Prognostic prediction of electroacupuncture for low back pain based on surface electromyographic map and CNN-SE model

Chronic low back pain (CLBP) is a common and refractory musculoskeletal disease, which seriously affects patients' work and life. Electroacupuncture (EA) has become a popular method for the treatment of CLBP due to its advantages of high safety and few side effects, but its effect is still controversial. One of the important reasons is that individual differences make EA therapy not effective for everyone. Through machine learning method, the therapeutic effect prediction model was constructed from the surface electromyography (sEMG) data shows potential to help screen patients suitable for specific treatments efficiency. Thus, this study aims to establish a prediction model for EA treatment of CLBP by using convolutional neural network (CNN) based on sEMG topographic map. Each map was categorized with a label according to the changes of pain indicators. The results show that the prediction accuracy of CNN-SE (Squeeze-and-Excitation) model proposed in this study is 75%, which proves the potential of CNN-SE model to predict the efficacy of EA. This study initially confirmed the effectiveness of the EA effect prediction model for CLBP based on multi-channel sEMG signal, which is also of great significance for further research on the pathogenesis and treatment optimization of CLBP.