HRV-based Classification of Psychiatric Disorder Using Wearable ECG

Objective assessment of psychiatric conditions using wearable sensing is becoming increasingly important for intelligent healthcare systems. This study examined whether heart rate variability (HRV) features derived from electrocardiogram (ECG) recordings measured during daily activities can distinguish individuals with psychiatric disorders from healthy controls and whether the identified key HRV feature is associated with symptom severity. Thirty patients with schizophrenia or bipolar disorder and 30 age-and sex-matched healthy controls wore a wearable ECG sensor for approximately 1.5–2 hours during habitual daily activities. Twenty-one HRV features including time-, frequency- and nonlinear-domain metrics were extracted from the ECG recordings. A random forest classifier was trained and evaluated using stratified five-fold cross-validation, achieving an overall accuracy of 80.0%. Permutation-based feature importance highlighted the HRV triangular index as a major contributor to classification performance, and a lower triangular index was significantly associated with higher PANSS general psychopathology scores (r = −0.44, Bonferroni-corrected p = 0.045). These findings demonstrate the feasibility of interpretable HRV-based classification in daily-life settings and suggest that reduced HRV may reflect clinically meaningful symptom burden in psychiatric disorders.