The DITESMU project (Early Diagnosis of Depressive Symptoms through Statistical Multimodal Machine Learning) aims to enhance mental health diagnostics by leveraging artificial intelligence (AI) and multimodal biomarker analysis. Traditional clinical interviews and self-report measures are often insufficient due to high false-positive rates and subjectivity. DITESMU introduces a novel computational psychiatry framework that integrates physiological, behavioral, and linguistic data for more accurate and scalable depression diagnosis.

Project Objectives:
• Multimodal Data Processing for Depression Detection: The project develops a system that collects and processes multiple data sources, including eye-tracking, body movements, vocal features, and heart rate variability, to identify depressive biomarkers.
• Machine Learning for Biomarker Identification: By employing advanced machine learning techniques, such as deep learning (CNNs, LSTMs) and transformer-based models (BERT, Audio2Vec), DITESMU enhances the detection of depressive symptoms beyond traditional methods.
• Virtual Humans for Diagnostic Assessment: The project integrates AI-driven virtual humans that engage in real-time, voice-based interactions to elicit behavioral and physiological responses indicative of depression.
• Personalized and Scalable Mental Health Screening: Using multimodal AI, the system stratifies patients based on symptom severity, optimizing clinical decision-making and early intervention.

Impact:
• DITESMU represents a breakthrough in precision psychiatry by introducing an AI-driven, multimodal assessment tool that overcomes biases in current diagnostic methods. This scalable solution enhances mental health screening, enabling early and accurate detection of depression in primary care and hospital settings.