veterinary radiation oncology with most
199
modern treatment planning systems now facilitating the use of a secondary datasets for target and
200
OAR delineation. It is also possible to map dose information to the secondary dataset following
201
an appropriate registration.
202
Dose distributions generated for conformal radiation therapy (CRT) have been accurately
203
predicted with CT-based treatment planning systems and are already widely used in veterinary
204
oncology.53 CT data includes the Hounsfield units as a linear transformation of radiation beam
205
attenuation that varies with electron densities of materials as the beam progresses through tissue.
206
Most RT planning software obtains the relative electron density from the relationship between
207
the linear attenuation coefficients and Hounsfield unit values in order to determine heterogeneity
208
in the dog or cat’s body. Non-contrast CT images are typically utilized for treatment planning,
209
given that contrast agents are high-Z radio-opaque materials, which would attenuate a radiation
210
beam more than normal, resulting in higher than normal electron density. Post-contrast CT data
211
used for a treatment plan would theoretically give rise to higher monitor unit (MU) values, and
212
therefore radiation dose, compared to MU values taken from calculating a plan using pre-contrast
213
CT data. As veterinary radiation oncologists often use commercially available human treatment
214
planning software, automatic registration methods may be built-in to the software, although other
215
“in house” registration algorithms may be adapted for use.22, 24-29 One of the most common
216
automatic registrations used in routine treatment planning is the merging of pre-contrast and
217
post-contrast CT data in order to delineate the GTV, CTV, PTV and OAR. Because regions of
11
218
interest are often best identified on post contrast CT images, many commercial software
219
programs (such as Eclipse™, Varian Medical Systems, Palo Alto, used at the authors’
220
institutions) automatically permit DICOM-coordinated registration and facilitate contouring on
221
fused multimodality images. However, there are often limited registration capabilities between
222
multimodal image acquisitions; our current planning system is capable of a fully automatic,
223
mutual-information-based rigid registration but results are often unsatisfactory between
224
modalities (i.e. CTs obtained with different DICOM origins or MRI and CT data acquired with
225
vastly different animal positioning). Automatic registration partially eliminates a potential
226
source of error as radiation oncologists historically may have estimated tumor volume on the pre-
227
contrast CT data based on visual examination of alternative imaging studies. Upon evaluation of
228
the veterinary literature to date, it is often difficult to firmly understand how tumor volumes were
229
identified prior to treatment; for example in one study evaluating RT for canine intracranial
230
tumors, the GTV was defined on planning CT scans by the contrast-enhancing area on CT or MR
231
data.54 It is presumed that contours for the GTV were drawn on pre-contrast CT images after
232
evaluation of diagnostic scans, however this presumption may be incorrect. In most other
233
veterinary studies, it is unclear how the GTV was derived, as irradiated volumes have been
234
inconsistently defined.52 The delineation of target volumes is paramount to effective RT, as
235
geographic miss could occur if the GTV, CTV and PTV are ineffectually contoured and/or
236
unexpected toxicity could result from a poorly defined OAR. Proper radiation reporting is also
237
critical to permit reproducibility of clinical studies across institutions.52 It is also important to
12
238
note that dose distributions generated by treatment planning are estimated dose distributions,
239
thereby introducing an additional uncertainty to planning. The variation in registration methods
240
introduces bias that could lead to erroneous conclusions about the extent and location of various
241
volumes compounded by variations in dose distribution algorithms. As veterinary reporting of
242
RT planning improves, information regarding registration should be included.
