A cascaded CNN-LSTM framework for quantifying respiratory motion from surface electromyographic signals | JoVE Visualize

Area of Science:

Biomedical Engineering
Signal Processing
Machine Learning

Background:

Diaphragm surface electromyographic (sEMG) signals are crucial for real-time respiratory monitoring in clinical settings like radiotherapy and intensive care.
Electrocardiographic (ECG) interference commonly contaminates sEMG signals, hindering accurate respiratory motion estimation.
Existing methods for ECG artifact removal often introduce delays and rely on linear assumptions, limiting real-time clinical applicability.

Purpose of the Study:

To develop a robust deep learning framework for real-time respiratory motion quantification from sEMG signals.
To achieve high-fidelity artifact suppression and accurate estimation of respiratory motion.
To overcome the limitations of traditional signal processing techniques in clinical respiratory monitoring.

Main Methods:

A cascaded deep learning framework was proposed, integrating a CNN-LSTM model for respiratory sEMG component isolation and a multi-scale CNN for nonlinear feature abstraction.
sEMG and respiratory data were collected from 45 subjects (20 for training, 25 for validation).
Cross-correlation analysis was used to evaluate the correlation between sEMG-derived respiration and reference signals.

Main Results:

The proposed deep learning method achieved a superior correlation coefficient (Pearson's r = 0.949 ± 0.030) with abdominal pressure-derived respiration compared to gating (0.910 ± 0.046) and template subtraction (0.859 ± 0.081).
The method demonstrated significantly higher correlation with reference signals even without post-processing, highlighting its real-time artifact suppression capabilities.
This indicates robust performance in accurately quantifying respiratory motion from contaminated sEMG signals.

Conclusions:

The developed deep learning framework offers an efficient solution for high-fidelity artifact suppression in sEMG signals.
It enables accurate and real-time respiratory monitoring, crucial for clinical applications.
This approach advances the field of physiological signal processing for improved patient care and monitoring.