Journal Article

DKFZ-2025-02060

http://join2-wiki.gsi.de/foswiki/pub/Main/Artwork/join2_logo100x88.png Pathologist-like explainable AI for interpretable Gleason grading in prostate cancer.

Mittmann, G. (First author)DKFZ* ; Laiouar-Pedari, S. (First author)DKFZ* ; Mehrtens, H. A. (First author)DKFZ* ; Haggenmüller, S.DKFZ* ; Bucher, T.DKFZ* ; Chanda, T.DKFZ* ; Gaisa, N. T. ; Wagner, M. ; Klamminger, G. G. ; Rau, T. T. ; Neppl, C. ; Compérat, E. M. ; Gocht, A. ; Haemmerle, M. ; Rupp, N. J. ; Westhoff, J. ; Krücken, I. ; Seidl, M. ; Schürch, C. M. ; Bauer, M. ; Solass, W. ; Tam, Y. C. ; Weber, F. ; Grobholz, R. ; Augustyniak, J. ; Kalinski, T. ; Hörner, C. ; Mertz, K. D. ; Döring, C. ; Erbersdobler, A. ; Deubler, G. ; Bremmer, F. ; Sommer, U. ; Brodhun, M. ; Griffin, J. ; Lenon, M. S. L. ; Trpkov, K. ; Cheng, L. ; Chen, F. ; Levi, A. ; Cai, G. ; Nguyen, T. Q. ; Amin, A. ; Cimadamore, A. ; Shabaik, A. ; Manucha, V. ; Ahmad, N. ; Messias, N. ; Sanguedolce, F. ; Taheri, D. ; Baraban, E. ; Jia, L. ; Shah, R. B. ; Siadat, F. ; Swarbrick, N. ; Park, K. ; Hassan, O. ; Sakhaie, S. ; Downes, M. R. ; Miyamoto, H. ; Williamson, S. R. ; Holland-Letz, T.DKFZ* ; Wies, C.DKFZ* ; Schneider, C. V. ; Kather, J. N. ; Tolkach, Y. ; Brinker, T. (Last author)DKFZ*

2025
Springer Nature [London]

This record in other databases:  

Please use a persistent id in citations: doi:10.1038/s41467-025-64712-4

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.

Keyword(s): Humans (MeSH) ; Male (MeSH) ; Prostatic Neoplasms: pathology (MeSH) ; Prostatic Neoplasms: diagnosis (MeSH) ; Neoplasm Grading: methods (MeSH) ; Artificial Intelligence (MeSH) ; Pathologists (MeSH) ; Observer Variation (MeSH) ; Prostate: pathology (MeSH)

Note: #EA:C140#LA:C140#

---

Contributing Institute(s):

1. Digitale Prävention, Diagnostik und Therapiesteuerung (C140)
2. C060 Biostatistik (C060)

Research Program(s):

1. 313 - Krebsrisikofaktoren und Prävention (POF4-313) (POF4-313)

Appears in the scientific report 2025

---

Database coverage:
; ; ; Article Processing Charges ; BIOSIS Previews ; Biological Abstracts ; Clarivate Analytics Master Journal List ; Current Contents - Agriculture, Biology and Environmental Sciences ; Current Contents - Life Sciences ; Current Contents - Physical, Chemical and Earth Sciences ; DOAJ Seal ; Essential Science Indicators ; Fees ; IF >= 15 ; JCR ; SCOPUS ; Science Citation Index Expanded ; Web of Science Core Collection ; Zoological Record

---