Robust, credible, and interpretable AI-based histopathological prostate cancer grading

medRxiv

Abstract

Background: Prostate cancer (PCa) is among the most common cancers in men and its diagnosis requires the histopathological evaluation of biopsies by human experts. While several recent artificial intelligence-based (AI) approaches have reached human expert-level PCa grading, they often display significantly reduced performance on external datasets. This reduced performance can be caused by variations in sample preparation, for instance the staining protocol, section thickness, or scanner used. Another limiting factor of contemporary AI-based PCa grading is the prediction of ISUP grades, which leads to the perpetuation of human annotation errors. Methods: We developed the prostate cancer aggressiveness index (PCAI), an AI-based PCa detection and grading framework that is trained on objective patient outcome, rather than subjective ISUP grades. We designed PCAI as a clinical application, containing algorithmic modules that offer robustness to data variation, medical interpretability, and a measure of prediction confidence. To train and evaluate PCAI, we generated a multicentric, retrospective, observational trial consisting of six cohorts with 25,591 patients, 83,864 images, and 5 years of median follow-up from 5 different centers and 3 countries. This includes a high-variance dataset of 8,157 patients and 28,236 images with variations in sample thickness, staining protocol, and scanner, allowing for the systematic evaluation and optimization of model robustness to data variation. The performance of PCAI was assessed on three external test cohorts from two countries, comprising 2,255 patients and 9,437 images. Findings: Using our high-variance datasets, we show how differences in sample processing, particularly slide thickness and staining time, significantly reduce the performance of AI-based PCa grading by up to 6.2 percentage points in the concordance index (C-index). We show how a select set of algorithmic improvements, including domain adversarial training, conferred robustness to data variation, interpretability, and a measure of credibility to PCAI. These changes lead to significant prediction improvement across two biopsy cohorts and one TMA cohort, systematically exceeding expert ISUP grading in C-index and AUROC by up to 22 percentage points. Interpretation: Data variation poses serious risks for AI-based histopathological PCa grading, even when models are trained on large datasets. Algorithmic improvements for model robustness, interpretability, credibility, and training on high-variance data as well as outcome-based severity prediction gives rise to robust models with above ISUP-level PCa grading performance.

Authors 26

1. Fabian Westhaeusser

   Universität Hamburg · University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

   Spearpoint Analytics AB, Stockholm, Sweden

2. Patrick Fuhlert

   Universität Hamburg · University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

   Spearpoint Analytics AB, Stockholm, Sweden

3. Esther Dietrich

   Universität Hamburg · University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

4. Maximilian Lennartz

   University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   Institute of Pathology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

5. Robin Khatri

   Universität Hamburg · University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

6. Nico Kaiser

   Universität Hamburg · University Medical Center Hamburg-Eppendorf

   Affiliation as printed

   III. Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

   Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

7. Pontus Röbeck

   Uppsala University Hospital

   Affiliation as printed

   Department of Urology, Uppsala University Hospital, Uppsala, Sweden

8. Roman David Bülow Aachen Institute of Pathology

   RWTH Aachen University · Universitätsklinikum Aachen

   Affiliation as printed

   Institute of Pathology, RWTH Aachen University Hospital, Aachen, Germany

9. Saskia Freifrau von Stillfried Aachen Institute of Pathology

   RWTH Aachen University · Universitätsklinikum Aachen

   Affiliation as printed

   Institute of Pathology, RWTH Aachen University Hospital, Aachen, Germany

10. Anja Witte

    Universität Hamburg · University Medical Center Hamburg-Eppendorf

    Affiliation as printed

    Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

11. Sam Ladjevardi

    Uppsala University Hospital

    Affiliation as printed

    Department of Urology, Uppsala University Hospital, Uppsala, Sweden

12. Anders Drotte
    Affiliation as printed

    Spearpoint Analytics AB, Stockholm, Sweden

13. Peter Severgårdh
    Affiliation as printed

    Spearpoint Analytics AB, Stockholm, Sweden

14. Jan Baumbach

    Universität Hamburg

    Affiliation as printed

    Institute of Computational Systems Biology, University of Hamburg, Germany

15. Victor G. Puelles

    Aarhus University · University Medical Center Hamburg-Eppendorf · Aarhus University Hospital

    Affiliation as printed

    Department of Clinical Medicine, Aarhus University, Aarhus, Denmark

    Department of Pathology, Aarhus University Hospital, Aarhus, Denmark

    III. Department of Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

16. Mikael Häggman

    Uppsala University Hospital

    Affiliation as printed

    Department of Urology, Uppsala University Hospital, Uppsala, Sweden

17. Michael Brehler

    Universität Hamburg · University Medical Center Hamburg-Eppendorf

    Affiliation as printed

    Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

18. Peter Boor Aachen Institute of Pathology

    RWTH Aachen University · Universitätsklinikum Aachen

    Affiliation as printed

    Institute of Pathology, RWTH Aachen University Hospital, Aachen, Germany

19. Peter Walhagen
    Affiliation as printed

    Spearpoint Analytics AB, Stockholm, Sweden

20. Anca Dragomir

    Uppsala University · Uppsala University Hospital

    Affiliation as printed

    Department of Pathology, Uppsala University Hospital and Department of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden

21. Christer Busch

    Uppsala University Hospital

    Affiliation as printed

    Department of Urology, Uppsala University Hospital, Uppsala, Sweden

    Spearpoint Analytics AB, Stockholm, Sweden

22. Markus Graefen

    Universität Hamburg · University Medical Center Hamburg-Eppendorf · Martini-Klinik

    Affiliation as printed

    Martini-Klinik Prostate Cancer Center, University Hospital Hamburg-Eppendorf, Hamburg, Germany

23. Ewert W. Bengtsson

    Uppsala University

    Affiliation as printed

    Spearpoint Analytics AB, Stockholm, Sweden

    Uppsala University, Department of Information Technology, Centre for Image Analysis, Uppsala, Sweden

24. Guido Sauter

    University Medical Center Hamburg-Eppendorf

    Affiliation as printed

    Institute of Pathology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

25. Marina Zimmermann

    Universität Hamburg · University Medical Center Hamburg-Eppendorf

    Affiliation as printed

    Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

26. Stefan Bonn corresponding

    Universität Hamburg · University Medical Center Hamburg-Eppendorf

    Affiliation as printed

    Institute of Medical Systems Biology, Center for Biomedical AI (bAIome), Center for Molecular Neurobiology Hamburg (ZMNH), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

    Spearpoint Analytics AB, Stockholm, Sweden