Using surface electromyography (SEMG) to classify low back pain based on lifting capacity evaluation with principal component analysis neural network method.

Abstract

Low back pain (LBP) is a leading cause of disability. The population with low back pain is continuously growing in the recent years. This study tries to distinguish LBP patients with healthy subjects by using the objective surface electromyography (SEMG) as a quantitative score for clinical evaluations. There are 26 healthy and 26 low back pain subjects who involved in this research. They lifted different weights by static and dynamic lifting process. Multiple features are extracted from the raw SEMG data, including energy and frequency indexes. Moreover, false discovery rate (FDR) omitted the false positive features. Then, a principal component analysis neural network (PCANN) was used for classifications. The results showed the features with different loadings (including 30%, and 50% loading) on lifting which can be used for distinguishing healthy and back pain subjects. By using PCANN method, more than 80% accuracies are achieved when different lifting weights were applied. Moreover, it is correlated between some EMG features and clinical scales, on exertion, fatigue, and pain. This technology can be potentially used for the future researches as a computer-aid diagnosis tool of LBP evaluation.