Novel approach on estimation of deep muscle activation level

Abstract

Non-specific low back pain (LBP) is one of the world’s major neuromuscular diseases. Recent studies have shown that deep muscle activation abnormalities are related to non-specific low back pain (LBP). In particular, transverse abdominis activation abnormalities are often associated with patients that suffer from LBP. Hence, deep muscle activation assessment has a potential clinical used in non-specific LBP. Traditionally, needle electromyography (EMG) has been used for deep muscle activation assessment. However, it is invasive and only local muscle activities can be assessed. In contrast, surface EMG is a non-invasive assessment method of muscle activities. It can also reflects the global characteristics of motor unit activities. Hence, there is a strong urge to explore the potential uses of surface EMG in assessing deep muscle activity. In this study, the main objective is to evaluate the potentials of surface EMG in accessing deep muscles activities and proposed a systematic approach to estimate deep muscles activation level.

In order to investigate how deep muscle EMG would be recorded on the surface, a new structure-based surface EMG model with capability to model fiber alignment direction is built. It allows us to investigate the surface EMG characteristics of deep muscles with different fiber alignment directions. Moreover, surface EMG data are collected using an electrode array from three normal subjects (age: 25± 2.65) during the performance of three motion groups—trunk flexion, trunk rotation and compress abdominal contents—which involved deep muscles transverse abdominis and internal oblique. Through comparing the results from the computation model and the experimental setup, useful characteristics in surface EMG that capable to infer deep muscle activities can be isolated.

Using the new surface EMG model, we found that surface EMG array data can estimated the depth of a single muscle fiber by the trace of eigenvalues derived from the surface EMG array data. Furthermore, eigenvector matrix U derived from surface EMG array data can separate different motions that involved deep muscles in both the computation model and experimental setup. Based on this findings, a systematic approach to estimated deep muscle activities is proposed in this study. Using the computation model to evaluate the algorithm, about 70% of extra information is gained about the deep muscle EMG source. In the experimental setup, it is verified (96% ± 3%) against the traditional binary muscle-motion table.

To conclude, surface EMG array data has the ability to estimate deep muscle activities. In particular, both eigenvector matrix and eigenvalue derived from surface EMG exhibit high potential in deep muscle EMG extraction. The method proposed in this study have the potential to be used clinically for neuromuscular disease assessment.