Abstract
Purpose:
Validation of MRI-guided tumor boundary delineation for targeted prostate cancer therapy is achieved via correlation with gold-standard histopathology of radical prostatectomy specimens. Challenges to accurate correlation include matching the pathology sectioning plane with the in vivo imaging slice plane and correction for the deformation that occurs between in vivo imaging and histology. A methodology is presented for matching of the histological sectioning angle and position to the in vivo imaging slices.
Methods:
Patients (n = 4) with biochemical failure following external beam radiotherapy underwent diagnostic MRI to confirm localized recurrence of prostate cancer, followed by salvage radical prostatectomy. High-resolution 3-D MRI of the ex vivo specimens was acquired to determine the pathology sectioning angle that best matched the in vivo imaging slice plane, using matching anatomical features and implanted fiducials. A novel sectioning device was developed to guide sectioning at the correct angle, and to assist the insertion of reference dye marks to aid in histopathology reconstruction.
Results:
The percentage difference in the positioning of the urethra in the ex vivo pathology sections compared to the positioning in in vivo images was reduced from 34% to 7% through slicing at the best match angle. Reference dye marks were generated, which were visible in ex vivo imaging, in the tissue sections before and after processing, and in histology sections.
Conclusions:
The method achieved an almost fivefold reduction in the slice-matching error and is readily implementable in combination with standard MRI technology. The technique will be employed to generate datasets for correlation of whole-specimen prostate histopathology with in vivo diagnostic MRI using 3-D deformable registration, allowing assessment of the sensitivity and specificity of MRI parameters for prostate cancer. Although developed specifically for prostate, the method is readily adaptable to other types of whole tissue specimen, such as mastectomy or liver resection.
Keywords: pathology correlation, histopathology, prostate cancer, magnetic resonance imaging, registration, MRI-guided targeted therapy
1. INTRODUCTION
MRI has been demonstrated as a powerful tool for mapping the spatial extent of prostate cancer (PCa) for targeted therapy such as high dose rate brachytherapy1 and MRI-guided biopsy,2 e.g., T2-weighted (T2w) signal,3 and parameters from diffusion-weighted imaging (DWI)3 and dynamic contrast enhanced MRI (DCE-MRI).4 However, the sensitivity and specificity of imaging parameters is determined by comparison with the gold-standard of histology. Retrospective evaluation can be made after radical prostatectomy, by correlation with a digital reconstruction of whole-mount histopathology.5 This task is highly challenging, as deformation occurs from multiple influences between the in vivo state and histology,6 and it can be difficult to relate the anatomy of an ex vivo specimen to its in vivo orientation. Precise slice matching to in vivo imaging is critical, as the foci of early stage PCa are likely to be smaller than the pathology section width (typically 3 mm).
This work presents a robust and readily implementable methodology for correlation of whole-mount prostate histopathology with in vivo imaging, using intermediate ex vivo MRI to determine the pathology sectioning plane, while employing a novel in-house developed sectioning device to guide slicing in this plane. The sectioning device also facilitates the insertion of dye tracks to guide reconstruction of the 3-D histopathology volume, and reference dye marks to verify the accuracy of the pathology slice planes. An initial clinical evaluation of the methodology is presented.
2. METHODS
The workflow for the methodology involved a sequence of in vivo and ex vivo imaging and correlation techniques (Fig. 1).
FIG. 1.
Pathology correlation workflow. Ex vivo MRI of the specimen determines the pathology sectioning angle for best match with in vivo imaging. A sectioning device is employed in four steps, to guide pathology sectioning and create reference marks for histopathology reconstruction.
2.A. Patient recruitment
Prostate cancer patients (n = 4) with biochemical failure after radiotherapy were enrolled in an institutional Research Ethics Board approved prospective clinical trial to map sites of local recurrence. Subsequent to a confirmed diagnosis from MRI and MR-guided biopsy, patients proceeded to radical prostatectomy.
