Predicting Prostate Cancer Progression During Active Surveillance Using Longitudinal bpMRI Scans and A Multi-scale Foundation Model

Yifan Wang, Bin Lou, Heinrich von Busch, Robert Grimm, Sanoj Punnen, Dorin Comaniciu, Ali Kamen, Henkjan Huisman, Angela Tong, David Winkel, Tobias Penzkofer, Ivan Shabunin, Moon Hyung Choi, Qingsong Yang, Dieter Szolar, Steven Shea, Fergus Coakley, Mukesh Harisinghani

Keywords: Prostate Cancer Diagnosis, Active Surveillance, Foundation Model, Longitudinal bpMRI

TL;DR: Predicting PCa Progression During AS using Longitudinal bpMRI

Abstract: Active Surveillance (AS) is the recommended management strategy for patients with low- or intermediate-risk Prostate Cancer (PCa), providing a safe alternative that helps avoid the adverse effects of overtreatment. While artificial intelligence (AI)-based models for PCa detection have been extensively studied, their application in AS remains challenging, with limited research addressing the detection of PCa progression in AS scenarios. In this study, we present a novel framework for predicting PCa progression within AS protocols using bi-parametric MRI (bpMRI). Due to the limited availability of longitudinal bpMRI scans (206 patients in our study), we first developed a multi-scale foundation model trained on a large cohort of single-year bpMRI scans, comprising 5,162 patients from 10 different institutions. Building on this foundation model, we designed a three-module framework: (1) a lesion detection module to identify PCa lesions in full bpMRI scans, (2) a lesion classification module to perform detailed analysis of the identified lesion regions, and (3) a multi-scan lesion progression prediction module to assess changes in lesions over time using longitudinal bpMRI patches. The proposed framework was evaluated on a cohort from an AS clinical trial and demonstrated significant performance improvements over baseline models and radiologists, highlighting its potential to enhance clinical decision-making in AS management.

Primary Subject Area: Detection and Diagnosis

Secondary Subject Area: Application: Radiology

Paper Type: Both

Registration Requirement: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Created a single midl25_NNN.zip file with midl25_NNN.tex, midl25_NNN.bib, all necessary figures and files., Includes \documentclass{midl}, \jmlryear{2025}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package, Did not use the times package., All authors and co-authors are correctly listed with proper spelling and avoid Unicode characters., Author and institution details are de-anonymized where needed. All author names, affiliations, and paper title are correctly spelled and capitalized in the biography section., References must use the .bib file. Did not override the bibliographystyle defined in midl.cls. Did not use \begin{thebibliography} directly to insert references., Tables and figures do not overflow margins; avoid using \scalebox; used \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., Appendices and supplementary material are included in the same PDF after references., Main paper does not exceed 9 pages; acknowledgements, references, and appendix start on page 10 or later.

Submission Number: 200