High-resolution MRI of excised human prostate specimens acquired with 9.4T in detection and identification of cancers: Validation of a technique

Authors


Abstract

Purpose:

To evaluate feasibility of high-resolution, high-field ex vivo prostate magnetic resonance imaging (MRI) as an aid to guide pathologists' examination and develop in vivo MRI methods.

Materials and Methods:

Unfixed excised prostatectomy specimens (n = 9) were obtained and imaged immediately after radical prostatectomy under an Institutional Review Board-approved protocol. High-resolution T2-weighted (T2W) MRI of specimens were acquired with a Bruker 9.4 T scanner to correlate with whole-mount histology. Additionally, T2 and apparent diffusion coefficient (ADC) maps were generated.

Results:

By visual inspection of the nine prostate specimens imaged, high-resolution T2W MRI showed improved anatomical detail compared to published low-resolution images acquired at 4 T as published by other investigators. Benign prostatic hyperplasia, adenocarcinomas, curvilinear duct architecture distortion due to adenocarcinomas, and normal radial duct distribution were readily identified. T2 was ≈10 msec longer (P < 0.03) and the ADC was ≈1.4 times larger (P < 0.002) in the normal peripheral zone compared to the peripheral zone with prostate cancer.

Conclusion:

Differences in T2 and ADC between benign and malignant tissue are consistent with in vivo data. High-resolution, high-field MRI has the potential to improve the detection and identification of prostate structures. The protocols and techniques developed in this study could augment routine pathological analysis of surgical specimens and guide treatment of prostate cancer patients. J. Magn. Reson. Imaging 2011. © 2011 Wiley-Liss, Inc.

RADICAL PROSTATECTOMY is the standard surgical treatment option for the patients with localized prostate cancer (1). Successful treatment requires complete resection of the tumor, which is currently assessed only by pathologists. The pathological assessment of the radical prostatectomy specimen is the gold standard used to decide the presence or absence of cancer at the surgical margin, aggressiveness of the tumor, and the tumor burden within the prostate (2). The pathologic information derived from radical prostatectomy specimens is used to predict patient outcome after surgery (3, 4), and to decide if a patient needs adjuvant treatment such as radiotherapy or hormone therapy (5, 6). Other promising prognostic information that can be obtained from prostatectomy specimen includes tumor volume (intraglandular extent), histologic subtype, vascular space, and perineural invasion. Therefore, it is important to accurately evaluate surgical specimens. A survey conducted by American Society of Clinical Pathologists indicated that only 12% of laboratories in the United States submitted the entire radical prostatectomy specimens for detailed histological analysis (7). A variety of partial submission strategies have been developed in the literature; however, they are all susceptible to errors caused by sampling bias, which can lead to some loss of prognostic information (8, 9).

Handling of the specimen samples and reporting the findings face significant challenges (10, 11). First, the submission of the entire specimen for histopathologic analysis is a time-consuming and expensive procedure for pathologists. Second, improper handling or processing of the specimen can cause artifactual positivity of the margins (12). Delicate strands of periprostatic connective tissue can sometimes be disrupted, resulting in ink being put on a portion of the gland that does not represent the margin (12). In addition, the histologic slides are relatively thick and this may lead to inaccuracies, especially in determination of extraprostatic extension and margin positivity (8, 9). Third, although tumor volume correlates with disease progression following radical prostatectomy, currently there is no standard method for assessing the tumor volume within the gland (13). Usually, the tumor volume is determined by a subjective visual inspection of the percentage of specimen involved by cancer. This difficulty is further compounded by partial sampling techniques that severely limit the accuracy with which intraglandular extent of the tumor is determined. Therefore, there is a critical need for an automated method to evaluate prostatectomy specimens so the distribution and extent of disease can be determined at histological resolution for comparison with in vivo imaging results.

Ex vivo imaging of prostate specimens has been proposed by a number of groups as a tool that can enhance the efficiency and accuracy of pathologic evaluation. Madabhushi et al (14) presented a fully automated computer-aided detection system for detecting prostatic adenocarcinoma from 4 T ex vivo magnetic resonance imaging (MRI) of the prostate with in-plane resolution of 160 μm. Zhan et al (15) developed a novel method for deformable registration of histologic and MR images of the same prostate acquired with low resolution at 4 T. Although these and others groups have reported correlation of registered MRI and histology, they often lack whole-mount histology and high-resolution MRI acquired at high field (14–16).

