Imaging the Human Prostate Gland Using 1-μm-Resolution Optical Coherence Tomography

The benchtop μOCT system used in this study has been previously described.²⁵  Briefly, μOCT is conducted with a spectral-domain OCT system with a broad bandwidth supercontinuum light source spanning from 650 to 950 nm, which provides an axial resolution of less than 1 μm in tissue. Light from the broadband source enters a common path interferometer and is split into a reference and sample path. Light on the sample is apodized by a central obscuration, creating a 2-μm-size spot over an extended depth of focus of approximately 300 μm. Light returned from the sample and reference arms are recombined and detected by a spectrometer. The resultant digitized signal is Fourier transformed to create depth-resolved reflectivity profiles (A-scan). A 3-dimensional μOCT data set is created by recording multiple A-scans as the sample beam is scanned along 2 lateral dimensions.

Prostate specimens were acquired from either the Pathology Service at Massachusetts General Hospital (Boston) or the National Disease Research Interchange (Philadelphia, Pennsylvania). Prostate specimens acquired from the Massachusetts General Hospital Pathology Service were collected from patients undergoing whole or partial surgical resections of the prostate. After review of the organ, discarded deidentified surgical specimens were collected for this study, placed in phosphate-buffered saline solution, and imaged within 24 hours of resection. This ex vivo imaging study was approved by the Partners HealthCare (Somerville, Massachusetts) institutional review board (No. 2010P000845). Whole prostate organs received from National Disease Research Interchange were recovered at autopsy from men older than 70 years who did not have a history of prostate surgery, prostate implants, or other prostatic surgical procedures. Specimens were recovered within 24 hours after death, placed in phosphate-buffered saline solution, and shipped on wet ice. Specimens were imaged in less than 48 hours after death to limit tissue degradation. This National Disease Research Interchange study was approved under a separate protocol (No. 2013P001310).

Before imaging, whole prostates were “breadloafed” orthogonal to the urethra in approximately 1-cm sections. Sections were visualized grossly to identify diseased features such as nodules or discoloration of the parenchyma. Regions of interest were cut from the involved tissue. All specimens were trimmed to provide a clean front-imaging plane and a second orthogonal cut was made on the left edge as a means for registering the tissue between μOCT imaging and subsequent histology. After tissue preparation, the prostate specimens were placed on phosphate-buffered saline-soaked gauze before imaging.

Each prostate specimen was placed on a 5-axis stage with the front edge of the specimen aligned with the fast-scanning axis and imaged at room temperature. Because of the depth range of μOCT (approximately 300–500 μm), the scanning beam was advanced approximately 1 mm into tissue from the front edge and the first scan was acquired on the right side with subsequent scans moved to the left. Using galvanometer beam scanners, the μOCT system acquired 512 frames over a 1-mm by 1-mm area. All μOCT image stacks were frame averaged (n = 3) and saved using a logarithmic gray-scale look-up tables.

After imaging, a small ink spot was placed on the back left corner of each specimen. To preserve the orientation during histologic processing, both the bottom side and left edge were painted with tissue-marking ink. Specimens were placed in 10% buffered formalin for a period of not less than 48 hours and submitted for routine paraffin-embedded histology. During embedding, processed specimens were oriented such that the histology sections were in the same plane as the images. Tissue sections (5 μm thick) were captured for hematoxylin-eosin histology at level spacing of 100 μm for up to 20 sections. All histology slides were digitized with a whole slide scanner (Nanozoomer, Hamamatsu Photonics, Hamamatsu, Japan) at a native resolution of ×40.

Mosaic images of digitized histology sections and μOCT image stacks from different sites on the same tissue were created in ImageJ (US National Institutes of Health, Bethesda, Maryland), which allowed all histology images be viewed and compared with μOCT image stacks. For each specimen, a diagnosis was rendered by a pathologist. Then, the mosaic of μOCT image stacks was reviewed to identify landmark features, which could then be searched for within the histology mosaic. When matches were not exact, the match was termed representative. Once corresponding features were identified, the appropriate histology section was determined and opened with the Nanozoomer NDPVIEW (Hamamatsu Photonics) application for visualizing the digitized histology at high resolution, so that the scale was approximately the same as that of the μOCT data. High-resolution images of both the matching μOCT and digitized histology frames were selected as matches. Features in the matching data sets were identified and annotated on the histology and matching μOCT image data.

