Early detection of vasovagal syncope using artificial intelligence–driven analysis of smartwatch-derived photoplethysmographic heart rate variability

Background Vasovagal syncope (VVS) is a common cause of transient loss of consciousness and may lead to injury and impaired quality of life.1 Timely identification of impending events could enable preventive strategies.2 However, reliable short-term prediction remains challenging. Purpose To determine whether heart rate variability (HRV) derived from smartwatch-based photoplethysmography (PPG) can predict VVS during head-up tilt (HUT) testing, and to identify the optimal observation window and prediction lead time. Methods We prospectively enrolled 132 patients with suspected neurally mediated syncope who underwent standardized HUT testing while wearing a Samsung Galaxy Watch 6 for continuous multiwavelength PPG acquisition (25 Hz). A total of 107 HRV-related features were extracted from PPG signals. An Extra Trees ensemble classifier (600 trees) was trained using an 80/20 subject-level split and evaluated on a hold-out test set. Model performance was assessed using the area under the receiver operating characteristic curve (AUROC). Threshold-based metrics were calculated at a fixed sensitivity of 0.90. Results Sixty-three participants were HUT-positive and 69 were HUT-negative. A 5-minute presyncope observation window yielded the highest discrimination (AUROC 0.91; 95% CI 0.77–1.00). At 90% sensitivity, specificity was 0.64 (95% CI 0.40–0.85). Using a fixed 5-minute observation window, prediction remained robust with a 5-minute lead time before syncope (AUROC 0.91; 95% CI 0.76–1.00; accuracy 84.6%). (Figure 1) Key predictive features included nonlinear complexity measures (approximate entropy, composite multiscale entropy) and indices of autonomic modulation (normalized low-frequency power, log-transformed high-frequency power, cardiac vagal index). Conclusion Artificial intelligence–based analysis of smartwatch-derived PPG HRV enables prospective prediction of VVS during HUT using a short observation window and maintains clinically meaningful performance several minutes prior to event onset. These findings support the feasibility of wearable, real-time early warning systems for syncope.