In this research, we performed high-resolution (≈137 μm in-plane resolution) MRI fresh prostatectomy specimens using a 9.4 T small-bore scanner prior to routine histopathological specimen. The feasibility of correlating MRI findings of with standard prostate histopathology was investigated. The T2 and apparent diffusion coefficient (ADC) values were also calculated and compared between normal peripheral zone (PZ) and PZ prostate cancer.

PATIENTS AND METHODS

Patients

An Institutional Review Board-approved protocol was followed to identify and recruit patients who were appropriate for this study. The protocol allowed research scans of radical prostatectomy specimens in cases where patients gave their informed consent. In vivo MR images of the prostate were acquired as a part of the clinical work-up of each patient using an endorectal coil and a 1.5 T MRI scanner. To date, we have recruited nine patients (average age = 64 ± 5 years) who had undergone radical prostatectomy and preoperative multiparametric MRI at our institution and agreed to participate in the study. These patients had prostate-specific antigen (PSA) values 4.9–23.7 ng/mL. Seven patients had a Gleason score (GS) of 7, one had GS of 9, and one had GS of 6. In addition, extracapsular extension was seen in three patients, and two patients had seminal vesicle invasion. On average, there were ≈43 days (range 10–97 days) between prostate biopsy and surgery. Any artifacts from the biopsy procedure would be minimal on ex vivo MRI.

MRI of Prostate Specimen Acquired at 9.4T

Immediately after surgery, prostate specimens were obtained and delivered by the radiologist to the MRI laboratory for ex vivo MRI. The specimen was sealed in a plastic bag with some crushed ice inside. The specimen size was typically 5–7 cm in the transverse dimension and 3–4 cm in the anteroposterior dimension (Fig. 1). For that reason, the Bruker (Billerica, MA) 72 mm birdcage volume coil was selected for all ex vivo imaging. Ex vivo MRI experiments were conducted with a 9.4 T Bruker BioSpec 33 cm horizontal bore scanner with 11.6 cm self-shielded gradient coil insert (maximum strength 660 mT/m). The orientation of the specimen in the scanner was similar to the orientation for in vivo MRI. The imaging protocol was tested on a phantom prior to the scans of first specimen and the final optimized protocol was determined after the first specimen. For each specimen, low-resolution coronal T2-weighted (T2W) fast spin echo images were acquired first using the rapid acquisition with refocused echoes (RARE) sequence (TR/TEequivalent = 2000/26.7 msec, in-plane resolution = ≈500 μm, array size = 128 × 128, slice thickness = 1 mm, RARE fact = 4, and number of excitations [NEX] = 1). These images were acquired to identify the prostate orientation so that subsequent high-resolution axial slices could be prescribed accurately. Two sets of axial high-resolution RARE T2-weighted images (TR/TEequivalent = 5000/40 msec, in-plane resolution = ≈137 μm, array size = 512 × 256, slice thickness = 1 mm, slice gap = 1 mm, RARE fact = 4, and NEX = 2) were acquired to facilitate registration of MR data with histology. The first and second sets were offset by 1 slice thickness and were collated offline to produce multislice images covering the whole specimen. This scan was also used to guide the subsequent T2 imaging and diffusion-weighted images (DWI). A total of 15 axial slices were imaged with multiple echo T2W imaging (TR = 5000 msec, smallest TE = 12.5 msec, number of echo = 16, array size = 256 × 128, in-plane resolution = ≈234 μm, slice thickness = 2 mm, slice gap = 1 mm, NEX = 1) and DWI (TR/TE = 4000/27.7 msec, b-value = 0, 1000, and 3000 s/mm2, in-plane resolution and slice thickness were same as for T2 imaging, NEX = 1). Data were analyzed to produce T2 and ADC maps. This protocol typically requires about 90 minutes including experiment setup time. Immediately after the ex vivo MRI scans the specimen was returned to the pathology laboratory for formalin fixation and processing for histology.

Figure 1.