243
In addition to RT planning, image registration has made its way into the routine
244
assessment of daily patient positioning, enabling better tumor targeting.25-29, 55-57 One of the key
245
concepts of IMRT, stereotactic radiation therapy (SRT) or radiosurgery (SRS) and IGRT
246
planning is that tight dose constraints are placed around the PTV provided the clinician is
247
confident that the animal’s positioning will be reproducible during treatment. One of the most
248
common practices immediately prior to treatment delivery is to generate a pair of orthogonal
249
digitally reconstructed radiographs (DRRs) from the planning CT and to register simulated
250
radiographs with MV radiographs acquired by a flat-panel imager (electronic portal imaging
251
devices or EPIDs) attached to the linear accelerator (Fig. 6).55,56 More advanced on-board
252
imaging equipment involves cone-beam CT systems or helical CT systems built into the
253
treatment unit.25-29, 57 Automated image registration at the treatment unit can determine the
254
rotations and translations required to align the datasets, thus decreasing some of the variability
255
with daily patient setup and intra-patient shifts.25-29, 40 Challenges still exist with registration of
256
different modalities with different positioning and therefore fiducials and other landmarks may
257
be needed to improve alignment (Fig. 7). Dose delivered to patients throughout a treatment
13
258
protocol also rely on accurate patient positioning and registration (i.e. daily cone beam CT
259
registered with treatment-planning CT) in order to sum dose delivered over time; monitoring of
260
dose deviations from the expected distribution allows for intra-treatment adjustments in order to
261
better target tumor yet limit normal tissue dose. Little veterinary literature has evaluated the
262
utility or validity of various registration algorithms other than an early study demonstrating an
263
in-house computer program that was developed for brain imaging.22 This study highlighted the
264
widespread potential of image registration by evaluating corresponding images from CT and
265
MR, CT images before and after surgical treatment, and CT and post-mortem cryosection images
266
in dogs or cats.22 Currently at our institution, for follow-up imaging, standardized response
267
criteria are applied when assessing tumor response following treatment.58 Responses for tumors
268
of the head and neck (intracranial, sino-nasal, cervical) are often assessed by the radiology and
269
radiation oncology team on serial CT scans, without consideration for image registration over
270
time, which may not accurately reflect the response of the tumor or the assessment of treatment-
271
induced toxicity.27 Toxicities are often inferred by tissue changes, however easier
272
implementation of image registration in the clinic may provide more objective information
273
regarding tumor and normal tissue changes following radiation, particularly as varying
274
fractionated protocols are used across veterinary radiation oncologists.52,53,59,60
275
With the incorporation of functional imaging into veterinary oncology, determination of
276
structural and, indirectly, pathological changes may be better assessed. Changes such as fibrosis
277
and tumor recurrence can be difficult to distinguish on CT and MR; functional imaging with PET
14
278
may help clarify differences and alter future therapy. Currently in veterinary medicine, 2-deoxy-
279
2-[18F]-fluoro-deoxyglucose (FDG), 3’-deoxy-3’-[18F]-fluorothymidine (FLT), and copper(II)-
280
diacetyl-bis (N4-methylthiosemicarbazone) (Cu-ATSM) have been used to assess various
281
neoplastic conditions in dogs and cats.61-70 In particular, several studies have evaluated functional
282
imaging over time in an attempt to correlate pre-treatment scans to post-treatment scans; ideal
283
comparisons would require registration of image datasets to avoid interpretation errors.67-70 One
284
study that assessed pre-radiation therapy FDG, FLT and Cu-ATSM PET/CT scans to post-
285
treatment FDG PET/CT scans demonstrated the complexity and utility of registration methods in
286
dogs with sinonasal tumors.70 Dogs in this study underwent a radiation-planning CT scan that
287
served as the reference dataset; bony anatomy from pre-radiation PET/CT scans was registered
288
to the bony anatomy of the radiation-planning CT; rigid registration using cross-correlation
289
resulted in affine transformations that were applied to the corresponding PET images. The
290
translated and rotated PET matrices were subsequently resampled using a spline filter defined by
291
the reference PET image that was obtained at the same time as the treatment-planning CT.70
292
While complex, the registration process ensured that each voxel index of the pre-treatment
293
PET/CT scans corresponded as closely as possible to the same region in the post-treatment FDG
294
PET image for each patient, enabling assessment of changes in tracer uptake.70 Imaging and
295
registration advances will undoubtedly lead to more work being performed in a similar fashion
296
and may alter current criteria for response assessment and/or response prediction.
297
15
298
Proposed Advances in Veterinary Image Registration
299
In general, RT is delivered to patients over the course of several weeks in both human and
300
veterinary oncology. While the GTV, CTV and PTV are currently considered standard targets in
301
radiation oncology, the PTV is created to make sure that dose plans are robust to uncertainties.71
302
The GTV and CTV are oncological-anatomical concepts, which are heavily dependent on the
303
experience and judgment of the radiation oncologist and radiologist. The PTV, on the other hand,
304
is a geometrical scheme used for treatment planning in order to ensure that the prescribed dose
305
targets the CTV.71 Alternatively, one might view the PTV as decreasing the risk that patient
306
positioning, setup uncertainties, and OAR or target position uncertainties will lead to a severe
307
under-dosage of the CTV. As the impact of setup uncertainties decreases with IGRT, the CTV to
308
PTV margins may be reduced, thus decreasing normal tissue toxicity. Taking this a step further,
309
if diagnostic imaging, interpretation, and registration with the initial treatment plan are
310
performed frequently while a patient is on therapy, the treatment plan can be “adapted” to
311
optimize tumor control and minimize normal tissue damage.
