HRV-derived data similarity and distribution index based on ensemble neural network for measuring depth of anaesthesia

PeerJ. 2017 Nov 16;5:e4067. doi: 10.7717/peerj.4067. eCollection 2017.

Abstract

Evaluation of depth of anaesthesia (DoA) is critical in clinical surgery. Indices derived from electroencephalogram (EEG) are currently widely used to quantify DoA. However, there are known to be inaccurate under certain conditions; therefore, experienced anaesthesiologists rely on the monitoring of vital signs such as body temperature, pulse rate, respiration rate, and blood pressure to control the procedure. Because of the lack of an ideal approach for quantifying level of consciousness, studies have been conducted to develop improved methods of measuring DoA. In this study, a short-term index known as the similarity and distribution index (SDI) is proposed. The SDI is generated using heart rate variability (HRV) in the time domain and is based on observations of data distribution differences between two consecutive 32 s HRV data segments. A comparison between SDI results and expert assessments of consciousness level revealed that the SDI has strong correlation with anaesthetic depth. To optimise the effect, artificial neural network (ANN) models were constructed to fit the SDI, and ANN blind cross-validation was conducted to overcome random errors and overfitting problems. An ensemble ANN was then employed and was discovered to provide favourable DoA assessment in comparison with commonly used Bispectral Index. This study demonstrated the effectiveness of this method of DoA assessment, and the results imply that it is feasible and meaningful to use the SDI to measure DoA with the additional use of other measurement methods, if appropriate.

Keywords: Artificial neural network; Depth of anesthesia; Expert assessment of consciousness level; Heart rate variability; Similarity and distribution index.

Associated data

  • figshare/10.6084/m9.figshare.5254426.v1

Grant support

This research is supported by the Center for Dynamical Biomarkers and Translational Medicine, National Central University, Taiwan, which is sponsored by National Science Council (Grant Number: NSC102-2911-I-008-001). This research is also supported by Innovation Center for Big Data and Digital Convergence, Yuan Ze University, Taiwan. Additional funding comes from Wuhan University of Technology international exchange program (Grant Number: 2015-JL-012) and National Natural Science Foundation of China (Grant Number: 51475342, 51675389). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.