Predicting Prostate Cancer Molecular Subtype With Deep Learning on Histopathologic Images.

Microscopic examination of prostate cancer has failed to reveal a reproducible association between molecular and morphologic features. However, deep-learning algorithms trained on hematoxylin and eosin (H&E)-stained whole slide images (WSI) may outperform the human eye and help to screen for clinically-relevant genomic alterations. We created deep-learning algorithms to identify prostate tumors with underlying ETS-related gene (ERG) fusions or PTEN deletions using the following 4 stages: (1) automated tumor identification, (2) feature representation learning, (3) classification, and (4) explainability map generation. A novel transformer-based hierarchical architecture was trained on a single representative WSI of the dominant tumor nodule from a radical prostatectomy (RP) cohort with known ERG/PTEN status (n = 224 and n = 205, respectively). Two distinct vision transformer-based networks were used for feature extraction, and a distinct transformer-based model was used for classification. The ERG algorithm performance was validated across 3 RP cohorts, including 64 WSI from the pretraining cohort (AUC, 0.91) and 248 and 375 WSI from 2 independent RP cohorts (AUC, 0.86 and 0.89, respectively). In addition, we tested the ERG algorithm performance in 2 needle biopsy cohorts comprised of 179 and 148 WSI (AUC, 0.78 and 0.80, respectively). Focusing on cases with homogeneous (clonal) PTEN status, PTEN algorithm performance was assessed using 50 WSI reserved from the pretraining cohort (AUC, 0.81), 201 and 337 WSI from 2 independent RP cohorts (AUC, 0.72 and 0.80, respectively), and 151 WSI from a needle biopsy cohort (AUC, 0.75). For explainability, the PTEN algorithm was also applied to 19 WSI with heterogeneous (subclonal) PTEN loss, where the percentage tumor area with predicted PTEN loss correlated with that based on immunohistochemistry (r = 0.58, P = .0097). These deep-learning algorithms to predict ERG/PTEN status prove that H&E images can be used to screen for underlying genomic alterations in prostate cancer.

Identifying Men Who Can Remain on Active Surveillance Despite Biopsy Reclassification to Grade Group 2 Prostate Cancer.

PURPOSE: Men on active surveillance with Grade Group 1 prostate cancer who reclassify to Grade Group 2 on surveillance biopsy often leave active surveillance. We aimed to identify subgroups of men who can safely remain on active surveillance despite preoperative reclassification to Grade Group 2. MATERIALS AND METHODS: We studied 249 active surveillance patients with surveillance biopsies classified as Grade Group 1 or Grade Group 2 who underwent radical prostatectomy. Perineural invasion, cancer volume, linear length and maximum percentage of Gleason pattern 4, and prostate-specific antigen density were evaluated. Radical prostatectomy adverse pathology was defined by any of: pN1; ≥pT3; ≥Grade Group 2 with ≥20% Gleason pattern 4; intraductal carcinoma; large cribriform glands. RESULTS: A multivariable logistic regression model incorporating prostate-specific antigen density and perineural invasion stratified radical prostatectomy adverse pathology risk among Grade Group 1 and Grade Group 2 active surveillance patients. 57% (39/68) of Grade Group 1 men reclassified to Grade Group 2 while on active surveillance had favorable radical prostatectomy pathology. Those without biopsy perineural invasion and with low prostate-specific antigen density were more likely to have favorable radical prostatectomy pathology. CONCLUSIONS: Most Grade Group 1 men who enter active surveillance and subsequently reclassify to Grade Group 2 have favorable findings at radical prostatectomy and can remain on active surveillance. Among patients reclassified to Grade Group 2, those with low prostate-specific antigen density and without perineural invasion had the lowest risk of radical prostatectomy adverse pathology, comparable to (or below) that of Grade Group 1 patients who were not reclassified to Grade Group 2 preoperatively. Prostate-specific antigen density and perineural invasion stratify risk in active surveillance patients reclassified to Grade Group 2 and, if concordant with other clinicopathological and radiographic findings, can enable more patients to remain on active surveillance. Reclassification to Grade Group 2 alone should not disqualify men from remaining on active surveillance.

Utilization of NKX3.1, P501S, Prostate-Specific Antigen, and Steroidogenic Factor 1 to Distinguish Malignant Leydig Cell Tumor From Metastatic Prostatic Adenocarcinoma to the Testis.

CONTEXT.­: A recent study demonstrated that NKX3.1-positive staining can uncommonly be seen in testicular Sertoli cell tumors (1 of 4 cases). Also, it was reported that 2 of 3 Leydig cell tumors of the testis showed diffuse cytoplasmic staining for P501S, although it was unclear whether it was specific granular staining that defines true positivity. However, Sertoli cell tumors do not typically pose a diagnostic dilemma with metastatic prostate carcinoma to the testis. In contrast, malignant Leydig cell tumors, which are exceedingly rare, can closely resemble Gleason score 5 + 5 = 10 prostatic adenocarcinoma metastatic to the testis. OBJECTIVE.­: To evaluate the expression of prostate markers in malignant Leydig cell tumors and steroidogenic factor 1 (SF-1) in high-grade prostate adenocarcinoma, as no data are currently published on these topics. DESIGN.­: Fifteen cases of malignant Leydig cell tumor were collected from 2 large genitourinary pathology consult services in the United States from 1991 to 2019. RESULTS.­: All 15 cases were negative immunohistochemically for NKX3.1, and all 9 with available additional material were negative for prostate-specific antigen and P501S and positive for SF-1. SF-1 was negative immunohistochemically in a tissue microarray with cases of high-grade prostatic adenocarcinoma. CONCLUSIONS.­: The diagnosis of malignant Leydig cell tumor and its distinction from metastatic adenocarcinoma to the testis can be made immunohistochemically on the basis of SF-1 positivity and negativity for NKX3.1.