312
ART is defined as changing a radiation plan during RT to account for anatomy or biology
313
changes of the patients.72 Studies have shown clear advantages and gains in clinical application
314
of adaptive radiation therapy. However, clinical implementation is limited as ART is time-
315
consuming and places an additional workload on clinicians to quickly contour follow-up images
316
manually. Adaptive planning could be used at different fractions (off-line) and at different time
317
points within one fraction (on-line).73 Identifying changes in the GTV between images acquired
16
318
for the purpose of RT planning and images acquired for the purpose of assessing response to RT
319
is challenging. This is mainly because of: 1. uncertainties of soft tissues and internal organ
320
motion;74 2. the response of the GTV to radiation or other treatments is difficult to predict, which
321
may cause anatomical changes;75 3. distortions or contrast variability between different
322
modalities, which makes it difficult to register the information between multi-modality images.76
323
To overcome these difficulties, image guided adaptive radiation therapy (IGART) is the most
324
reliable method for a comprehensive ART approach.77 Research into known, documented
325
uncertainties in soft tissue and organ motions in prostate,78 head and neck79 and lung80 is well
326
developed and directly impact ART approaches. Furthermore, registration methods that combine
327
information and realize deformable mapping between structural and functional images are
328
essential to make individualized IGART rapid and feasible.81
329
330
331
Conclusion
In its simplest context, image registration simply refers to the comparison of one dataset
332
to another. The complexity behind making quantitative use of this imaging data has been
333
reviewed here, illustrating that it is necessary to determine the transformations needed to relate
334
coordinates of one dataset to another. The impact of image registration on RT from treatment
335
planning to image-guided delivery to post-treatment follow-up is enormous and has already
336
impacted veterinary patients. Novel registration frameworks may be ideally tested on companion
337
animals undergoing RT given similarities between human and veterinary radiation oncology.
17
338
Ultimately optimized image registration and data fusion techniques should aid in targeting tumor
339
volumes and organs-at-risk, improve response assessment and increase the availability and utility
340
of an efficient ART approach.
341
342
Acknowledgements:
343
We are grateful for the generous support from NHS Lothian, Edinburgh and Lothians Health
344
Foundation (charity number SC007342), the James Clerk Maxwell Foundation, the Jamie King
345
Uro-Oncology Endowment Fund, a University of Edinburgh Darwin Award and a University of
346
Edinburgh Individual Stipend.
347
References
348
1. Paoloni M, Khanna C. Translation of new cancer treatments from pet dogs to humans.
349
Nat Rev Cancer. 2008;8:147-56.
350
2. Mell LK, Song WY, Pawlicki T, Mundt AJ. Image-guided radiation therapy. In: Halperin EC,
351
Wazer DE, Perez CA, Brady LW (eds): Principles and Practice of Radiation Oncology, 6th
352
ed. Philadelphia: Lippincott Williams and Wilkins, 2013;246-76.
353
354
3. Pluim
JPM,
Fitzpatrick
JM.
Image
registration.
IEEE
Trans
Med
Imaging.
2003;22(11):1341–3.
355
4. Gooding MJ, Rajpoot K, Mitchell S, Chamberlain P, Kennedy SH, Noble JA. Investigation
356
into the fusion of multiple 4-D fetal echocardiography images to improve image quality.
357
Ultrasound Med Biol. 2010;36(6):957-66.
18
358
359
360
361
362
363
5. Maintz JBA, Viergever MA. A survey of medical image registration. Med Image Anal.
1998;2(1):1-36.
6. Freeborough PA, Fox NC. Modeling brain deformations in Alzheimer disease by fluid
registration of serial 3D MR images. J Comput Assist Tomogr. 1998;22(5):838–43.
7. Collins DL, Evans AC. ANIMAL: validation and applications of non-linear registrationbased segmentation. Intern J Pattern Recognit Artif Intell. 1997;11(8):1271–94.
364
8. Lavely WC, Scarfone C, Cevikalp H, Li R, Byrne DW, Cmelak AJ, et al. Phantom validation
365
of co-registration of PET and CT for image-guided radiotherapy. Med Phys.
366
2004;31(4):1083–92.
367
9. Maurer CR, Fitzpatrick JM, Wang MY, Galloway RL, Maciunas RJ, Allen GS. Registration of
368
head volume images using implantable fiducial markers. IEEE Trans Med Imaging.
369
1997;16(4):447–62.
370
10. Grosland NM, Bafna R, Magnotta VA. Automated hexahedral meshing of anatomic
371
structures using deformable registration. Comput Methods Biomech Biomed Engin.
372
2009;12(1):35-43.
373
374
11. Maksimov D, Hesser J, Brockmann C, Jochum S, Dietz T, Schnitzer A. Graph-Matching
based CTA. IEEE Trans Med Imaging. 2009;28(12):1940–1954.
