A COMPREHENSIVE REVIEW OF RECENT ADVANCES IN AI-BASED BIOSIGNAL PROCESSING SYSTEMS FOR DEPRESSION DETECTION

Depression represents one of the most common neurological and mental health issues worldwide, imposing a significant burden on quality of life and contributing to high rates of morbidity and mortality. Traditional diagnosis methods are primarily based on clinical interviews and self-reporting, which can be subjective and time-consuming, thereby motivating the need for alternative approaches. Artificial Intelligence (AI) and intelligent biosignal processing represent a promising direction in depression detection, offering increased objectivity and precision. This study provides a comprehensive survey of recent literature on AI-based biosignal processing for depression detection, covering EEG, speech, physiological signals, facial expressions, textual data, and multimodal approaches. This study discusses the current state-of-the-art in machine learning and deep learning algorithms, including Convolutional Neural Networks (CNN), Long Short-Time Memory (LSTM), Transformers, Graph Neural Networks (GNN), Attention Mechanisms, and Explainable AI (XAI) for health. The results demonstrate that deep learning algorithms are generally more effective than traditional machine learning methods, achieving classification accuracy in some cases exceeding 95%. Multimodal approaches were also found to be more accurate and robust than unimodal ones. The study highlights the key limitations of the current research, including the lack of standardized data sets and external validation, as well as the need for increased model interpretability and clinical applicability. This study concludes that future directions in intelligent depression detection include multimodal learning, XAI, wearables, and large-scale clinical investigations. The results of this study contribute to the growing body of literature on AI-based biosignal processing for depression detection and can help clinicians and researchers identify viable options for developing novel biosignal processing algorithms.