מושגי ליבה
The author proposes a multimodal approach for Parkinson's Disease classification using contrastive graph cross-view learning, achieving high accuracy and AUC in five-fold cross-validation.
תקציר
The study introduces a novel approach integrating image and clinical features for PD classification. By leveraging contrastive cross-view graph fusion, the model achieves 91% accuracy and 92.8% AUC. The research highlights the importance of incorporating both image and non-image data for improved predictive capabilities. The proposed method outperforms traditional machine learning-based approaches by utilizing a multimodal co-attention module and contrastive loss-based fusion method. By combining graph representation learning with deep neural networks, the model extracts robust features for enhanced multi-view data analysis.
סטטיסטיקה
Our graph-view multimodal approach achieves an accuracy of 91%.
The model demonstrates an AUC of 92.8% in five-fold cross-validation.