375
12. Martin B, Julien M, Jouke D, Tobias DH, Axel WW, Ivo Q, et al. Atlas-based whole-body
376
segmentation of mice from low-contrast Micro-CT data. Med Image Anal.
19
377
2010;14(5):723-37.
378
13. Anand AJ, Abhijit JC, Changqing L, Joyita D, Simon RC, David WS, et al. DigiWarp: a
379
method for deformable mouse atlas warping to surface topographic data. Phys Med Biol.
380
2010;55(20):6197–214.
381
14. Johnson CA, Seidel J, Carson RE, Gandler WR, Sofer A, Green MV. Evaluation of 3D
382
reconstruction algorithms for a small animal PET camera. IEEE Trans Nucl Sci.
383
1997;44(3):1303-8.
384
15. Green MV, Seidel J, Vaquero JJ, Jagoda E, Lee I, Eckelman WC. High resolution PET, SPECT
385
and projection imaging in small animals. Comput Med Imaging Graph. 2001;25(2):79-
386
86.
387
16. Wildeman M, Baiker M, Reinders MJT, Lowik CWGM, Reiber JHC, Lelieveldt BPF. 2D/3D
388
registration of micro-CT data to multi-view photographs based on a 3D distance map.
389
Proc IEEE Int Symp Biomed Imaging. In: Proceedings of IEEE International Symposium
390
on Biomedical Imaging; 2009; Boston, Massachusetts, USA.
391
17. Zheng X, Xishi H, Xiaobo Z, Youxian S, Vasilis N, Stephen TCW. Registration of 3-D CT and
392
2-D flat images of mouse via affine transformation. IEEE Trans Inf Technol Biomed.
393
2008;12(5):569-78.
394
18. Yang CC, Wu TH, Lin MH, Huang YH, Guo WY, Chen CL, et al. Multimodality imaging
395
combination in small animal via point-based registration. Nucl Instrum Methods Phys
20
396
Res A. 2006;569(2):240-4.Johnson HJ, McCormick M. ITK software guide. 2013.
397
19. Hayakawa N, Uemura K, Ishiwata K, Shimada Y, Ogi N, Nagaoka T. A PET-MRI
398
registration technique for PET studies of the rat brain. Nucl Med Biol. 2000;27(2):121-5.
399
20. Sakdinawat AE, Iwata K, Hwang AB, Tang HR, Wong KH, Hasegawa BH. Development of
400
external fiducial markers for image registration in small animal SPECT/CT. IEEE Nucl
401
Sci Symp Conf Rec. 2002;2:842-5.
402
21. Woo SK, Kim KM, Lee T, Kim JY, Jung JH, Woo KS, et al. Experimental condition and
403
registration method for the tumor detection of lung metastasis small animal PET and CT
404
whole body images. IEEE Nucl Sci Symp Conf Rec. 2007;5:3372-5.
405
406
22. Yubei S, Koji U, Babak AA, Tsukasa N, Kiichi I, Hinako T, et al. Application of PET-MRI
registration techniques to cat brain imaging. J Neurosci Methods. 2000;101(1):1-7.
407
23. Bottcher P, Maierl J, Hecht S, Matis U, Leibich HG. Automatic image registration of three-
408
dimensional images of the head of cats and dogs by use of maximization of mutual
409
information. Am J Vet Res. 2004;65:1680-7.
410
24. Shimada Y, Uemura K, Ardenkani BA, Nagaoka T, Ishiwata K, Toyama H, et al. Application
411
of PET-MRI registration techniques to cat brain imaging. J Neurosci Methods.
412
2000;101:1-7.
413
25. Nolan MW, Kogan L, Griffin LR, Custis Jt, Harmon JF, Biller BJ, et al. Intensity-modulated
414
and image-guided radiation therapy for treatment of genitourinary carcinomas in dogs.
21
415
J Vet Intern Med. 2012;26:987-95.
416
26. Harmon J, Yoshikawa H, Custis J, LaRue S. Evlauation of canine prostate intrafractional
417
motion using serial cone beam computed tomography imaging. Vet Radiol Ultrasound.
418
2013;54:93-8.
419
27. Murphy S, Gutierrez A, Lawrence J, Bjorling D, Mackie T, Forrest L. Laparoscopically
420
implanted tissue expander radiotherapy in canine transitional cell carcinoma. Vet Radiol
421
Ultrasound. 2008;49:400-5.
422
28. Lawrence JA, Forrest LJ, Turek MM, Miller PE, Mackie TR, Jaradat HA, et al. Proof of
423
principle of ocular sparing in dogs with sinonasal tumors treated with intensity-
424
modulated radiation therapy. Vet Radiol Ultrasound. 2012;51:561-70.
