Multidimensional decomposition framework for electrocardiographic interference, noise and artifacts removal from...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Respiratory Physiology

Background:

• Diaphragmatic electromyography (dEMG) extraction is challenging due to electrocardiographic (ECG) interference, motion artifacts, and noise.
• Existing denoising methods may inadvertently remove physiologically relevant respiratory information.
• Robust suppression of artifacts while preserving dEMG amplitude modulation is critical for accurate respiratory analysis.

Purpose of the Study:

• To develop a novel higher-order singular value decomposition (HO-SVD) algorithm for enhanced dEMG signal processing.
• To enable robust suppression of ECG and motion artifacts while preserving respiratory amplitude variations in dEMG recordings.
• To evaluate the performance of HO-SVD against a 2D SVD method using synthetic and clinical data.

Main Methods:

• A higher-order singular value decomposition (HO-SVD) algorithm was developed, exploiting temporal, QRS complex, and spatial dimensions of dEMG data.
• The HO-SVD method allows targeted suppression of coherent and repetitive artifacts before signal reconstruction.
• Performance was assessed using synthetic datasets and clinical data from patients with obstructive sleep apnea, compared to a 2D SVD (SVD_RI) algorithm.

Main Results:

• HO-SVD significantly improved signal-to-noise ratio (SNR) by up to 27 dB (131%) on synthetic data and 19 dB (88%) with artifacts.
• HO-SVD demonstrated significantly lower symmetric mean absolute error and mean frequency deviation compared to SVD_RI (p<0.0001).
• Clinical data showed HO-SVD had higher correlation with esophageal pressure (p<0.02), better ECG suppression (p<0.0001), and fewer outliers (p<0.0001) than SVD_RI.

Conclusions:

• The HO-SVD framework effectively removes ECG and motion artifacts while preserving respiratory dEMG modulation by incorporating spatial dimensions and adaptive thresholding.
• This novel approach offers a flexible and effective solution for denoising multichannel biomedical signals.
• HO-SVD represents a significant advancement in accurately extracting dEMG for respiratory monitoring and research.