2.B. In vivo imaging
Patients were imaged on a 1.5-tesla (T) MRI scanner (GE Signa Infinity) with a pelvic coil and an endorectal coil (MEDRAD MRInnervu). To minimize prostate deformation, the balloon of the endorectal coil was not inflated. The imaging examination included DWI, DCE-MRI, and T2w imaging. The T2w imaging was a conventional diagnostic axial fast spin-echo (FSE) acquisition: echo time (TE) = 102 ms; repetition time (TR) = 4650 ms; echo train length (ETL) = 16; field of view (FOV ) = 140 × 140 mm2; imaging matrix = 512 × 512; pixel in-plane dimension = 0.273 × 0.273 mm2; and slice thickness = 3 mm. Imaging was acquired axial to the endorectal coil. The areas suspicious for cancer were selected on the images by the oncologist (CM) using MIPAV (NIH, Bethesda, MD), on the basis of reduced T2w signal and evidence of extra-capsular extension.
2.C. Postsurgery anatomical marking with tissue dye
The surfaces of the prostate specimens were marked with tissue dye as a reference of anatomical orientation: apex = yellow; right side = red; left side = green; and black dye was inserted through the urethra.
2.D. High-resolution MRI of ex vivo prostate at 7 T
To estimate the pathology sectioning angle and to determine the deformation during fixation, the ex vivo prostate was MRI scanned pre- and postfixation. To prevent MRI susceptibility artifacts at air–tissue boundaries, the specimen was embedded in a block (7 × 7 × 7 cm3) of gel (4% gelatin, 1.5% agarose), resting flat on the posterior surface above 1 cm of gel, and such that the axes of the specimen and block were aligned. The block was imaged in a 7-T preclinical MRI scanner (70/30 BioSpec, Bruker, Ettlingen, Germany), using a quadrature volume resonator (15.5 cm inner diameter) for transmission and reception. High magnetic field strength facilitated access to higher signal to noise ratios at high resolution, for enhanced anatomical mapping. A spin-echo sequence, rapid acquisition with relaxation enhancement (RARE) was used to obtain a T2w 3-D isotropic high-resolution acquisition of the block. The sequence parameters were TE = 68 ms, TR = 2400 ms, ETL = 16, FOV = 70 × 70 × 70 mm3, imaging matrix = 233 × 233 × 233, voxel dimension = 0.3 × 0.3 × 0.3 mm3 (approximately matching the in vivo in-plane voxel dimension). The scan acquisition time was about 2 h. Postfixation, TE was reduced to 40 ms, to compensate for fixation-mediated reduction in the transverse relaxation time, T2.9
2.E. Pathology sectioning angle estimation from ex vivo MRI
Prostate specimens are typically positioned on the posterior surface (closest to the rectum) prior to sectioning. Therefore, to match sectioning to in vivo imaging, estimation must be made of the angle in the anterior–posterior direction between the posterior surface and the coil axis (axis of imaging), denoted α1, and also the left–right angle between the main axes of the prostate and coil, denoted α2. Angles α1 and α2 were initially estimated using the prefixation ex vivo MRI data, and updated postfixation. The oncologist (CM) contoured the prostate boundaries and identified salient anatomical features on all images using matlab (The MathWorks, Natick, MA). Three implanted gold seed fiducials (length 3–5 mm, diameter 1 mm) were visible as focal hypointensities for three patients (#2–#4), both in vivo and ex vivo. Contours were drawn around each marker artifact, allowing centroid calculation. A 3-D rotational transform between the in vivo and ex vivo states was calculated from the marker centroids, which provided estimates of α1 and α2.
Implanted fiducial markers were not visible for patient #1, necessitating the use of distance metrics for angle estimation (Fig. 2). Four distances were measured to the center of the urethra in corresponding in vivo and ex vivo slices: d1: the distance from the anterior surface; d2: from the posterior surface; d3: from the left surface; d4: from the right surface. The ratios of these distances were calculated as
| (1) |
The axial (to the posterior surface) ex vivo data were digitally resampled at varying axial oblique angles (α1) from −30° to +30° in steps of 5°, and down-sampled to create 3-mm thick slices, to match the in vivo slices. The α1 was identified that minimized the mean square difference between in vivo and ex vivo ratios. For final refinement, α1 was varied through increments of 1° in the vicinity of the minimum. Angle α2 was determined by rotation of the α1-rotated data through increments of 1° from −5° to +5°.
