Diagnosis and Classification of Depressive Disorders using ML and DL Models

The diagnosis and classification of depressive disorders pose significant challenges in mental healthcare, mainly due to overlapping symptoms, subjective evaluations, and variations in patient presentations. Traditional diagnostic approaches often lack objectivity and fail to capture the complex nature of depression across diverse populations. This study introduces a comprehensive framework that leverages advanced Machine Learning (ML) and Deep Learning (DL) models to improve the accuracy and reliability of diagnosing depressive disorders. Using the SAMM (Spontaneous Micro-Facial Movement) dataset, comprising 11,800 high-resolution facial images capturing spontaneous facial expressions, the proposed framework integrates dual embedding methods (GloVE and BERT) with hierarchical attention mechanisms for feature extraction. Parallel processing streams of LSTM and CNN architectures allow the recognition of intricate patterns across multimodal data. Experimental results showed superior performance across key metrics, achieving an accuracy of 94%, precision of 92%, recall of 93%, F1-score of 92.5%, and an AUC-ROC of 0.96. The proposed framework provides an efficient, interpretable, and scalable solution to advance mental health diagnostics, addressing the urgent need for objective and standardized tools in psychiatric care.