Heart Rate Variability-Derived Features Based on Deep Neural Networks for Monitoring Depth of Anaesthesia

Abstract

Background: Estimating the depth of anaesthesia (DoA) is critical in clinical anaesthesiology. Electroencephalograms (EEGs) have been widely used for monitoring the DoA; however, they may be inaccurate under certain conditions. Methods: In this study, we propose a novel method to evaluate the DoA based on multiple heart rate variability (HRV)-derived features combined with a discrete wavelet transform and deep neural networks (DNNs). Four features were extracted from an electrocardiogram, including the HRV high-frequency power, low-frequency power, high-to-low-frequency power ratio, and sample entropy. Next, these features were used as inputs for the DNN, which used the expert assessment of consciousness level as the reference output. Finally, the DNN was compared with the logistic regression (LR), support vector machine (SVM), and decision tree (DT) models. The data of 23 anaesthesia patients were used to assess the proposed method. Results: The results demonstrated that the accuracies of the four models, in distinguishing the anaesthesia states, were 86.2% (LR),87.5% (SVM),87.2% (DT), and 90.1%(DNN). Our method outperformed the LR, SVM, and DT methods.Conclusions: The proposed method could accurately distinguish between different anaesthesia states, thus, providing an alternative or supplementary method to EEG monitoring for the assessment of DoA.