FIG. 2.
Estimating α1 and α2 by comparison of distance metrics between in vivo and (prefixation) ex vivo MRI data sets. (a) On the in vivo image distances are measured between the center of the urethra and the anterior (d1), posterior (d2), left (d3), and right (d4) edges, from which the ratios r1,2 and r3,4 are calculated. (b) The metrics are measured on the ex vivo MRI slice and the ratios compared with in vivo data. (An in-plane rotation was also required to compensate for rotation of the specimen caused by extraprostatic tissue on the posterior surface.) (c) The ex vivo data rotated by α1 = 20° give the best match to in vivo ratios.
2.F. Sectioning device
A sectioning device (Fig. 3) was designed for MRI-guided matching of the sectioning planes to in vivo imaging. It consisted of an acrylic mold with detachable sides, and a series of attachments and guidance plates made from polycarbonate (ABS blend), which were generated by fast prototyping using a Stratasys Vantage 3-D printer (Stratasys, MN). The hypodermic needles employed were 21G and 12 cm long (B. Braun Melsungen AG, Germany), and the tissue dye (Bradley Products, Inc., MN) was diluted in a 1:5 ratio with ultrasound (U/S) gel, to limit diffusion and leakage.
FIG. 3.
Sectioning device. (a) Plate type A (width = 9 cm, height = 5.5 cm, thickness = 3 mm) with 1-mm deep grooves for needle-guidance, and to create MR-visible reference tracks when filled with ultrasound gel. (b) Plate type B (same dimensions as type A) with slots for guiding needles through plate. [(c), (f)] Device includes a rectangular mold (length = 25 cm, width = 10 cm), and guide rails for holding plates, with a pivot to set the angle. (c) Use #1: guiding needles through type B plate and perpendicular to the slicing plane. (d) Use #2: setting the angle of the type A plate to the slicing angle. (Side view depicts the urethra trajectory.) (e) Use #3: All type A plates are fitted at the slicing angle, and needles are guided into tissue surface to create reference dye marks for sectioning. (f) Use #4: Type B plate is fitted to the last slot and at the same angle, and needles are positioned through the plate to create tracks through the gel that will embed the tissue.
2.G. Sectioning device use #1: Insertion of needle tracks through the tissue
On completion of the fresh state ex vivo MRI, α1 and α2 were estimated. The specimen was removed from gel, and two needles were inserted in the tissue perpendicular to the determined sectioning plane [Fig. 3(c)], which were left in place during fixation to create holes that would be visible in histology. For fixation, the specimen was submersed for 60 h in 10% neutral buffered formaldehyde solution (formalin). Once removed from formalin, dye was inserted during needle removal, to help distinguish the needle marks from other features.
2.H. Sectioning device use #2: Setting angle and position of the sectioning guidance plate
On completion of the fixed state ex vivo MRI, α1 and α2 were updated. The mold was preset with a 1-cm gel layer, upon which the specimen was placed on its posterior surface. (α2 was not found to differ from zero for any specimen, however in the instance where α2 is nonzero, a protractor may be employed to position the specimen accordingly.) For immobilization, another 1-cm layer of gel was poured around the specimen and allowed to set.
The pathology section thickness matched the in vivo MRI slice thickness (3 mm). For optimum sampling of disease margins at the apex, the width of the first section at the apex should be close to 3 mm. Hence, the objective was to position the first plate in the 3-mm slice immediately before the first slice containing tissue [Fig. 3(d)]. A set of 1-mm thick T2w MR imaging slices was acquired at α1. It was deemed that the plate position and angle was correct when the U/S-gel filled grooves were visible in the last 1-mm slice before the apex.
