Detection of genioglossus myoelectric activity using ICA of multi-channel mandible sEMG

Abstract

BACKGROUND: Genioglossus myoelectric activity is of great significance in evaluating clinical respiratory function. However, there is a tradeoff in genioglossus EMG measurement with respect to accuracy versus convenience. OBJECTIVE: This paper presents a way to separate the characteristics of genioglossus myoelectric activity from multi-channel mandible sEMG through independent component analysis. METHODS: First, intra-oral genioglossus EMGgenioglossus EMG and three-channel mandible sEMG were recorded simultaneously. The FastICA algorithm was applied to three independent components from the sEMG signals. Then the independent components with the intra-oral genioglossus EMG were compared by calculating the Pearson correlation coefficient between them. RESULTS: An examination of 60 EMG samples showed that the FastICA algorithm was effective in separating the characteristics of genioglossus myoelectric activity from multi-channel mandible sEMG. The results of analysis were coincident with clinical diagnosis through intra-oral electrodes. CONCLUSIONS: Genioglossus myoelectric activity can be evaluated accurately by multi-channel mandible sEMG, which is non-invasive and easy to record.