Deep fusion for multimodal subtype classification of major depressive disorders based on TSCANet

Major Depressive Disorder (MDD) represents a major global health burden, characterized by high recurrence rates and an increased risk of suicide. Traditional clinical diagnoses rely primarily on subjective scales, which limits their objectivity in accurately distinguishing MDD subtypes.To address these challenges, this study introduces a multimodal assessment framework that integrates EEG functional connectivity with clinical scale data.We developed TSCANet (Transformer Local-Global Graph Neural Network Scales Cross-Attention Network), a deep learning multimodal model designed to address the complexity of psychiatric data. TSCANet achieved classification accuracies of 82.62% and 81.36% in visual and auditory paradigms, respectively, outperforming single-modality models.