Photograph of a typical prostate specimen prior to imaging. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Ex Vivo MRI Data Analysis

Ex vivo MRI data analysis was performed using software written in IDL (ITT Visual Information Solutions, Boulder, CO). The two sets of axial high-resolution T2W images were collated into a single set to facilitate correlation with histopathology. The T2 maps and ADC maps were generated by fitting the signal intensity curve as a function of echo time and b values, respectively, with a single exponential decay.

Histopathological Evaluation and Correlation With Ex Vivo MRI

After ex vivo imaging the specimen was fixed with formalin. The fixed specimen was sliced accurately into 4-mm thick slices and scanned with a desktop scanner (Epson Perfection V300). The 3D stack of the optical scans facilitated assembly of the 3D stack of histology scans and served as a “bridge” to the ex vivo MRI. Paraffin embedded blocks were sectioned ≈5-μm thick and stained with hematoxylin and eosin. The locations of tumors were recorded for the peripheral zone (PZ) or the transition zone (TZ) and the base, middle, or apex of the prostate. The pathologic stage and the presence of extraprostatic extension, seminal vesicle invasion, lymph node metastasis, and tumor-positive margins were also recorded. All imaging studies were evaluated jointly by a senior radiologist (12 years experience), a radiology resident, a physicist (10 years experience), and the experienced genitourinary pathologist (10 years experience).

Statistical Analysis

Normal PZ and PZ prostate cancers were outlined on whole-mount pathology slices and visually matched with high-resolution T2-weighted images, so that the physicist could calculate T2 and ADC values for these regions of interest. The Wilcoxon Mann–Whitney Test was performed to analyze whether T2 and ADC values were statistically different in normal PZ and PZ cancer. P < 0.05 was considered significantly different.

RESULTS

For all nine prostate specimens the ex vivo high-resolution T2W MR images had much more anatomical detailed features than in vivo images, which could not be seen in the in vivo images. Figure 2 shows an example of high-resolution T2W MR images (left panel) with roughly corresponding matched histology slices (right panel) for four different patients: (Fig. 2a) 65 years old, (2b) 67 years old, (2c) 71 years old, and (2d) 54 years old. By visual inspection, correlation between ex vivo prostate MR images and histology slices was good despite the fact that the histology slices were usually cut at a slightly different angle than MR images, and despite some distortion of anatomy in fixed tissues slices. On histology slides cancer was marked as dotted line by the pathologist.

Figure 2.

The ex vivo T2W fast spin echo prostate images (left panel) and approximately corresponding histology slices (right panel) from four different prostate cancer patients with Gleason scores of 7: (a) 65 years old, (b) 67 years old, (c) 71 years old, and (d) 54 years old. On histology slides everything marked with a dotted line is cancer. Identified features drawn on each T2W image: (a) Red: Prostate cancer within benign prostatic hyperplasia (BPH) nodule presenting as mixed intensity, Blue: Mixed stromal and glandular BPH nodule; (b) Red: Prostate cancer (hypointense lesion), Green: normal PZ; (c) Blue: PZ nodule with atrophic dilated gland, Yellow: Anterior fibromuscular stroma; and (d) Purple: Prostatic intraepithelial neoplasia (PIN). Solid line denotes PIN in the histology.

Figure 3 shows a high-resolution T2W MRI prostate specimen (3a) from a 71-year-old patient with corresponding histology slice (3b), calculated T2 maps (3c), and ADC maps (3d). The figure demonstrates that the normal PZ, PZ prostate cancer, and stromal benign prostatic hyperplasia (BPH) nodule could all be matched on MRI and histology. T2 and ADC values were smaller in regions of cancer than in regions of normal tissue.

Figure 3.

For a 71-year-old patient, (a) the ex vivo T2W fast spin echo prostate images, showing a low signal intensity lesion in the right PZ with a few high signal intensity foci suggestive of malignancy as well as a mixed signal intensity BPH nodule in right side; (b) approximately corresponding histology slices; (c) T2 map; and (d) ADC map. On histology slides everything marked as dotted line is cancer. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Figure 4 shows a high-resolution T2W MRI prostate specimen (4a) from a 63-year-old patient with corresponding histology slice (4b), calculated T2 maps (4c), and ADC maps (4d). The detailed features in a mixed BPH nodule and normal PZ matched well on MRI and histology. This region did not contain cancer, and T2 values were higher than in cancers, but the ADC values were comparable to those in the cancer found in the prostate of patient in Fig. 3.

