Abstract

Info

Title:: Prediction of pain thresholds under transcutaneous vagus nerve electrical stimulation based on machine learning and multimodal physiological signals

Author(s):: LIU Chunliang; ZHANG Yu; LIU Qi; WANG Jing; ZHUANG Lin; LIU Peirong; The Seventh Peoples Hospital Affiliated to Shanghai University of Traditional Chinese Medicine, Shanghai 200137, China

Keywords:: machine learning multimodal physiological signal transcutaneous vagus nerve electrical stimulation pain threshold

Abstract:: Objective To predict individual pain thresholds under transcutaneous vagus nerve stimulation (taVNS) using 3 modal physiological signals (electroencephalogram, electrodermal response, and heart rate variability) combined with machine learning algorithms, and further evaluate the predictive performance of different models. Methods The collected physiological data were processed with wavelet packet decomposition and feature extraction. Four models, namely support vector regression, long short-term memory network, gated recurrent unit, and bidirectional long short-term memory network (BiLSTM), were selected for experiments which were conducted using two different dataset partitioning methods. Predictive performance of each model was evaluated using normalized root mean square error, normalized mean absolute error, and coefficient of determination (R2). Results Machine learning models, especially BiLSTM, exhibited the superior predictive performance under both partitioning methods. BiLSTM achieved the highest R2, and the lowest normalized root mean square error and normalized mean absolute error, outperforming the other models. Conclusion Machine learning models integrated with multimodal physiological signals have great promise for pain threshold prediction, and notably, they demonstrate excellent performance when processing taVNS-induced complex physiological signals.