Related Experiment Video
Updated: Oct 26, 2025
04:04
Transauricular Vagus Nerve Stimulation and Electroencephalographic Assessment in Disorders of Consciousness
Published on: July 11, 2025
625
Preoperative Heart Rate Variability During Sleep Predicts Vagus Nerve Stimulation Outcome Better in Patients With
Xi Fang1, Hong-Yun Liu1,2, Zhi-Yan Wang1
1National Engineering Laboratory for Neuromodulation, School of Aerospace Engineering, Tsinghua University, Beijing, China.
Frontiers in Neurology
|July 26, 2021
Summary
Machine learning models using preoperative heart rate variability (HRV) during sleep can predict vagus nerve stimulation (VNS) success in drug-resistant epilepsy (DRE) patients. This approach helps identify suitable candidates, avoiding unnecessary surgery and associated risks.
Area of Science:
- Neurology
- Biomedical Engineering
- Data Science
Background:
- Vagus nerve stimulation (VNS) is a standard adjunctive therapy for drug-resistant epilepsy (DRE).
- Predicting VNS treatment outcomes remains challenging, leading to difficulties in patient selection.
- Preoperative assessment is crucial for optimizing VNS therapy efficacy.
Purpose of the Study:
- To develop a machine learning model for predicting VNS outcomes in DRE patients.
- To utilize multidimensional preoperative heart rate variability (HRV) indices for outcome prediction.
- To identify key HRV features that correlate with VNS treatment success.
Main Methods:
- Analysis of preoperative electrocardiography (ECG) data from 59 DRE patients and 50 healthy controls.
- Comparison of time, frequency, and non-linear HRV indices between VNS responders and non-responders during awake and sleep states.
- Application of machine learning algorithms, including Random Forest (RF) with Recursive Feature Elimination (RFE), for feature selection and model training.
- Evaluation of the prediction model using leave-one-out (LOO) cross-validation.
Main Results:
- Significant differences in 49 out of 52 HRV indices were observed between DRE patients and controls.
- More HRV indices showed significant differences between responders and non-responders during sleep (35) compared to awake states (16).
- The VNS outcome prediction model achieved 74.6% accuracy, 80% precision, 70.6% recall, and 75% F1 score using HRV indices from sleep states, outperforming awake state predictions.
Conclusions:
- Preoperative HRV analysis, particularly during sleep, combined with machine learning, can effectively predict VNS treatment outcomes in DRE patients.
- This predictive model can aid in selecting appropriate candidates for VNS surgery, potentially avoiding costly and risky procedures for non-responders.
- The study highlights the potential of non-invasive HRV measures for personalized epilepsy treatment strategies.
