Oncology

Pathologist-like explainable AI for interpretable Gleason grading in prostate cancer

Nature Communications

Nature Communications

Research Authors: Gesa Mittmann, Sara Laiouar-Pedari, Hendrik A Mehrtens, Sarah Haggenmüller, Tabea-Clara Bucher, Tirtha Chanda, Nadine T Gaisa, Mathias Wagner, Gilbert Georg Klamminger, Tilman T Rau, Christina Neppl, Eva Maria Compérat, Andreas Gocht, Monika Haemmerle, Niels J Rupp, Jula Westhoff, Irene Krücken, Maximilian Seidl, Christian M Schürch, Marcus Bauer, Wiebke Solass, Yu Chun Tam, Florian Weber, Rainer Grobholz, Jaroslaw Augustyniak, Thomas Kalinski, Christian Hörner, Kirsten D Mertz, Constanze Döring, Andreas Erbersdobler, Gabriele Deubler, Felix Bremmer, Ulrich Sommer, Michael Brodhun, Jon Griffin, Maria Sarah L Lenon, Kiril Trpkov, Liang Cheng, Fei Chen, Angelique Levi, Guoping Cai, Tri Q Nguyen, Ali Amin, Alessia Cimadamore, Ahmed Shabaik, Varsha Manucha, Nazeel Ahmad, Nidia Messias, Francesca Sanguedolce, Diana Taheri, Ezra Baraban, Liwei Jia, Rajal B Shah, Farshid Siadat, Nicole Swarbrick, Kyung Park, Oudai Hassan, Siamak Sakhaie, Michelle R Downes, Hiroshi Miyamoto, Sean R Williamson, Tim Holland-Letz, Christoph Wies, Carolin V Schneider, Jakob Nikolas Kather, Yuri Tolkach, Titus J Brinker

Research Authors: Gesa Mittmann, Sara Laiouar-Pedari, Hendrik A Mehrtens, Sarah Haggenmüller, Tabea-Clara Bucher, Tirtha Chanda, Nadine T Gaisa, Mathias Wagner, Gilbert Georg Klamminger, Tilman T Rau, Christina Neppl, Eva Maria Compérat, Andreas Gocht, Monika Haemmerle, Niels J Rupp, Jula Westhoff, Irene Krücken, Maximilian Seidl, Christian M Schürch, Marcus Bauer, Wiebke Solass, Yu Chun Tam, Florian Weber, Rainer Grobholz, Jaroslaw Augustyniak, Thomas Kalinski, Christian Hörner, Kirsten D Mertz, Constanze Döring, Andreas Erbersdobler, Gabriele Deubler, Felix Bremmer, Ulrich Sommer, Michael Brodhun, Jon Griffin, Maria Sarah L Lenon, Kiril Trpkov, Liang Cheng, Fei Chen, Angelique Levi, Guoping Cai, Tri Q Nguyen, Ali Amin, Alessia Cimadamore, Ahmed Shabaik, Varsha Manucha, Nazeel Ahmad, Nidia Messias, Francesca Sanguedolce, Diana Taheri, Ezra Baraban, Liwei Jia, Rajal B Shah, Farshid Siadat, Nicole Swarbrick, Kyung Park, Oudai Hassan, Siamak Sakhaie, Michelle R Downes, Hiroshi Miyamoto, Sean R Williamson, Tim Holland-Letz, Christoph Wies, Carolin V Schneider, Jakob Nikolas Kather, Yuri Tolkach, Titus J Brinker

AIIM Authors: Suman Sanghera, Cedric Bruges, Reda Riffi

AIIM Authors: Suman Sanghera, Cedric Bruges, Reda Riffi

Publication Date: Oct 8, 2025

Publication Date: Oct 8, 2025

Comprehensive Summary

This study developed a novel AI approach to predict Gleason scores by training the AI model on over 1,000 tissue microarray core images, each annotated by 54 pathologists using standardized guidelines. In particular, this approach allows for AI to contribute to primary prostate cancer diagnoses through Gleason scoring in a way that resembles the accepted clinical decision-making process for such diagnoses. The researchers found that this AI model achieved comparable scores to the typical diagnoses approach (Dice score: 0.713 vs. 0.691). Overall, this paper provides an important overview of a more clinically trained AI model that was found to produce comparable or superior Gleason scores, providing insight on both a different tool for diagnosis as well as the factors taken into consideration when formulating Gleason scores by pathologists.

Outcomes and Implications

The Gleason score is used to assess tumor aggressiveness and prognosis in patients with prostate cancer. A major limitation of the Gleason scoring system is the interobserver variability, a product of sampling bias and the subjectivity of tumor build. This study developed a novel AI approach in predicting Gleason score and, thus, contributes to prostate cancer diagnosis. Because the study found that this model is comparable/superior to the current method for Gleason score calculation, it may be used as a tool in the future by healthcare providers to more reliably predict Gleason score. It may also be used alongside the traditional Gleason score calculation to provide additional insight into sources of variability and highlight factors that may be overlooked in manual assessments. As a result, this AI method may improve diagnostic accuracy on a patient-to-patient basis, potentially improving treatment methods and ensuring optimal care is provided.

Our mission is to

Connect medicine with AI innovation.

No spam. Only the latest AI breakthroughs, simplified and relevant to your field.

Our mission is to

Connect medicine with AI innovation.

No spam. Only the latest AI breakthroughs, simplified and relevant to your field.

Our mission is to

Connect medicine with AI innovation.

No spam. Only the latest AI breakthroughs, simplified and relevant to your field.

AIIM Research

Articles

© 2025 AIIM. Created by AIIM IT Team

AIIM Research

Articles

© 2025 AIIM. Created by AIIM IT Team

AIIM Research

Articles

© 2025 AIIM. Created by AIIM IT Team