425
29. Kubicek LN, Seo S, Chappell RJ, Jeraj R, Forrest LJ. Helical tomotherapy setup variations
426
in canine nasal tumor patients immobilized with a bite block. Vet Radiol Ultrasound.
427
2012;53:474-81.
428
30. Forrest LJ, Mackie TR, Ruchala K, Turek M, Kapatoes J, Jaradat H, et al. The utility of
429
megavoltage computed tomography images from a helical tomotherapy system for
430
setup verification purposes. Int J Radiat Oncol Biol Phys. 2004;60:1639-44.
431
31. Kaczka DW, Cao K, Christensen GE, Bates JHT, Simon BA. Analysis of regional mechanics
432
in canine lung injury used forced oscillations and 3D image registration. Ann Biomed
433
Eng. 2011;39:1112-24.
22
434
435
436
437
32. Vail DM, MacEwen EG. Spontaneously occurring tumors of companion animals as
models for human cancer. 2000;18:781-92.
33. Khanna C, Lindblad-Toh K, Vail D, London C, Bergman P, Barber L, Breen M, et al. The
dog as a cancer model. Nat Biotechnol. 2006;24:1065-6.
438
34. Paoloni M, Webb C, Mazcko C, Cherba D, Hendricks W, Lana S, et al. Prospective
439
molecular profiling of canine cancers provides a clinically relevant comparative model
440
of evaluating personalized medicine (PMed) trials. PLoS ONE 2014;9:e90028.
441
doi:10.1371/journal.pone.0090028
442
443
35. Modersitzki J. Numerical Methods for Image Registration. Oxford University Press.
2004.
444
36. Murphy K, van Ginneken B, Klein S, Staring M, de Hoop BJ, Viergever MA, et al. Semi-
445
automatic construction of reference standards for evaluation of image registration. Med
446
Image Anal. 2011;15:71-84.
447
448
449
450
37. Markelj P, Tomazevic D, Likar B, Pernus F. A review of 3D/2D registration methods for
image-guided interventions. Med Image Anal. 2012;16:642-61.
38. Livyatan H, Yaniv Z, Joskowicz L. Gradient-based 2-D/3-D rigid registration of
fluoroscopic X-ray to CT. IEEE Trans Med Imaging. 2003;22(11):1395–406.
451
39. Andreetto M, Cortelazzo GM, Lucchese L. Frequency domain registration of computer
452
tomography data. In: John A, Gabriel T, editors. Proceedings of the 2nd International
23
453
Symposium on 3D Data Processing, Visualization and Transmission; 2004 Sep 6-9;
454
Thessaloniki, Greece.
455
456
40. Kessler ML. Image registration and data fusion in radiation therapy. Br J Radiol.
2006;79:S99-S108.
457
41. Mattes D, Haynor DR, Vesselle H, Lewellen TK, Eubank W. PET-CT image registration in
458
the chest using free-form deformations. IEEE Trans Med Imaging. 2003;22(1):120-8.
459
42. Hellier P, Barillot C. A hierarchical parametric algorithm for deformable multimodal
460
image registration. Comput Methods Programs Biomed. 2004;75(2):107-15.
461
43. Rueckert D, Sonoda LI, Hayes C, Hill DLG, Leach MO, Hawkes DJ. Non-rigid registration
462
using free-form deformations: application to breast MR images. IEEE Trans Med
463
Imaging. 1999;18(8):712–21.
464
44. Meyer CR, Boes JL, Kim B, Bland PH, Zasadny KR, Kison PV, et al. Demonstration of
465
accuracy and clinical versatility of mutual information for automatic multimodality
466
image fusion using affine and thin-plate spline warped geometric deformations. Med
467
Image Anal. 1997;1(3):195–206.
468
469
45. Kessler ML, Pitluck S, Petti PL, Castro JR. Integration of multimodality imaging data for
radiotherapy treatment planning. Int J Radiat Oncol Biol Phys. 1991;21:1653-67.
470
46. Pelizzari CA, Chen GT, Spelbring DR, Weichselbaum RR. Accurate three-dimensional
471
registration of CT, PET, and/or MR images of the brain. J Comput Assist Tomogr.
24
472
1989;13:20-26.
473
47. Studholme C, Constable RT, Duncan JS. Accurate alignment of functional EPI data to
474
anatomical MRI using a physics-based distortion model. IEEE Trans Med Imaging.
475
2000;19(11):1115–27.
476
48. Schnabel JA, Tanner C, Castellano-Smith AD, Degenhard A, Martin OL, Hose DR, et al.
477
Validation of nonrigid image registration using finite element methods: application to
478
breast MR images. IEEE Trans Med Imaging. 2003;22(2):238–47.
479
49. Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P. Multimodality medical
480
image registration by maximization of mutual information. IEEE Trans Med Imaging.
