Multi-Model AI Approaches for heart failure risk prediction based on fusion of Echocardiographic Images and clinical data

Abstract

Heart failure (HF) is a chronic illness that has a high global incidence and mortality rate. Its direction is complex and incredibly varied. Traditional prediction methods are usually based on a single information source, such as organized medical records or echocardiographic images, and are thus not able to completely reflect the medical condition of patients. The use of artificial intelligence (AI) in cardiovascular medicine has been steadily growing in recent years. Specifically, the multi-model connection and multi-modal data fusion methods held great potential in integrating diverse information sources and enhancing prediction accuracy. It is a review of modeling, fusion, and implementation approaches to multi-model diagnosis from clinical and image data. It summarizes recent advances in the prediction of heart failure risk from single- and multi-modal AI, presents challenges such as interpretability, clinical acceptability, data quality, and generalizability, and envisions future directions such as time series modeling, cross-modal deep learning, and multi-center data sharing.  The significance of this review lies in serving future research and further promoting the clinical applications and development of AI technology in predicting heart failure.