EXPLAINABLE MULTIMODAL AI IN CARDIOVASCULAR DISEASE PREDICTION
Received: 2026-06-21 12:49:34
Published: 2025-12-21
Abstract
Cardiovascular disease (CVD) is the leading global cause of mortality, and prediction is difficult in resource-limited settings. Traditional models like SCORE2 and Framingham use limited variables and offer low interpretability. We propose an Explainable Multimodal AI (EMAI) framework combining echocardiography images, ECG waveforms, and clinical data. Modality-specific encoders and a fusion transformer learn cross-modality interactions for improved prediction. Explainability is provided through SHAP for feature attribution, Grad-CAM for echo visualization, and temporal saliency for ECG interpretation. Preliminary results show higher accuracy than single-modality models. The framework enhances clinical trust and supports patient-specific cardiac risk assessment.
Keywords
List of references
About the Authors
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
How to Cite