481
1997;16(2):187–98.
482
50. Alterovitz R, Goldberg K, Pouliot J, Hsu IC, Kim Y, Noworolski SM, et al. Registration of
483
MR prostate images with biomechanical modeling and nonlinear parameter estimation.
484
Med Phys. 2006;33(2):446-54.
485
51. Vercauteren T, Pennec X, Perchant A, Ayache N. Nonparametric diffeomorphic image
486
registration with the demons algorithm. In: Ayache N, Ourselin S, Maeder A, editors.
487
Proceedings of the 10th International Conference on Medical Image Computing and
488
Computer-Assisted Intervention; 2007; Brisbane, Australia.
489
52. Keyerleber MA, McEntee MC, Farrelly J, Podgorsak M. Completeness of reporting of
490
radiation therapy planning, dose, and delivery in veterinary radiation oncology
25
491
492
493
494
495
manuscripts from 2005-2010. Vet Radiol Ultrasound. 2012;53:221-30.
53. LaRue SM, Gordon IK. Radiation Therapy. In: Withrow SJ, Vail DM, Page RL (eds): Small
Animal Clinical Oncology, 5th ed. St Louis: Elsevier Inc; 2013:180-97.
54. Rohrer Bley C, Sumova A, Roos M, Kaser-Hotz B. Irradiation of brain tumors in dogs with
neurologic disease. J Vet Intern Med. 2005;19:849-54.
496
55. Kent MS, Gordon IK, Benavides I, Primas P, Young J. Assessment of the accuracy and
497
precision of a patient immobilization device for radiation therapy in canine head and
498
neck tumors. Vet Radiol Ultrasound. 2009;50:550-4.
499
500
56. McEntee MC. Portal radiography in veterinary radiation oncology: options and
considerations. Vet Radiol Ultrasound. 2009;49:S57-S61.
501
57. Nieset JR, Harmon JF, LaRue SM. Use of cone-beam computed tomography to
502
characterize daily urinary bladder variations during fractionated radiotherapy for
503
canine bladder cancer. Vet Radiol Ultrasound. 2011;52:580-8.
504
58. Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New
505
response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1).
506
Eur J Cancer. 2009;45:228-47.
507
508
509
59. Farrelly J, McEntee MC. A survey of veterinary radiation facilities in 2010. Vet Radiol
Ultrasound 2014;55:638-643.
60. McEntee MC. A survey of veterinary radiation facilities in the United States during 2001.
26
510
Vet Radiol Ultrasound 2004;45:476-479.
511
61. Randall E, Loeber S, Kraft S. Physiologic variants, benign processes, and artifacts from
512
106 canine and feline FDG-PET/computed tomography scans. Vet Radiol Ultrasound.
513
2014;55:213-26.
514
515
516
517
62. Hansen AE, McEvoy F, Engelholm SA, Law I, Kristensen AT. FDG PET/CT imaging in
canine cancer patients. Vet Radiol Ultrasound. 2011;52:201-6.
63. Leblanc AK, Jakoby BW, Townsend DW, Daniel GB. 18FDG-PET imaging in canine
lymphoma and cutaneous mast cell tumor. Vet Radiol Ultrasound. 2009;50:215-23.
518
64. Yoshikawa H, Randall EK, Kraft SL, LaRue SM. Comparison between 2-18F-fluoro-2-
519
deoxy-D-glucose positron emission tomography and contrast-enhanced computed
520
tomography for measuring gross tumor volume in cats with oral squamous cell
521
carcinoma. Vet Radiol Ultrasound. 2013;54:307-13.
522
65. Bradshaw TJ, Bowen SR, Jallow N, Forrest LJ, Jaraj R. Heterogeneity in intratumor
523
correlations of 18F-FDG, 18-FLT, and 61Cu-ATSM PET in canine sinonasal tumors. J Nucl
524
Med. 2013;54:1917-31.
525
526
66. LeBlanc AK, Peremans K. PET and SPECT imaging in veterinary medicine. Semin Nucl
Med. 2014;44:47-56.
527
67. LeBlanc AK, Miller AN, Galyon GD, Moyers TD, Long MJ, Stuckey AC, et al. Preliminary
528
evaluation of serial (18)FDG-PET/CT to assess response to toceranib phosphate therapy
27
529
in canine cancer. Vet Radiol Ultrasound. 2012;53:348-57.
530
68. Lawrence J, Vanderhoek M, Barbee D, Jeraj R, Tumas DB, Vail DM. Use of 3’-deoxy-3’-
531
[18F]fluorothymidine PET/CT for evaluating response to cytotoxic chemotherapy in
532
dogs with non-Hodgkin’s lymphoma. Vet Radiol Ultrasound. 2009;50:660-8.
