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

Area of Science:

• Biomedical Engineering
• Respiratory Physiology
• Signal Processing

Background:

• Diaphragmatic electromyogram (EMGdi) is crucial for understanding human respiration and monitoring respiratory conditions.
• Noninvasive surface EMGdi (sEMGdi) recording is convenient but faces challenges in extracting weak signals from noise.
• Current methods struggle to isolate sEMGdi effectively, limiting its clinical utility compared to invasive esophageal methods.

Purpose of the Study:

• To present a novel, modified progressive FastICA peel-off (PFP) framework for robust sEMGdi extraction.
• To enhance the accuracy and reliability of noninvasive respiratory signal analysis.
• To overcome limitations of existing methods in extracting weak sEMGdi from complex, noisy data.

Main Methods:

• Utilized a modified progressive FastICA peel-off (PFP) framework.
• Employed an initial FastICA for strong interference removal, followed by constrained FastICA for sEMGdi refinement.
• Validated the approach using both synthetic and clinical datasets, assessing signal-to-interference ratio (SIR) and correlation coefficient (CORR).

Main Results:

• Demonstrated efficient extraction of clean sEMGdi signals with minimal distortion.
• Outperformed state-of-the-art methods in SIR and CORR on synthetic data across various noise levels.
• Achieved high accuracy (95.06%) and F2-score (96.73%) for breath identification in clinical data.

Conclusions:

• The proposed modified PFP framework offers a valuable solution for noninvasive sEMGdi signal extraction.
• This method provides a practical approach for improving ventilator synchrony.
• Significant potential exists for applications in respiratory rehabilitation and overall health monitoring.