MACHINE LEARNING MODEL FOR DEPRESSION SCREENING IN PATIENTS WITH NASOPHARYNGEAL CARCINOMA

Objective: To develop and internally validate a one-dimensional convolutional neural network (1D-CNN) model for depression screening among newly diagnosed patients with nasopharyngeal carcinoma. Methods: This prediction-model development study was conducted at Hanoi Medical University Hospital from January 2025 to March 2026 among 812 newly diagnosed, treatment-naïve patients with nasopharyngeal carcinoma. Depression risk was screened using the Patient Health Questionnaire-9 with a cutoff score of ≥ 10. The dataset was divided into training, validation and test sets. Model performance was assessed using AUC, sensitivity, specificity, positive predictive value, negative predictive value, F1-score and Brier score. Results: The prevalence of depression was 25.1% (204/812). The 1D-CNN model achieved an AUC of 0.902, sensitivity of 84.4%, specificity of 81.3%, positive predictive value of 60.1%, negative predictive value of 93.8%, F1-score of 0.701 and Brier score of 0.124. Conclusion: The 1D-CNN model may support early screening of patients at high risk of depression in nasopharyngeal carcinoma, although external validation is required before routine clinical implementation.