Pain-Level Detection Using Heart Rate Variability

Pain is one of the most felt and complex sensation of human body. Almost 1/3rd of world population suffers from chronic pain. Most of the pain assessment techniques depend on verbal responses from the subjects. Therefore, patients who are unable to communicate in conventional way cannot participate in these methods and express their severity of injury. As pain has an emotional component, facial video signal-based pain estimation methodologies are reported by many researchers. However, video signal-based analysis requires costly equipment and follows complex algorithm. Instead of video signals, pain-induced facial expression analysis through biopotentials captured from certain facial muscles are also reported with lesser accuracy in lower pain levels. Considering the pain stimulated activation of autonomic nervous system, in this study, alteration of sympathetic activity of nervous system due to pain was analyzed using heart rate. Time Absolute Integral, Mean, Variance and Sectional Slopes were extracted from heart rate to recognize pain levels using various machine learning approaches such as SVM, k-NN and Logistic regression. Further, detection ability of no pain and other pain levels was investigated for each classifier individually by various parameters. Moreover, the proposed method shows satisfactory response during lower pain-level detection.