2.I. Sectioning device use #3: Insertion of dye marks at the sectioning positions
All the type-A plates were fitted and set to α1 [Fig. 3(e)]. A needle was guided along the grooves in each plate to insert dye markers of about 1 cm in depth into the tissue, by first depositing a small quantity on the surface, followed by slowly inserting the needle while not applying pressure to the syringe. Marks were made at 5-mm intervals across the slicing planes, which were separated by 3 mm. The dye color was varied between plates in a repeating sequence of three colors, to distinguish neighboring planes. On completion, all plates, with the exception of the first, were removed, and the last plate was replaced by a type-B plate. These two plates would remain in place to guide pathology sectioning, and furthermore the type-B plate would be used for the work described in Sec. 2.J.
2.J. Sectioning device use #4: Insertion of needle tracks through the gel
Four needles were pushed through the type-B plate and positioned alongside the tissue [Fig. 3(f)]. These would create tracks in the embedding gel, which would guide stacking of the digital images of the tissue sections during histopathology reconstruction. The mold with the specimen, plates, and needles in place was filled with gel and refrigerated overnight. To increase track visibility, dye was deposited during needle removal.
2.K. Pathology sectioning
To prepare the specimen for sectioning, first the sides of the mold were detached from the gel block in which the specimen was embedded, and the guide rails were also removed. The specimen would remain embedded in gel during sectioning. Furthermore, the two guidance plates (see Sec. 2.I) would remain embedded in the gel with the specimen, to help guide sectioning. A rotary blade meat slicer was used for sectioning and the sectioning angle on the slicer was set by use of a rectangular wedge of gel, which was prepared at the best match angle α1. The gel block containing the specimen was placed on top of the wedge block, and both were trimmed so that the combined cutting edge was parallel to the embedded plates. The block and wedge were placed on the holding tray of the slicer so that the apex would be sliced first. The pathologist (TvdK) performed all histological sectioning. Once slicing had reached the first (type A) plate, the block and wedge were removed from the slicer and the plate was detached. Some gel protruded from where the rectangular opening in the plate had been, and this was carefully trimmed by hand to create a tight fit with the back plate of the slicer. The block and wedge were repositioned on the slicer, verifying that the slice plane was parallel to the remaining plate. The next slice consisted of the first slice of the apex, and slicing continued for the whole specimen. The sections were ordered in sequence during slicing, with the surrounding gel kept in place, and were digitally captured at high resolution using a camera mounted at a fixed height, while resting on a 1-mm spaced grid to indicate the dimensions.
2.L. Histopathology processing
The tissue sections were processed in a histoprocessor, embedded in paraffin and sliced on a sledge microtome to produce one 5-μm whole-mount histology slide section from the top of each block, which was stained with Hematoxylin and Eosin (H&E). The slides were digitized at 5 μm resolution on a TISSUEscope™ 4000 (Biomedical Photometrics, Inc., ON, Canada) and digitally annotated by the pathologist using aperio imagescope software (Aperio Technologies, Inc., CA).
3. RESULTS
3.A. Pathology slicing angle estimation from ex vivo MRI
For specimen #1, estimates of α1 and α2 were 20° and 0°, respectively, before fixation, and were unchanged after fixation. The estimated angles for specimens #2–#4 (Table I) (i.e., with fiducial markers) were similar to those of specimen #1, and approximately consistent before and after fixation, with an average α1 of 17° both pre- and postfixation, while α2 was 0° for all specimens. Angle variation after fixation is likely due to tissue deformation. A possible added influence is pressure from the plate guiding the needles during use #1, and care was taken to avoid this (subsequent to the differing angle result of specimen #2), by ensuring that the plate was supported underneath.
TABLE I.
