PCAI is the result of more than 20 years of medical research at Uppsala University, Akademiska sjukhuset, Martini-Klinik and UKE Hamburg Partners. A summary of what the studies show is available under clinical results Clinical results. Below is a complete list of scientific publications, in the order in which the work developed.
The development, publication by publication
In our development we used, from the very beginning, the large biobank at Martini-Klinik described in:
- Sauter G et al., Clinical Utility of Quantitative Gleason Grading in Prostate Biopsies and Prostatectomy Specimens. European Urology, Volume 69, Issue 4, 2016, Pages 592–598. doi.org/10.1016/j.eururo.2015.10.029
A set of digitised biopsies from Uppsala University Hospital that we made publicly available:
- Walhagen P, Röbeck P, Bengtsson E, Busch C, Häggman M. Spear Prostate Biopsy 2020 (SPROB20) (2020). datahub.aida.scilifelab.se/10.23698/aida/sprob20
The first version of our system:
- Dietrich E, Fuhlert P, Ernst A, Sauter G, Lennartz M, Stiehl HS, Zimmermann M, Bonn S. Towards explainable end-to-end prostate cancer relapse prediction from H&E images combining self-attention multiple instance learning with a recurrent neural network. Proceedings of Machine Learning Research, 158:38–53, 2021
Our team in Uppsala–Stockholm published our first outcome assessments:
- Walhagen P, Bengtsson E, Lennartz M, Sauter G, Busch C. AI-based prostate analysis system trained without human supervision to predict patient outcome from tissue samples. J Pathol Inform. 2022 Sep 8;13:100137. doi: 10.1016/j.jpi.2022.100137
Robustness is a key question when we develop a system for clinical work. An analysis of the state of the art in the literature in that area led to the published review article:
- Brehler M, Walhagen P, Busch C, Bonn S, Bengtsson E. Difficulties and Recommendations for AI-Based Prediction of Prostate Cancer Aggressiveness in Digital Pathology. Medical Research Archives. 2023;11(11). doi.org/10.18103/mra.v11i10.4586
The current robust PCAI (Prostate Cancer Aggressiveness Index) was then created, and a first manuscript was submitted to MedRxiv:
- Westhaeusser F, Fuhlert P, Dietrich E, Lennartz M, Khatri R, Kaiser N, Röbeck P, Bülow R, von Stillfried S, Witte A, Ladjevardi S, Drotte A, Severgardh P, Baumbach J, Puelles VG, Häggman M, Brehler M, Boor P, Walhagen P, Dragomir A, Busch C, Graefen M, Bengtsson E, Sauter G, Zimmermann M, Bonn S. Robust, credible, and interpretable AI-based histopathological prostate cancer grading. MedRxiv 2024.07.09.24310082. doi.org/10.1101/2024.07.09.24310082
Main publication
Our most recent accepted article in Medical Image Analysis, November 2025:
- Fuhlert P, Westhaeusser F, Dietrich E, Lennartz M, Khatri R, Kaiser N, Röbeck P, Bülow R, von Stillfried S, Witte A, Ladjevardi S, Drotte A, Severgardh P, Baumbach J, Puelles VG, Häggman M, Brehler M, Boor P, Walhagen P, Dragomir A, Busch C, Graefen M, Bengtsson E, Sauter G, Zimmermann M, Bonn S. A systematic analysis of the impact of data variation on AI-based histopathological grading of prostate cancer. Medical Image Analysis, november 2025.
That is the study that forms the basis for PCAI in its current form, and the one that describes how the method holds up when sample handling varies between laboratories.
The research continues
Research is under way to develop the analysis with parameters such as PSA PSA levels, Pi-RADS and the patient's age, which will further improve the precision and personalisation of treatment recommendations.
The work is carried out together with the institutions that have been involved from the beginning Partners, and the background to the collaboration is described under our story.
Related: Clinical results · How PCAI works · PCAI vs. Gleason
Would you like to know more?
We are happy to go through what the research means in your particular case. The initial meeting is free of charge. If you feel afterwards that this is right for you, you can simply order a second opinion.