Smart Devices and Multimodal Systems for Mental Health Monitoring: From Theory to Application.

Area of Science:

• Digital health and biomedical engineering.
• Artificial intelligence in healthcare.
• Mental health technology and biosignal processing.

Background:

• Wearable and multimodal biosignal systems, including electroencephalography (EEG), electrocardiography (ECG), and electromyography (EMG), are increasingly integrated with artificial intelligence (AI).
• These technologies aim to enhance the assessment and longitudinal monitoring of mental health conditions.
• Current evidence is heterogeneous, with limited clinical translation due to variability in protocols, analysis, and validation.

Purpose of the Study:

• To systematically review current applications, signal-processing techniques, and methodological limitations of AI-powered smart systems for mental health monitoring.
• To synthesize findings on the use of biosignals for detecting, monitoring, and managing mental health outcomes.
• To identify challenges and future directions for clinical translation.

Main Methods:

• A PRISMA 2020-guided systematic review of studies published between 2013 and 2026.
• Searched major scientific databases (PubMed/MEDLINE, Scopus, Web of Science, IEEE Xplore, ACM Digital Library).
• Included studies on human applications of wearable/smart devices or multimodal biosignals for mental health, analyzing data using machine learning/deep learning models.

Main Results:

• Literature clustered into depression (37%), stress/anxiety (18%), PTSD/trauma (5%), technological innovations (25%), brain-state stimulation (3%), and socioeconomic context (7%).
• Common analytical pipelines involved artifact suppression, feature extraction, and machine/deep learning models.
• Significant limitations include small sample sizes (<100 in 67% of studies), limited ecological validity, lack of external validation, and heterogeneity in protocols and outcomes.

Conclusions:

• Multimodal biosignal systems, especially wearable cardiac metrics and passive sensing, hold significant potential for augmenting mental health assessment.
• Current evidence is largely based on proof-of-concept studies.
• Future research must focus on standardized reporting, rigorous validation in diverse cohorts, transparent model evaluation, and ethical considerations (privacy, fairness, governance) to facilitate clinical practice.