Results of angle estimation from ex vivo MRI data for specimens #2–#4.
| Prefixation | Postfixation | |||
|---|---|---|---|---|
| Specimen No. | α1 (deg) | α2 (deg) | α1 (deg) | α2 (deg) |
| 2 | 10 | 0 | 5 | 0 |
| 3 | 20 | 0 | 20 | 0 |
| 4 | 20 | 0 | 25 | 0 |
| Mean(±std) | 17(±6) | 0 | 17(±10) | 0 |
The in vivo MRI r1,2 values for an example slice near the apex were compared with the ratios in postfixation ex vivo MRI axial to the posterior surface, postfixation ex vivo MRI resampled at α1, and the tissue slice photograph. The r1,2 errors were calculated from estimated uncertainties of 1 mm for the tissue edge and the urethra center. For specimen #1, the r1,2 ratio measured on in vivo MRI was 1.9 ± 0.4, and on the ex vivo postfixation slice without digital resampling r1,2 was 2.6 ± 0.6 (37% difference). After digital resampling of the ex vivo data, the r1,2 ratio was 1.9 ± 0.4, which matched the in vivo ratio. After sectioning, r1,2 on the tissue slice photograph was 1.8 ± 0.4; a difference of −5% with the in vivo r1,2. The ratios and percentages were similar for specimens #2–#4 (Table II). A substantial improvement in spatial matching was achieved with resampling at α1, i.e., the mean absolute percentage difference of r1,2 was reduced from 34(±18)% to 12(±4)%.
TABLE II.
For specimens #2–#4, r1,2 values for example matching slices near the apex for in vivo MRI, ex vivo MRI, ex vivo MRI resampled at α1, and in the photographed tissue slices. For the ex vivo ratios, the percentage differences from the in vivo ratios are also listed.
| Specimen No. | r1,2 in vivo MRI | r1,2 ex vivo MRI | % Δ (r1,2 ex vivo MRI −r1,2 in vivo MRI) (%) | r1,2 ex vivo MRI resampled at α1 | % Δ (r1,2 ex vivo MRI resampled −r1,2 in vivo MRI) (%) | r1,2 tissue slice | % Δ (r1,2 tissue slice −r1,2 in vivo MRI) (%) |
|---|---|---|---|---|---|---|---|
| 2 | 1.5 ± 0.2 | 1.7 ± 0.3 | 13 | 1.6 ± 0.3 | 7 | 1.5 ± 0.3 | 0 |
| 3 | 1.6 ± 0.3 | 2.3 ± 0.5 | 44 | 1.4 ± 0.3 | −13 | 1.6 ± 0.3 | 0 |
| 4 | 2.0 ± 0.4 | 2.9 ± 0.7 | 45 | 2.3 ± 0.6 | 15 | 1.6 ± 0.3 | −20 |
| Mean absolute % Δ | 34(±18) | 12(±4) | 7(±12) |
3.B. The sectioning device
3.B.1. Use #1
The dye tracks were visible in the postfixation ex vivo MRI [Fig. 4(a)], and in the tissue sections [Fig. 4(b)]. While the dye was found to mostly wash away during the histology processing before paraffin embedding [Fig. 4(c)], the holes created by the needles were still clearly visible with some indication of the dye around the rim. Furthermore, these holes were visible on the histology sections [Fig. 4(d)].
FIG. 4.
Matching images for MRI and histology. [(a)–(d)] Slice position toward prostate base. [(e)–(h)] Central slice position. (a) Ex vivo postfixation MRI resampled at α1. Arrows indicate dye tracks from use #1. (b) Photograph of tissue section with gel. Arrows inside tissue indicate dye tracks from use #1, while arrows outside tissue indicate dye tracks from use #4. (c) Paraffin block after removal of histology slide section. Arrows indicate holes corresponding to tissue dye tracks from use #1. (d) H&E stained histology slide with tumor areas marked by pathologist. Arrows indicate holes corresponding to dye tracks from use #1. Some residual gel is present around the tissue. (e) In vivo T2w MRI showing contour of prostate and tumor areas as determined by the oncologist. (f) Ex vivo prefixation MRI of matching slice resampled at α1. (g) Paraffin block after removal of histology section. Arrows indicate the dye marks from use #3. (h) H&E stained histology slide with tumor areas marked by the pathologist. Arrows indicate tracks of marks from use #3.