533
69. Bradshaw TJ, Yip S, Jallow N, Forrest LJ, Jeraj R. Spatiotemporal stability of Cu-ATSM and
534
FLT positron emission tomography distributions during radiation therapy. Int J Radiat
535
Oncol Biol Phys. 2014; doi:10.1016/j.ijrobp.2014.02.016.
536
70. Bowen SR, Chappell RJ, Bentzen SM, Deveau MA, Forrest LJ, Jeraj R. Spatially resolved
537
regression analysis of pre-treatment FDG, FLT and Cu-ATSM PET from post-treatment
538
PDG PET: an exploratory study. Radiother Oncol. 2012;105:41-8.
539
71. ICRU Report 62. Prescribing, recording, and reporting photon beam therapy
540
(supplement to ICRU Report 50). Bethesda, USA: International Commission on
541
Radiation Unites and Measurements; 1999.
542
543
544
545
546
547
72. Yan D, Vicini F, Wong J, Martinez A. Adaptive radiation therapy. Phys Med Biol.
1997;42(1):123-32.
73. Li T, Zhu X, Thongphiew D, Lee WR, Vujaskovic Z, Wu Q, et al. On-line adaptive radiation
therapy: feasibility and clinical study. J Oncol. 2010; doi:10.1155/2010/407236.
74. Birkner M, Yan D, Alber M, Liang J, Nusslin F. Adapting inverse planning to patient and
organ geometrical variation: algorithm and implementation. 2003;30(10):2822-31.
28
548
75. Wu Q, Ivaldi G, Liang J, Lockman D, Yan D, Martinez A. Geometric and dosimetric
549
evaluations of an online image-guidance strategy for 3D-CRT of prostate cancer. Int J
550
Radiat Oncol Biol Phys. 2006;64(5):1596-609.
551
552
76. Jans HS, Syme AM, Rathee S, Fallone BG. 3D interfractional patient position verification
using 2D-3D registration of orthogonal images. Med Phys. 2006;33(5):1420-39.
553
77. Court LE, Tishler RB, Petit J, Cormack R, Chin L. Automatic online adaptive radiation
554
therapy techniques for targets with significant shape change: a feasibility study. Phys
555
Med Biol. 2006;51(10):2493-501.
556
78. Thongphiew D, Wu QJ, Lee WR, Chankong V, Yoo S, McMahon R, et al. Comparison of
557
online IGRT techniques for prostate IMRT treatment: adaptive vs repositioning
558
correction. Med Phys. 2009;36(5):1651-62.
559
79. Peroni M, Ciardo D, Spadea MF, Riboldi M, Comi S, Alterio D, et al. Automatic
560
segmentation and online virtualCT in head-and-neck adaptive radiation therapy. Int J
561
Radiat Oncol Biol Phys. 2012;84(3):e427-33. doi:10.1016/j.ijrobp.2012.04.003.
562
563
80. Sonke JJ, Belderbos J. Adaptive radiotherapy for lung cancer. Semin Radiat Oncol.
2010;20(2):94-106. doi:10.1016/j.semradonc.2009.11.003.
564
81. Jobse M, Davelaar J, Hendriks E, Kattevilder R, Reiber H, Stoel B. A new algorithm for the
565
registration of portal images to planning images in the verification of radiotherapy, as
566
validated in prostate treatments. Med Phys. 2003;30(9):2274-81.
29
567
30
568
569
Tables
Table 1: The Classification Of Registration Method By Maintz And Viergever In 1998.5
Criteria
Classification
I. Dimensionality
A. Spatial dimensions only: 2D-2D; 2D-3D; 3D-3D
B. Time series (more than two images), with spatial dimensions:
2D-2D; 2D-3D; 3D-3D
II. Nature of Registration Basis
A. Extrinsic: 1. Invasive (stereotactic frame, fiducials); 2. Noninvasive (mould-frame, dental adaptor)
B. Intrinsic: 1. Landmark based; 2. Segmentation-based; 3.
Voxel-property-based
C. Non-image based
III. Nature of Transformation
A. Rigid; B. Affine; C. Projective; D. Curved
IV. Domain of Transformation
A. Local; B. Global
V. Interaction
A. Interactive; B. Semi-automatic; C. Automatic
VI. Optimization Procedure
A. Parameters computed; B. Parameters searched for
VII. Modalities Involved
A. Mono-modal; B. Multimodal; C. Modality to model; D.
Patient to modality
VIII. Subject
A. Intrasubject; B. Intersubject; C. Atlas
IX. Object
A. Head; B. Thorax; C. Abdomen; D. Pelvis and peritoneum; E.
Limbs; F. Spine and vertebrae
570
31
571
Figure Captions
572
573
Fig. 1. A typical flow chart of image-guided radiation therapy (IGRT), highlighting the need for
574
registration, using a canine nasal tumor as an example. Diagnostic images are obtained and
575
registered to the reference image for treatment-planning. Each fraction of radiation is
576
administered with the aid of an electronic portal imaging device (EPID) or cone-beam computed
577
tomography (CBCT) images that are registered to the treatment-planning reference image.
