Chronic pain classification using PPG and ECG parameters selected via hybrid feature selection

Current chronic pain intensity assessment methods are large based on self-description, which is the gold-standard in clinical practice. However, for patients who lack communication ability or consciousness, this challenging task falls on medical staff. Our study aimed to develop a pain intensity classification system to perform this task automatically. Electrocardiogram and photoplethysmography (PPG) recordings from 43 patients with pain were analyzed with recordings from 20 healthy volunteers as a control group. A numerical rating scale was derived to assess pain intensity, which was divided into four categories of no pain, low pain, moderate pain, and severe pain as specific labels. To explore the robustness of PPG and heart rate variability (HRV) features, a hybrid feature selection (HFS) method was applied to identify important indicators based on majority rule. Additionally, a multi-class support vector machine (SVM) was utilized for classification with 10-fold cross validation. The overall estimation accuracy of the SVM for the four pain intensity levels was approximately 74.6%. It was determined that PPG and HRV features with HFS can provide sufficient information to discriminate different pain intensities in myofascial chronic pain patients. Additionally, our study yielded a novel perspective regarding the trends of PPG features in clinical practice.