3.B.2. Use #2
The U/S gel in the grooves of the type-A plate was predominantly visible in the last 1-mm slice before the apex. The uncertainty on the positioning of the 1-mm wide and 15-mm long grooves in a 1-mm wide imaging slice was estimated as half the depth of the groove, at 0.5 mm (i.e., partial volume averaging with neighboring slices). Therefore, the uncertainty on the setting of α1 was arctan(0.5/15) = 1.9°. The average error on r1,2 for specimens #2–#4 for the tissue slice photographs was 7(±12)%, which is lower than the average error for the resampled ex vivo MRI [12(±4)%] (Table II), demonstrating that the step of setting the sectioning angle by use of the sectioning device did not inflate the errors.
3.B.3. Use #3
The dye marks inserted at the planned sectioning positions were visible in some sections after slicing. Owing to variable pressure applied during sectioning on the slicer, the section thickness was found to vary slightly (typically between 2.9 and 3.1 mm) and these errors could accumulate through slicing, and hence sectioning did not always occur at the predefined positions. However, when visible it was possible to verify accuracy of the slice plane angle in the left–right direction, by comparing marker visibility on the left and right. The dye marks were found to be robust to all processing steps (cycles of formalin, ethanol, xylene, and paraffin in the histoprocessor) to create the paraffin blocks [Fig. 4(g)]. This added a further reference for the left–right angle during histology sectioning on the microtome. Needle tracks corresponding to these dye marks were visible on the histology slides [Fig. 4(h)].
3.B.4. Use #4
The needle tracks made through the gel were visible on photography of the sections, with some of the dye remaining in the tracks [Fig. 4(b)].
3.C. Comparison of slices from in vivo to histology
Correlation of anatomical features was observed between in vivo MRI, ex vivo MRI, section photography, paraffin blocks, and histology (Fig. 4). Furthermore, there was initial indication of coincidence of disease margins between diagnostic MRI [Fig. 4(e)] and histology [Fig. 4(h)]. Extraprostatic tissue removed with the prostate provided extra cross-referencing features for slice matching between ex vivo data. However, in histology, this extra tissue (sometimes combined with residual gel from the block) tended to fragment [Figs. 4(d) and 4(h)].
4. DISCUSSION
4.A. Comparison with other methods
In the prostate, the urethra follows an arc-shaped path [Fig. 3(d)], and therefore slicing perpendicular to an arbitrary axis of the urethra7,8 can give rise to an angle mismatch of as much as 15°.7 Langer et al.10 imaged ex vivo specimens at multiple axial oblique angles in increments of 5° to match in vivo imaging. In another study, an organ mold was designed from in vivo images, in which the specimen was fixed, imaged, and sliced.11 However, tissue can deform in the mold during fixation. Image-guided magnetically tracked probes have also been used to indicate the sectioning plane on the ex vivo specimen.6 In contrast, the method presented here is readily implementable with standard MRI technology.
4.B. Pathology slice angle estimation
Angle estimation from in vivo imaging was inaccurate, as the 3-mm thick imaging slices provided very approximate information on the prostate shape, and it was not known how the ex vivo specimen would align on the posterior surface. The angle estimates from ex vivo imaging were generally similar between specimens, suggesting a common anatomical orientation and a predictable compression by the coil. For the distance metric approach, errors were estimated based on a 1 mm uncertainty in contour and feature selection. For the gold seeds method, the uncertainty was dependent on the artifact boundary selection. However, as the artifact dimensions vary between frequency-encode and phase-encode directions, and with magnetic field strength and readout bandwidth, the overall uncertainty was difficult to assess, and will be the subject of future work. However, further studies are required with larger numbers of specimens to fully explore the variability of prostate anatomy between individuals and the accuracy of pathology slice matching using this methodology.
