Extraction of Weak Surface Diaphragmatic Electromyogram Using Modified Progressive FastICA Peel-Off.

Abstract

The diaphragmatic electromyogram (EMGdi) holds critical information about human respiration and can be employed to monitor respiratory conditions. Recording EMGdi noninvasively using surface electrodes placed on the chest is convenient. However, the extraction of the weak surface EMGdi (sEMGdi) from a noisy environment remains challenging, limiting its clinical application compared to esophageal EMGdi. This paper presents a novel, modified version of the progressive FastICA peel-off (PFP) framework for extracting weak sEMGdi signals. This framework employs an initial fast independent component analysis (FastICA) and subsequent constrained FastICA approach to extract and refine strong, repetitive electrocardiogram interference and respiration-related sEMGdi signals, respectively. Furthermore, the peel-off strategy ensures the progressive and complete extraction of weaker sEMGdi components. The proposed method was validated with both synthetic and clinical data. Experimental results demonstrate efficient extraction of clean sEMGdi signals with minimal distortion. Our method outperformed state-of-the-art methods in terms of signal-to-interference ratio (SIR) and correlation coefficient (CORR) at various noise levels in synthetic data tests. Additionally, it achieved an accuracy of 95.06% and an F2-score of 96.73% for breath identification in clinical data. This study provides a valuable solution for the noninvasive extraction of sEMGdi signals, offering a practical approach for ventilator synchrony with significant potential in supporting respiratory rehabilitation and health.