Artificial intelligence (AI) has shown promise for diagnosing prostate cancer in biopsies. However, results have been limited to individual studies, lacking validation in multinational settings. Competitions have been shown to be accelerators for medical imaging innovations, but their impact is hindered by lack of reproducibility and independent validation. With this in mind, we organized the PANDA challenge—the largest histopathology competition to date, joined by 1,290 developers—to catalyze development of reproducible AI algorithms for Gleason grading using 10,616 digitized prostate biopsies. We validated that a diverse set of submitted algorithms reached pathologist-level performance on independent cross-continental cohorts, fully blinded to the algorithm developers. On United States and European external validation sets, the algorithms achieved agreements of 0.862 (quadratically weighted κ, 95% confidence interval (CI), 0.840–0.884) and 0.868 (95% CI, 0.835–0.900) with expert uropathologists. Successful generalization across different patient populations, laboratories and reference standards, achieved by a variety of algorithmic approaches, warrants evaluating AI-based Gleason grading in prospective clinical trials.

The Gleason score is the most important prognostic marker for prostate cancer patients, but it suffers from significant observer variability. Artificial intelligence (AI) systems based on deep learning can achieve pathologist-level performance at Gleason grading. However, the performance of such systems can degrade in the presence of artifacts, foreign tissue, or other anomalies. Pathologists integrating their expertise with feedback from an AI system could result in a synergy that outperforms both the individual pathologist and the system. Despite the hype around AI assistance, existing literature on this topic within the pathology domain is limited. We investigated the value of AI assistance for grading prostate biopsies. A panel of 14 observers graded 160 biopsies with and without AI assistance. Using AI, the agreement of the panel with an expert reference standard increased significantly (quadratically weighted Cohen's kappa, 0.799 vs. 0.872; p = 0.019). On an external validation set of 87 cases, the panel showed a significant increase in agreement with a panel of international experts in prostate pathology (quadratically weighted Cohen's kappa, 0.733 vs. 0.786; p = 0.003). In both experiments, on a grouplevel, AI-assisted pathologists outperformed the unassisted pathologists and the standalone AI system. Our results show the potential of AI systems for Gleason grading, but more importantly, show the benefits of pathologist-AI synergy. Members of the ISUP Pathology Imagebase Expert Panel are listed below Acknowledgements.

No abstract

Current endocrine treatment for advanced prostate cancer does not result in a complete ablation of adrenal androgens. Adrenal androgens can be metabolized by prostate cancer cells, which is one of the mechanisms associated with progression to castration-resistant prostate cancer (CRPC). Aldo-keto reductase family 1 member C3 (AKR1C3) is a steroidogenic enzyme that plays a crucial role in the conversion of adrenal androgen dehydroepiandrosterone (DHEA) into high-affinity ligands for the androgen receptor (testosterone [T] and dihydrotestosterone [DHT]). The aim of this study was to examine whether AKR1C3 could be used as a marker and therapeutic target for CRPC. AKR1C3 mRNA and protein levels were upregulated in CRPC tissue, compared with benign prostate and primary prostate cancer tissue. High AKR1C3 levels were found only in a subset of CRPC patients. AKR1C3 can be used as a biomarker for active intratumoral steroidogenesis and can be measured in biopsy or transurethral resection of the prostate specimens. DuCaP (a CRPC cell line that has high AKR1C3 expression levels) used and converted DHEA under hormone-depleted conditions into T and DHT. The DHEA-induced growth of DuCaP could be antagonized by indomethacine, an inhibitor of AKR1C3. This study indicates that AKR1C3 can be considered a therapeutic target in a subgroup of patients with high AKR1C3 expression.

De novo mutations are known to play a prominent role in sporadic disorders with reduced fitness. We hypothesize that de novo mutations play an important role in severe male infertility and explain a portion of the genetic causes of this understudied disorder. To test this hypothesis, we utilize trio-based exome sequencing in a cohort of 185 infertile males and their unaffected parents. Following a systematic analysis, 29 of 145 rare (MAF < 0.1%) protein-altering de novo mutations are classified as possibly causative of the male infertility phenotype. We observed a significant enrichment of loss-of-function de novo mutations in loss-of-function-intolerant genes (p-value = 1.00 × 10−5) in infertile men compared to controls. Additionally, we detected a significant increase in predicted pathogenic de novo missense mutations affecting missense-intolerant genes (p-value = 5.01 × 10−4) in contrast to predicted benign de novo mutations. One gene we identify, RBM5, is an essential regulator of male germ cell pre-mRNA splicing and has been previously implicated in male infertility in mice. In a follow-up study, 6 rare pathogenic missense mutations affecting this gene are observed in a cohort of 2,506 infertile patients, whilst we find no such mutations in a cohort of 5,784 fertile men (p-value = 0.03). Our results provide evidence for the role of de novo mutations in severe male infertility and point to new candidate genes affecting fertility.