Pathologist-like explainable AI for interpretable Gleason grading in prostate cancer
Recommended Citation
Mittmann G, Laiouar-Pedari S, Mehrtens HA, Haggenmüller S, Bucher TC, Chanda T, Gaisa NT, Wagner M, Klamminger GG, Rau TT, Neppl C, Compérat EM, Gocht A, Haemmerle M, Rupp NJ, Westhoff J, Krücken I, Seidl M, Schürch CM, Bauer M, Solass W, Tam YC, Weber F, Grobholz R, Augustyniak J, Kalinski T, Hörner C, Mertz KD, Döring C, Erbersdobler A, Deubler G, Bremmer F, Sommer U, Brodhun M, Griffin J, Lenon MSL, Trpkov K, Cheng L, Chen F, Levi A, Cai G, Nguyen TQ, Amin A, Cimadamore A, Shabaik A, Manucha V, Ahmad N, Messias N, Sanguedolce F, Taheri D, Baraban E, Jia L, Shah RB, Siadat F, Swarbrick N, Park K, Hassan O, Sakhaie S, Downes MR, Miyamoto H, Williamson SR, Holland-Letz T, Wies C, Schneider CV, Kather JN, Tolkach Y, and Brinker TJ. Pathologist-like explainable AI for interpretable Gleason grading in prostate cancer. Nat Commun 2025;16(1):8959.
Document Type
Article
Publication Date
10-8-2025
Publication Title
Nat Commun
Keywords
Humans, Male, Prostatic Neoplasms, Neoplasm Grading, Artificial Intelligence, Pathologists, Observer Variation, Prostate
Abstract
The aggressiveness of prostate cancer is primarily assessed from histopathological data using the Gleason scoring system. Conventional artificial intelligence (AI) approaches can predict Gleason scores, but often lack explainability, which may limit clinical acceptance. Here, we present an alternative, inherently explainable AI that circumvents the need for post-hoc explainability methods. The model was trained on 1,015 tissue microarray core images, annotated with detailed pattern descriptions by 54 international pathologists following standardized guidelines. It uses pathologist-defined terminology and was trained using soft labels to capture data uncertainty. This approach enables robust Gleason pattern segmentation despite high interobserver variability. The model achieved comparable or superior performance to direct Gleason pattern segmentation (Dice score: 0.713±0.003 vs. 0.691±0.010) while providing interpretable outputs. We release this dataset to encourage further research on segmentation in medical tasks with high subjectivity and to deepen insights into pathologists' reasoning.
Medical Subject Headings
ambulatory care; antimicrobial stewardship; outpatients; rapid diagnostics; upper respiratory tract infections; Humans; Male; Prostatic Neoplasms; Neoplasm Grading; Artificial Intelligence; Pathologists; Observer Variation; Prostate
PubMed ID
41062516
Volume
16
Issue
1
First Page
8959
Last Page
8959