4.C. Benefits of the sectioning device
The device provided a means of guiding pathology sectioning at the angles and positions which were estimated to best match the in vivo imaging planes. Even though prostate tissue is likely to undergo deformation from multiple influences between in vivo imaging and histology [i.e., deformable registration is included in the proposed pipeline to reduce errors from these effects (Fig. 1)], matching of the pathology slice angles to in vivo imaging is still a critically important step in the process. Given that the dimension of the foci of early stage PCa could be less than the pathology tissue slice thickness (3 mm), and that typically only a single histology slide is created from the top of each pathology section, it is vital to match the histology slice angles with the best estimates of the in vivo imaging angles. The device also ensured an adequate width of the first tissue section, which permits more precise sampling at the margins (e.g., with multiple-step histology).
The dye tracks through the tissue and gel provided a reference for histopathology reconstruction, and the tissue tracks could be used to assess deformation during processing. Although no needle deformation was discernable during needle insertion in the tissue or gel, slight deviations in the needle trajectory may occur. However, these deviations might be detectable in the postfixation MRI data, and examination of the data for these potential effects should be included in future implementations of this process. The dye marks at the planned sectioning positions indicated the slicing plane accuracy, and assisted histology slicing on the microtome.
The data obtained will be used in 3-D finite element modeling (FEM) biomechanical-based deformable registration12–14 of the histopathology volume to the in vivo volume in a stepwise process through the tissue states: (1) histopathology to ex vivo fixed MRI; (2) the result of 1 to ex vivo fresh MRI; (3) the result of step 2 to in vivo MRI.
5. CONCLUSIONS
A methodology has been presented for correlation of whole-specimen prostate histopathology with in vivo diagnostic MRI. Ex vivo MRI was employed to determine the pathology sectioning angle, and a novel sectioning device was demonstrated as an important tool to guide sectioning, and to allow creation of reference dye marks to aid in histopathology reconstruction. The process is easily implemented with standard MRI technology and is also readily adaptable to other tissue types, such as mastectomy or liver resections. The data obtained will be used to register histopathology with in vivo imaging to determine the sensitivity and specificity of diagnostic MRI parameters to prostate cancer.
ACKNOWLEDGMENTS
This research was funded in part by a grant from the Ontario Institute for Cancer Research and in part by a NIH Grant No. 5R21CA121586-2. K.K.B. was supported as a Cancer Care Ontario Research Chair. The authors would like to thank Matthew Filleti for technical assistance in development of the sectioning device and Dr. Andrew Evans for pathology support.
REFERENCES
- 1.Menard C., Susil R. C., Choyke P., Gustafson G. S., Kammerer W., Ning H., Millner R. W., Ullman K. L., Crouse N. S., Smith S., Lessard E., Pouliot J., Wright V., McVeigh E., Coleman C. N., and Camphausen K. C., “MRI-guided HDR prostate brachytherapy in standard 1.5T scanner,” Int. J. Radiat. Oncol., Biol., Phys. 59(5), 1414–1423 (2004). 10.1016/j.ijrobp.2004.01.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Menard C., Iupati D., Lee J., Simeonov A., Abed J., Publicover J., Chung P., Bayley A., Catton C., Milosevic M., Bristow R., Morton G., Warde P., Brock K., and Haider M. A., “MRI and biopsy performance in delineating recurrent tumor boundaries after radiotherapy for prostate cancer,” Proc. Int. Soc. Magn. Reson. Med. 19, 3072 (2011). [Google Scholar]