Figure 4.

For a 63-year-old patient, (a) the ex vivo T2W fast spine echo prostate images, showing multiple, large mixed signal intensity BPH nodules bilaterally compressing the PZ with no evidence of malignancy; (b) approximately corresponding histology slice; (c) T2 map; and (d) ADC map. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Finally, Fig. 5 shows the average T2 and ADC values and the standard deviations for normal PZ and PZ prostate cancers. In the normal PZ, the average T2 value is about 10 msec longer (P < 0.03) and the average ADC value is about 1.4 times larger (P < 0.002) than in the PZ prostate cancers.

Figure 5.

The averaged T2 and ADC values (±SD) for normal peripheral zone and peripheral zone prostate cancer.

DISCUSSION

In this study we demonstrated that high-resolution, high-field strength ex vivo prostate MR images show excellent anatomic detail, which could be easily correlated with standard prostate histopathology. In the PZ, T2 was considerably shorter and ADC was significantly smaller for cancers than for normal prostatic tissue. These preliminary results suggest that some of the same contrast mechanisms that are used to identify prostate cancers ex vivo also can be used to find cancers in vivo. Our results were consistent with a previous study at low field by Schiebler et al (17) that demonstrated BPH nodules showed low signal intensity similar to that of cancer.

Langer et al (18) measured average T2 values at 1.5 T in patient normal and tumor prostate in PZ were about 111.6 msec and 88.7 msec, respectively. These values were about 40 msec to 20 msec longer than those we measured in the ex vivo prostate specimen at 9.4 T. The ADC values measured in normal and tumor prostate (1.47 and 1.28 (×10−3 mm2/s)) in PZ at 1.5T were about two to three times larger than those we measured in ex vivo specimens at 9.4T (0.69 and 0.48 (×10−3 mm2/s)). The large differences in ADC measured in vivo and ex vivo may reflect microstructural changes following surgical excision, and in addition, modest temperature-dependent changes in ADC are expected. Although absolute values differ, the trend of lower T2s and ADCs in cancer is the same for in vivo and ex vivo scans. This consistency suggests that the fresh ex vivo prostate is a good model for the in vivo prostate in terms of T2-contrast and diffusion-based contrast. Therefore, studies of ex vivo prostate may guide the development of improved methods for in vivo scans. High-resolution ex vivo imaging has the potential to serve as an aid to pathologists to improve the sensitivity and specificity with which tissue can be evaluated—while at the same time increasing efficiency. As a result, ex vivo imaging could increase the probability that patients will receive optimal treatment postsurgery.

The present results suggest that discrimination between benign tissue and cancers based on T2, ADC, and perhaps other MRI parameters is similar in in vivo and ex vivo images. If this proves to be the case in larger studies, this means that ex vivo imaging can be used to improve understanding of MRI-detectable features that identify high-grade cancers and differentiate them from lower-grade cancers and benign lesions. This could lead to improved acquisition and analysis of in vivo data that will increase sensitivity and specificity.

In the research reported here anatomic features on ex vivo MR images were identified on histology and in vivo MR images by visual inspection. However, work is under way to fuse in vivo images, ex vivo images, and histology using accurate image registration methods including deformable registration. These methods would facilitate use of MR images by pathologists to guide more accurate and rapid evaluation of surgical specimens. However, high-field MRI will not replace the microscope to assess the margins status and establish the final pathological stage. The idea of correlating ex vivo MRI findings with histology still remains a valuable idea at this stage for research purposes to improve in vivo MRI accuracy for tumor foci localization and evaluation of cancer aggressiveness.

In summary, larger studies with improved MR and pathological correlation should be performed to more accurately identify MRI parameters or combinations of parameters that most accurately identify aggressive cancers and guide patient care postsurgery. In addition, spectroscopy should be also performed on the specimen to identify cancers similar to the study by Swindle et al (19). Nevertheless, the present results are encouraging and suggest important clinical applications for high-resolution ex vivo MRI of the prostate acquired at high field.

Ancillary