PRISM: A Framework Harnessing Unsupervised Visual Representations and Textual Prompts for Explainable MACE Survival Prediction from Cardiac Cine MRI

2025-08-26

AI Summary

Accurately predicting major adverse cardiac events (MACE) remains a critical challenge in cardiovascular prognostic modeling. This paper introduces a prompt-guided multimodal representation ensemble framework that, for the first time, jointly leverages unsupervised visual representations from non-contrast cine MRI and structured electronic health records for end-to-end survival analysis. Key methodological innovations include motion-aware multi-view self-supervised distillation, text-prompt-modulated imaging–clinical feature alignment, and an interpretable risk attribution mechanism that identifies three imaging biomarkers significantly associated with elevated MACE risk. Evaluated across four independent clinical cohorts—including internal and external validation—the model substantially outperforms conventional statistical models and state-of-the-art deep learning approaches (C-index improvement of 3.2–5.8%, *p* < 0.001), achieving both superior predictive performance and clinically meaningful interpretability.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Accurate prediction of major adverse cardiac events (MACE) remains a central challenge in cardiovascular prognosis. We present PRISM (Prompt-guided Representation Integration for Survival Modeling), a self-supervised framework that integrates visual representations from non-contrast cardiac cine magnetic resonance imaging with structured electronic health records (EHRs) for survival analysis. PRISM extracts temporally synchronized imaging features through motion-aware multi-view distillation and modulates them using medically informed textual prompts to enable fine-grained risk prediction. Across four independent clinical cohorts, PRISM consistently surpasses classical survival prediction models and state-of-the-art (SOTA) deep learning baselines under internal and external validation. Further clinical findings demonstrate that the combined imaging and EHR representations derived from PRISM provide valuable insights into cardiac risk across diverse cohorts. Three distinct imaging signatures associated with elevated MACE risk are uncovered, including lateral wall dyssynchrony, inferior wall hypersensitivity, and anterior elevated focus during diastole. Prompt-guided attribution further identifies hypertension, diabetes, and smoking as dominant contributors among clinical and physiological EHR factors.

Problem

• Predicting major adverse cardiac events from cardiac MRI
• Integrating visual representations with electronic health records
• Enabling fine-grained risk prediction using medical prompts

Innovation

• Unsupervised visual representations integration
• Medical textual prompts modulation
• Motion-aware multi-view distillation