A Novel ECG-Derived Respiration Method Combining Frequency-Domain Feature and Interacting Multiple Model Smoother.

Area of Science:

• Cardiorespiratory physiology
• Biomedical signal processing
• Electrocardiography (ECG) analysis

Background:

• ECG-derived respiration (EDR) is a valuable, low-cost method for monitoring respiratory activity.
• High-quality EDR is crucial for accurate cardiorespiratory studies.
• Existing EDR methods face challenges with noise and signal variations.

Purpose of the Study:

• To develop and validate a novel EDR method utilizing spectral features and advanced smoothing techniques.
• To enhance the accuracy and robustness of EDR signals.
• To improve the correlation between EDR and reference respiration.

Main Methods:

• A new EDR feature based on the first moment of the power spectrum (FMS) was proposed.
• The Interacting Multiple Model (IMM) smoother was applied to refine the EDR signal.
• Performance was evaluated against R-peak misalignment, noise, waveform correlation, and respiratory rate estimation using the Fantasia and Drivers databases.

Main Results:

• The FMS feature demonstrated improved robustness against R-peak offsets compared to existing methods.
• A slight 5% improvement in waveform correlation was observed under accurate R-peak conditions.
• The IMM smoother showed comparable performance to the Kalman filter in static data and enhanced EDR waveforms in ambulatory data.

Conclusions:

• The proposed method offers a promising approach for EDR signal enhancement by analyzing frequency domain characteristics of ECG.
• Treating EDR as a non-stationary time series facilitates better fitting of natural respiration.
• This work provides a foundation for improved cardiorespiratory monitoring using ECG.