Multimodal data fusion for early risk prediction of disease

This study utilizes the PEARL-Neuro dataset, which contains multimodal data from healthy middle-aged individuals. It aims to explore the integration of electroencephalogram (EEG) and magnetic resonance imaging (MRI) features, with apolipoprotein E (APOE) genotype information serving as risk labels for prediction. Both feature-level and model-level fusion strategies are adopted to leverage the complementary information across modalities and enhance classification performance. The results indicate that multimodal fusion significantly improves predictive performance compared to unimodal models. The fused model offers a more reliable solution for the early risk assessment of disease. Despite challenges such as the scarcity of early-stage case samples and data imbalance, this study still highlights the potential of multimodal data fusion in enhancing the risk prediction ability for disease and provides strong support for the formulation of personalized treatment strategies. Future research will focus on optimizing feature extraction methods, improving models, and exploring innovative fusion techniques to further increase the accuracy of risk prediction for disease.