- 3.Haider M. A., van der Kwast T. H., Tanguay J., Evans A. J., Hashmi A., Lockwood G., and Trachtenberg J., “Combined T2-weighted and diffusion-weighted MRI for localization of prostate cancer,” Am. J. Roentgenol. 189, 323–328 (2007). 10.2214/ajr.07.2211 [DOI] [PubMed] [Google Scholar]
- 4.Haider M. A., Chung P., Sweet J., Toi A., Jhaveri K., Menard C., Warde P., Trachtenberg J., Lockwood G., and Milosevic M., “Dynamic contrast-enhanced magnetic resonance imaging for localization of recurrent prostate cancer after external beam radiotherapy,” Int. J. Radiat. Oncol., Biol., Phys. 70(2), 425–430 (2008). 10.1016/j.ijrobp.2007.06.029 [DOI] [PubMed] [Google Scholar]
- 5.Langer D. L., van der Kwast T. H., Evans A. J., Trachtenberg J., Wilson B. C., and Haider M. A., “Prostate cancer detection with multi-parametric MRI: Logistic regression analysis of quantitative T2, diffusion-weighted imaging, and dynamic contrast-enhanced MRI,” J. Magn. Reson. Imaging 30, 327–334 (2009). 10.1002/jmri.21824 [DOI] [PubMed] [Google Scholar]
- 6.Ward A. D., Crukley C., McKenzie C., Montreuil J., Gibson E., Gomez J. A., Moussa M., Bauman G., and Fenster A., “Registration of in vivo prostate magnetic resonance images to digital histopathology images,” in Proceedings of MICCAI: International Conference on Medical Image Computing and Computer Assisted Intervenion, LNCS (Springer Berlin Heidelberg, 2010), Vol. 6367, pp. 66–76. [Google Scholar]
- 7.Jager G. J., Ruijter E. T. G., van de Kaa C. A., de la Rosette J. J. M. C. H., Oosterhof G. O. N., Thornbury J. R., and Barentsz J. O., “Local stagin of prostate cancer with endorectal MR imaging: Correlation with histopathology,” Am. J. Roentgenol. 166, 845–852 (1996). 10.2214/ajr.166.4.8610561 [DOI] [PubMed] [Google Scholar]
- 8.Graser A., Heuck A., Sommer B., Massmann J., Scheidler J., Reiser M., and Mueller-Lisse U., “Per-sextant localization and staging or prostate cancer: Correlation of imaging findings with whole-mount step section histopathology,” Am. J. Roentgenol. 188, 84–90 (2007). 10.2214/AJR.06.0401 [DOI] [PubMed] [Google Scholar]
- 9.Yong-Hing C., Obenaus A., Stryker R., Tong K., and Sarty G., “Magnetic resonance imaging and mathematical modeling of progressive formalin fixation of the human brain,” Magn. Reson. Med. 54, 324–332 (2005). 10.1002/mrm.20578 [DOI] [PubMed] [Google Scholar]
- 10.Langer D. L., van der Kwast T. H., Evans A. J., Plotkin A., Trachtenberg J., Wilson B. C., and Haider M. A., “Prostate tissue composition and MR measurements: Investigating the relationships between ADC, T2, Ktrans, Ve and corresponding features,” Radiology 255(2), 485–494 (2010). 10.1148/radiol.10091343 [DOI] [PubMed] [Google Scholar]
- 11.Shah V., Pohida T., Turkbey B., Mani H., Merino M., Pinto P. A., Choyke P., and Bernardo M., “A method for correlating in vivo prostate magnetic resonance imaging and histopathology using individualized magnetic resonance -based molds,” Rev. Sci. Instrum. 80(10), 104301 (2009). 10.1063/1.3242697 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.McNiven A., Moseley J., Langer D. L., Haider M. A., and Brock K. K., “Preliminary feasibility study: Modeling 3D deformations of the prostate from whole-mount histology to in vivo MRI,” Med. Phys. 36, 2712 (2009). 10.1118/1.3182294 [DOI] [Google Scholar]
- 13.Brock K., Ahmed S., Moseley J., Moulton C., Guindi M., Haider M., Gallinger S., and Dawson L., “Deformable registration for in vivo imaging and pathology correlation,” Med. Phys. 33, 1994 (2006). 10.1118/1.2240229 [DOI] [Google Scholar]
- 14.Brock K. K., Sharpe M. B., Dawson L. A., Kim S. M., and Jaffray D. A., “Accuracy of finite element model-based multi-organ deformable image registration,” Med. Phys. 32(6), 1647–1659 (2005). 10.1118/1.1915012 [DOI] [PubMed] [Google Scholar]