An anxiety screening framework integrating multimodal data and graph node correlation.

Area of Science:

• Psychiatry and Computational Neuroscience
• Machine Learning Applications in Healthcare

Background:

• Anxiety disorders represent a major global health challenge with significant personal and economic impacts.
• Current anxiety screening tools are limited by cost, cognitive biases, and an inability to model complex feature correlations.

Purpose of the Study:

• To develop an advanced anxiety screening framework utilizing multimodal data and Graph Convolutional Networks (GCN).
• To overcome the limitations of existing methods by capturing intricate spatiotemporal relationships and feature correlations.

Main Methods:

• Integration of multimodal data including physiological, behavioral, audio, and textual information.
• Application of a Graph Convolutional Network (GCN) framework, incorporating one-dimensional convolutional neural networks and gated recurrent units.
• Focus on modeling complex spatiotemporal relationships and correlations among diverse data features.

Main Results:

• Achieved high performance metrics: 93.48% accuracy, 94.58% AUC, 90.00% precision, 81.82% sensitivity, 97.14% specificity, and 85.71% F1 score.
• Demonstrated effectiveness even without questionnaire data, highlighting practical utility.
• The GCN framework successfully captured complex correlations within multimodal data.

Conclusions:

• The proposed GCN-based multimodal framework offers a novel and effective approach for anxiety screening.
• This method provides a strong scientific basis for early anxiety symptom identification, personalized treatment, and prevention.
• The framework's ability to analyze complex data structures enhances diagnostic accuracy and clinical applicability.