TRACE-PCa: Predicting Prostate Cancer Progression from Longitudinal MRI During Active Surveillance

2026-07-15

AI Summary

Current active surveillance strategies rely heavily on repeated biopsies, most of which are unnecessary due to lack of disease progression. Existing risk stratification tools often depend on single-timepoint imaging or require explicit lesion segmentation, limiting their ability to model longitudinal changes and excluding patients without MRI-visible lesions. This work proposes the first end-to-end temporal multimodal model that operates without lesion segmentation, leveraging a pretrained 3D MRI foundation model to encode serial scans. A temporal attention gating mechanism highlights focal imaging changes associated with progression, which are integrated with clinical variables to predict pathological progression. Evaluated on a longitudinal cohort, the method significantly outperforms baseline approaches, achieving a positive predictive value surpassing radiologist assessments while maintaining high negative predictive value—suggesting its potential to safely reduce unnecessary biopsies.

Abstract

Active surveillance (AS) is the preferred strategy for favorable-risk prostate cancer, yet current protocols rely on scheduled repeat biopsies, most of which reveal no progression and are unnecessary. Existing risk-stratification tools operate on single time-point imaging or depend on explicit lesion segmentation, limiting their ability to capture longitudinal change and excluding patients without an MRI-visible lesion. In this study, we propose an end-to-end temporal and multimodal model for predicting pathological progression during AS without lesion segmentation. We encode each serial scan with a pretrained 3D MRI foundation model and introduce a temporal attention gate that recalibrates the multi-visit features to amplify focal imaging changes associated with progression. The gated imaging representation is then fused with clinical variables in a multimodal framework to estimate the probability of progression. Validated on a longitudinal AS cohort, our approach consistently outperforms competing baselines and performs comparably to the radiologist assessment representing current clinical practice. It maintains high negative predictive value while achieving higher positive predictive value, demonstrating its potential to safely reduce unnecessary biopsies during surveillance.

Problem

• prostate cancer
• active surveillance
• longitudinal MRI
• disease progression
• risk stratification

Innovation

• temporal attention
• foundation model
• multimodal fusion
• lesion-free prediction
• longitudinal MRI