Machine Learning-Empowered Depression Detection With Wearable Skin Electronics: A Review

Depression, a widespread and debilitating psychiatric disorder, poses significant challenges to public health systems. Continuous real-time monitoring of depressive states is essential for facilitating timely interventions, optimizing treatment strategies, and improving patients’ quality of life. Traditional diagnostic methods are inherently subjective and limited, while objective assessments based on physiological signals provide an evidence- based foundation for accurate evaluation. The integration of wearable electronics and machine learning has spurred a growing trend toward real-time, long-term monitoring of depression within everyday environments. However, the recognition of depression through wearable technology encounters several challenges, including insufficient signal stability during long-term monitoring, difficulties in multimodal data acquisition, and limited algorithmic generalization across diverse individuals and deployment contexts. This review explores the physiological phenotypes associated with depression, evaluates machine learning algorithms for depression detection, and examines recent advances in wearable electronic devices. Finally, we highlight the ongoing challenges in signal stability, algorithmic generalization, and clinical validation, and outline future opportunities for interpretable modeling and privacy- preserving collaborative learning.