Key result
ECG-derived HRV with Random Forest identifies stress states with an ROC AUC of ~0.95.
Why the study?
The generalizability of HRV-based stress classifiers across individuals has been limited by subject leakage in standard train-test splits, requiring leave-one-subject-out cross-validation to establish unbiased performance estimates.
Does leave-one-subject-out cross-validation of ECG-derived HRV machine learning classifiers provide accurate, unbiased subject-independent stress-state discrimination compared to conventional train-test splits?
Observational (n=16)
No
Does leave-one-subject-out cross-validation of ECG-derived HRV machine learning classifiers provide accurate, unbiased subject-independent stress-state discrimination compared to conventional train-test splits?
ECG-derived HRV features analyzed with classical machine learning models achieve robust, subject-independent stress classification under leave-one-subject-out validation, establishing a methodological proof-of-concept for autonomic monitoring in alcohol use disorder.
Answers are grounded in this paper and cite their sources.
May support HRV-ML stress classifiers in research; leaves open clinical adoption pending larger validation.
Heart rate variability (HRV) is a validated non-invasive marker of autonomic dysregulation associated with stress reactivity and relapse risk in alcohol use disorder (AUD). However, the generalizability of HRV-based stress classifiers across individuals has been limited by subject leakage in standard train-test splits. We evaluated HRV-based stress-state classification using leave-one-subject-out (LOSO) cross-validation to establish unbiased performance estimates. Using the WESAD public dataset (15 subjects, chest-worn ECG), we extracted time-domain (RMSSD, SDNN, pNN50), frequency-domain (LF, HF, LF/HF), and nonlinear HRV features. Under LOSO validation, Random Forest achieved 86.7% accuracy with ROC AUC of 0.946 (95% CI: 0.871-0.998) for binary stress classification — substantially higher than the 0.848 obtained from a conventional train-test split with subject leakage. Logistic Regression achieved AUC 0.873 and SVM achieved AUC 0.893. Deep sequence models (LSTM, CNN) did not improve over feature-based approaches, likely
No takes yet. Share an insight, caveat, or question.
Chhaya et al. (2026) conducted an observational in Stress reactivity / Alcohol Use Disorder relapse risk proxy (n=16). ECG-derived heart rate variability (HRV) Random Forest classifier with leave-one-subject-out (LOSO) cross-validation vs. Conventional 70/30 stratified train-test split, Logistic Regression, SVM, LSTM, and CNN models was evaluated on Binary stress-state classification performance (Accuracy and ROC AUC) (AUC 0.946, 95% CI 0.871-0.998). ECG-derived HRV with Random Forest under leave-one-subject-out validation achieved 86.7% accuracy and an ROC AUC of 0.946 (95% CI: 0.871-0.998) for binary stress-state classification.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: