Proceedings of the Annual Meeting of the Cognitive Science Society

The rising prevalence of adolescent depression is a major public health concern. Current diagnostic methods, often burdensome and lacking objective biomarkers, hinder early detection. This paper proposes a novel multi-task framework integrating attention and resting EEG signals using Biosignal Learning and Agent Transformer (BLAT) for depression detection. EEG data is segmented, channeled, and positionally encoded, followed by feature extraction and classification via Agent Transformer. Experiments with 50 depressed adolescents and 50 controls achieved 85% accuracy using BLAT, offering a promising method for early adolescent depression screening.