AI-enabled Neurodiagnostic Framework for Migrant Mental Health: Integrating Ethical Governance in Stress, Depression, and Anxiety Assessment
Rufina Hussain, Safdar Tanweer, Sherin ZafarIntroduction/Objective:
Migrant workers in Gulf countries face elevated risks of depression, anxiety, and stress, while access to culturally sensitive screening remains limited. This study aimed to develop an interpretable AI-enabled framework for identifying probable depressive symptoms among Gulf migrants and to demonstrate a proof-of-concept neuroimaging extension.
Methods:
A cross-sectional secondary data analysis was conducted using an anonymized public DASS-21 dataset filtered to immigrants residing in Saudi Arabia, the United Arab Emirates, Qatar, and Oman. After applying the eligibility criteria, 124 respondents were analyzed using DASS-21 items, demographic variables, TIPI personality scores, and VCL indicators. A Random Forest classifier was evaluated using stratified validation, ROC-AUC, confusion matrices, and feature- importance analysis. A separate CNN-based MRI pipeline was included only as a technical proof of concept.
Results:
The Random Forest model achieved 86.8% accuracy, 0.94 ROC-AUC, 96.6% sensitivity, 55.6% specificity, 87.5% precision, and 0.92 F1-score. Higher depressive-symptom patterns were observed among respondents from the UAE and Oman, younger participants, and females.
Discussion:
The findings indicate that explainable machine learning can support early screening in underserved migrant populations.
Conclusion:
The framework offers a scalable screening-support tool, but requires larger, balanced, longitudinal validation.