PROFUSEme: PROstate Cancer Biochemical Recurrence Prediction via FUSEd Multi-modal Embeddings

2025-09-17

AI Summary

This study addresses the challenge of early and accurate prediction of biochemical recurrence (BCR) following radical prostatectomy. We propose a mid-fusion multimodal embedding framework that jointly integrates heterogeneous clinical, imaging, and pathological data. By leveraging deep representation learning, the framework constructs a unified cross-modal embedding space and seamlessly couples it with a Cox proportional hazards model to enable end-to-end survival analysis. Unlike late-fusion baselines, our approach explicitly models inter-modal interactions, enhancing prognostic discriminability. In internal five-fold cross-validation, the model achieves a concordance index (C-index) of 0.861; on the held-out test set of the CHIMERA 2025 Challenge, it attains a C-index of 0.7103—marking a substantial improvement over prior methods. The framework delivers an interpretable, robust, and clinically actionable tool for personalized postoperative management.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Almost 30% of prostate cancer (PCa) patients undergoing radical prostatectomy (RP) experience biochemical recurrence (BCR), characterized by increased prostate specific antigen (PSA) and associated with increased mortality. Accurate early prediction of BCR, at the time of RP, would contribute to prompt adaptive clinical decision-making and improved patient outcomes. In this work, we propose prostate cancer BCR prediction via fused multi-modal embeddings (PROFUSEme), which learns cross-modal interactions of clinical, radiology, and pathology data, following an intermediate fusion configuration in combination with Cox Proportional Hazard regressors. Quantitative evaluation of our proposed approach reveals superior performance, when compared with late fusion configurations, yielding a mean C-index of 0.861 ($σ=0.112$) on the internal 5-fold nested cross-validation framework, and a C-index of 0.7103 on the hold out data of CHIMERA 2025 challenge validation leaderboard.

Problem

• Predicting prostate cancer recurrence after surgery
• Fusing multi-modal clinical imaging and data
• Improving early biochemical recurrence risk assessment

Innovation

• Fuses clinical, radiology, pathology data
• Uses intermediate fusion configuration
• Combines with Cox Proportional Hazard regressors