A total of 18 samples were excised from 11 specimens from the human prostate resections. Three whole prostates were obtained at autopsy, generating an additional 32 samples. Per histopathology assessment, 7 (14%) of the 50 samples contained prostate cancer, and 43 samples (86%) were benign.

Benign prostate imaging by μOCT was characterized by features that include prostatic glands that were separated by large amounts of prostate stroma (Figure 1, A and B), the presence of large amounts of prostatic secretions (Figure 2, A and B), and corpora amylacea (Figure 3, A and B). The epithelium of the prostate was frequently seen as a moderately backscattering layer lining the glands (arrows in Figures 1, A and B; and 2, A and B). Sometimes, cell membranes could be faintly visualized (Figure 4, A and B), but this feature was not commonly identified. It was not possible to clearly differentiate basal from luminal epithelial cells.

Figure 5, A depicts a μOCT image of high-grade (Gleason score 4 + 4) invasive prostate cancer adjacent to corresponding histology (Figure 5, B). Individual cancer cells can be visualized in the μOCT image as clear open spaces with little μOCT signal. When compared with histology, those cells exhibit clear cell changes, which is an appearance that is seen for some prostate cancers. The μOCT and histology images also show an excellent example of corpora amylacea (red arrow) which serves as a fiducial marker for both μOCT and histology images.

A μOCT image of a Gleason grade 3 + 4 nodule of prostate cancer is shown in Figure 6, A, with corresponding histology (Figure 6, B and C). The cancer can be seen by μOCT as a nodule within surrounding uninvolved prostatic parenchyma (red arrows in Figure 6, A and B). Linear structures in the μOCT image (yellow arrows in Figure 6, A) are likely crystalloid structures that are frequently seen in prostate cancer, as also seen in the corresponding histology (blue arrows in Figure 6, C).

The preliminary results of this study provide information on the potential for use of μOCT to diagnose prostate cancer. High-quality image data were obtained at a penetration depth up to 500 μm. Invasive cancer, clear cell changes, cancer-related crystalloids, and corpora amylacea were evident by μOCT and were confirmed by the corresponding histopathology. Our findings show that the micrometer-scale resolution of this technology does indeed provide cellular-level detail that is diagnostically relevant. The improvements afforded by the high resolution and relatively high imaging speed of this technology make it a promising candidate for large area needle-based prostate diagnosis.

Even though the results presented here are encouraging, there are many cellular features seen by conventional brightfield histopathology that are not distinguishable by μOCT. Nuclei are not clearly discernable from cytoplasm and basal cells cannot be differentiated from luminal cells. These limitations of μOCT as currently configured could make it difficult to discriminate certain benign from malignant prostate glands based on standard histomorphometric patterns. As a result, clearer visualization of architectural morphology and associated findings (eg, crystals) seen uniquely by the higher-resolution technology will govern the differences in diagnostic performance between μOCT and conventional OCT.

The μOCT-imaging penetration depth found in this study ranged from 300 to 500 μm. Although that penetration depth is greater than that seen with other in vivo microscopy modalities (eg, confocal microscopy), it is insufficient for imaging the entire prostate with a realistic number of needle insertions, nominally 24. It is well known that OCT imaging at longer wavelengths increases penetration depth significantly and imaging within the 1.3 μm or 1.6 μm spectral windows will likely double or triple that value. These considerations suggest that the development of μOCT at longer wavelengths is merited.

The true potential of μOCT can only be investigated when this technology is implemented in a small-diameter probe that is capable of being inserted into a prostate biopsy needle. By providing in vivo microscopy images through a needle for large regions of the prostate in real time, this advance could significantly decrease biopsy sampling error with or without ultrasound or magnetic resonance imaging–based²⁸  image guidance. Recently, methods for creating such high-resolution probes have been reported,^27,29  making the translation of needle-based μOCT possible. Once μOCT is implemented in vivo, the findings in this study can be confirmed, and the potential utility of this imaging modality for comprehensive in situ prostate diagnosis can be validated clinically.