578
579
Fig. 2. A typical flow chart of image registration demonstrating the steps – transformation
580
function, similarity function, and optimization – that are required to obtain a registration result.
581
582
Fig. 3. An example of a B-spline control grid on the reference and target image of a dog with a
583
nasal carcinoma. This demonstrates a 20x20 control grid on pre-treatment and post-treatment
584
computed tomography (CT) (512x512) images (manually warped).
585
586
Fig. 4. The variability in the similarity function (SSD) is shown and demonstrated from canine
587
nasal tumor images when the transformation function (rotation) is applied on the target image
588
and makes it rotate around the reference image from -90° to +90°. The rotation degree was
589
defined as the angle between the directions of the reference (black) and the target (white) images.
32
590
As a result, the image in A illustrates an approximately -70° rotation while B illustrates an
591
approximately 30° rotation.
592
593
Fig.5. Illustration of a multi-step registration framework in a dog with a nasal tumor. Images
594
show registration of a pre-treatment computed tomography (CT) study with the post-treatment
595
CT 6 weeks after definitive intensity-modulated radiation therapy. The series demonstrates the
596
target image with the pre-treatment gross tumor volume (GTV) contour (blue), target GTV
597
contour (yellow) and an automatically generated contour using a multistep process (red). A
598
represents D1, registration between the original reference and the target contour; B represents
599
D2, registration between the rigid registered reference contour and target contour; C represents
600
D3, registration between non-rigid registered reference contour and target contour; D represents
601
D4, the final segmented result and target contour. The Dice coefficients increase following each
602
step in the registration framework as shown in Fig. 6: D1 = 42.65 %, D2 = 85.57%, D3 =
603
90.81%, D4 = 96.73%.
604
605
Fig. 6. Portal images obtained from a dog with a sinonasal tumor undergoing definitive intent
606
radiation therapy prior to 6 MV photon therapy with a linear accelerator (Varian 2100 CD;
607
Varian Medical Systems, Palo Alto, USA); orthogonal images were obtained and analyzed
608
immediately prior to treatment delivery. An on-board flat panel imager permits acquisition of the
609
dog’s position prior to treatment (image in the lower right), which can be compared with the
33
610
digitally reconstructed radiograph (DRR) obtained from the planning computed tomography (CT)
611
(image in the upper right). Images were evaluated using commercially available treatment
612
planning software (Eclipse v.11.0, Varian Medical Systems). The larger image on the left
613
demonstrates registration of the electronic portal image and the DRR. (A) Images were obtained
614
with the linear accelerator gantry at 0 degrees; the larger image on the left demonstrates
615
evaluation of the registered images using a “moving window” technique that allows the window
616
to move the DRR over the dog’s daily portal image. The orange graticule grid represents the
617
dog’s position at acquisition of the portal image while the green graticule grid represents shifts
618
needed in the dog’s position to match the ideal planning position. (B) Images were obtained
619
with the linear accelerator gantry at 90 degrees; the larger image on the left demonstrates
620
evaluation of the images using a “split window” technique, in which the central axis can be
621
moved to reveal more of the DRR image or the electronic portal imaging device (EPID) image as
622
needed. Manual alignment may be performed as was done here in which the daily image is
623
moved to match the DRR; alternatively semi-automated or point matching registration can be
624
performed.
625
626
Fig 7. A series of rigidly registered images acquired on a human patient treated for prostate
627
cancer highlighting the difficulty in interpreting critical structures in different imaging
628
modalities. T1- (A) and T2- (B) weighted MR images acquired on a Siemens Symphony 1.5T
629
scanner (Siemens, Munich, Germany) were registered to a radiotherapy planning CT (C)
34
630
acquired on a Philips Brilliance Big Bore CT scanner fitted with a flat couch. The hyper-intense
631
gold markers implanted within the prostate gland and used for positional verification are clearly
632
visible. The corresponding pre-treatment CBCT image (D) acquired on a Varian Clinac and used
633
for positional verification. Alignment is complete and treatment begins when the hyper-intense
634
markers are aligned on CT and CBCT. Soft tissue matching is difficult because of the atrophy
635
induced in the prostate by the neo-adjuvant hormones administered between MR and CT
636
acquisition.
637
638
639
640
35