Comprehensive Summary
This study introduces a novel anxiety screening framework, called the Anxiety Screening Framework integrating Multimodal Data and Graph Node Correlation, which combines physiological, behavioral, audio, and textual data using a graph convolutional network (GCN). The goal is to create an objective, cost-effective, and scalable screening tool to detect anxiety disorders more accurately than traditional methods such as questionnaires or interviews. The model integrates one-dimensional convolutional neural networks (1D-CNNs), gated recurrent units, and CNN-based text analysis to extract meaningful features from various data types. These features are then structured as graph nodes to model complex relationships among them. Tested on data from 217 seafarers, the framework achieved 93.48% accuracy, 94.58% AUC, 90.00% precision, 81.82% sensitivity, and 97.14% specificity, outperforming other deep learning and machine learning methods. Even when questionnaire data was unavailable, the system maintained strong performance, demonstrating robustness and adaptability.
Outcomes and Implications
This research highlights how multimodal AI models can revolutionize anxiety detection by integrating diverse physiological and behavioral signals into one cohesive framework. By moving beyond self-reported measures, ASF-MDGNC offers clinicians a more objective and data-driven tool for early anxiety screening and intervention. The system’s non-contact and multimodal design could make it particularly useful in environments where in-person evaluation is difficult or resource-limited. Clinically, such tools could enhance mental health monitoring and support personalized treatment plans. The authors note that broader validation is still needed, particularly across varied populations and real-world clinical settings, but their results suggest strong potential for future clinical implementation. As AI models like this evolve, they could form the basis for continuous, real-time mental health screening integrated into wearable or telehealth systems.