Explainable AI for CNN-based Prostate Tumor Segmentation in Multi-parametric MRI Correlated to Whole Mount Histopathology

Automatic prostate tumor segmentation is often unable to identify the lesion even if in multi-parametric MRI data is used as input, and the segmentation output is difficult to verify due to the lack of clinically established ground truth images. In this work we use an explainable deep learning model to interpret the predictions of a convolutional neural network (CNN) for prostate tumor segmentation. The CNN uses a U-Net architecture which was trained on multi-parametric MRI data from 122 patients to automatically segment the prostate gland and prostate tumor lesions. In addition, co-registered ground truth data from whole mount histopathology images were available in 15 patients that were used as a test set during CNN testing. To be able to interpret the segmentation results of the CNN, heat maps were generated using the Gradient Weighted Class Activation Map (Grad-CAM) method. With the CNN a mean Dice Sorensen Coefficient for the prostate gland and the tumor lesions of 0.62 and 0.31 with the radiologist drawn ground truth and 0.32 with wholemount histology ground truth for tumor lesions could be achieved. Dice Sorensen Coefficient between CNN predictions and manual segmentations from MRI and histology data were not significantly different. In the prostate the Grad-CAM heat maps could differentiate between tumor and healthy prostate tissue, which indicates that the image information in the tumor was essential for the CNN segmentation.

Frequency of Significant Prostate Cancer (Gleason Score ≥ 7) in Suspected Patients Having Pi-Rads Score ≥4: Descriptive Study

Background: Most common malignancy among males is prostate cancer causing many deaths. Aim: To determine the diagnostic accuracy of PI-RADS ≥4 lesions in predicting prostate tumor keeping histopathology as gold standard. Study Design: Cross-sectional validation. Methodology: The current project was conducted at Department of Radiology, Armed Forces Institute of Radiology and Imaging, Rawalpindi. Total 114 patients suspicion of prostate carcinoma between 40 to 80 years of age were included. Patients with already diagnosed carcinoma prostate and with inadequate biopsy specimens for diagnosing prostate cancer were excluded. After including the patients in this study, all patients were undergone MRI imaging findings to calculate PI-RADs score as per operational definition. After that biopsy specimens were taken and sent to the histopathology department for determination of Gleason score (GS), a patient was labelled as having significant prostate tumor. Results: Sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy of PI-RADS ≥4 lesions in predicting prostate tumor keeping histopathology as gold standard was 84.85%, 83.33%, 87.50%, 80% and 84.21% respectively. Conclusion: PI-RADS ≥4 is non-invasive modality of choice with high diagnostic accuracy in detecting ca prostate. Keywords: Prostate Cancer, Prostate Imaging Reporting and Sensitivity

Bone Marrow-Derived Mesenchymal Stem Cells Migrate toward Hormone-Insensitive Prostate Tumor Cells Expressing TGF-β via N-Cadherin

The prostate tumor microenvironment plays important roles in the metastasis and hormone-insensitive re-growth of tumor cells. Bone marrow-derived mesenchymal stem cells (BM-MSCs) are recruited into prostate tumors to facilitate tumor microenvironment formation. However, the specific intrinsic molecules mediating BM-MSCs’ migration to prostate tumors are unknown. BM-MSCs’ migration toward a conditioned medium (CM) of hormone-insensitive (PC3 and DU145) or hormone-sensitive (LNCaP) prostate tumor cells was investigated using a three-dimensional cell migration assay and a transwell migration assay. PC3 and DU145 expressed transforming growth factor-β (TGF-β), but LNCaP did not. Regardless of TGF-β expression, BM-MSCs migrated toward the CM of PC3, DU145, or LNCaP. The CM of PC3 or DU145 expressing TGF-β increased the phosphorylation of Smad2/3 in BM-MSCs. Inactivation of TGF-β signaling in BM-MSCs using TGF-β type 1 receptor (TGFBR1) inhibitors, SB505124, or SB431542 did not allow BM-MSCs to migrate toward the CM. The CM of PC3 or DU145 enhanced N-cadherin expression on BM-MSCs, but the LNCaP CM did not. SB505124, SB431542, and TGFBR1 knockdown prevented an increase in N-cadherin expression. N-cadherin knockdown inhibited the collective migration of BM-MSCs toward the PC3 CM. We identified N-cadherin as a mediator of BM-MSCs’ migration toward hormone-insensitive prostate tumor cells expressing TGF-β and introduced a novel strategy for controlling and re-engineering the prostate tumor microenvironment.

Multiphoton Microscopy for Identifying Collagen Signatures Associated with Biochemical Recurrence in Prostate Cancer Patients

Prostate cancer is a heterogeneous disease that remains dormant for long periods or acts aggressively with poor clinical outcomes. Identifying aggressive prostate tumor behavior using current glandular-focused histopathological criteria is challenging. Recent evidence has implicated the stroma in modulating prostate tumor behavior and in predicting post-surgical outcomes. However, the emergence of stromal signatures has been limited, due in part to the lack of adoption of imaging modalities for stromal-specific profiling. Herein, label-free multiphoton microscopy (MPM), with its ability to image tissue with stromal-specific contrast, is used to identify prostate stromal features associated with aggressive tumor behavior and clinical outcome. MPM was performed on unstained prostatectomy specimens from 59 patients and on biopsy specimens from 17 patients with known post-surgery recurrence status. MPM-identified collagen content, organization, and morphological tumor signatures were extracted for each patient and screened for association with recurrent disease. Compared to tumors from patients whose disease did not recur, tumors from patients with recurrent disease exhibited higher MPM-identified collagen amount and collagen fiber intensity signal and width. Our study shows an association between MPM-identified stromal collagen features of prostate tumors and post-surgical disease recurrence, suggesting their potential for prostate cancer risk assessment.