Research on Parkinson's disease diagnosis method based on multimodal information fusion
Year: 2026
Type: article
Abstract: Deep learning-based computer-aided diagnosis has attracted significant attention as an important tool to enhance the accuracy of Parkinson's Disease (PD) screening and outpatient diagnosis.To address the limitations of single-modal approaches in traditional research, including poor classification performance, the limited generalizability of public datasets, and small sample sizes, the authors propose a novel multimodal information fusion method for PD diagnosis.The proposed method integrates multimodal data from heterogeneous sources, including facial expressions, speech, and individual attributes, collected from patients during natural communication, for disease prediction.Specifically, this method uses a pre-trained multimodal model to extract robust facial expression and spectrogram features and enhances multimodal feature representation through a carefully designed DenseConv module.Additionally, the paper incorporates individual attribute information (gender and age) into the multimodal representation learning framework.Furthermore, the authors propose a novel cross-modal progressive hierarchical fusion framework, which employs a self-attention mechanism to efficiently integrate facial expressions, spectrogram features, and individual attributes.This approach alleviates modality discrepancies among heterogeneous features and significantly improves the generalization and robustness of the model.Experiments conducted on a real-world clinical PD dataset demonstrate the superior performance of the proposed method, achieving an accuracy of 84.4% and an F1-score of 84.0.These results provide strong support for large-scale, generalizable, and convenient screening of PD.
Language: Chinese
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7162702856
