A novel adaptive weighting approach for multimodal biomedical data fusion in prediction model

Abstract

Current coronary artery disease (CAD) prediction models predominantly rely on single-modal data, lack consideration of genetic factors, and suffer from insufficient model interpretability. To address these limitations, this study proposes an innovative multimodal data fusion and adaptive weighting network. This framework integrates clinical features, polygenic risk scores, and medical imaging data. By introducing an attention mechanism, it dynamically calculates the contribution weights of different data sources and employs multi-task learning to simultaneously predict CAD incidence risk and major adverse cardiovascular events. Experiments conducted on dataset that the proposed model outperforms traditional models in key metrics. This research provides a more accurate and interpretable new paradigm for CAD risk prediction.

Ähnliche Arbeiten