Application of kernel principal component analysis for single-lead-ECG-derived respiration

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiology

Background:

• Principal Component Analysis (PCA) of heartbeats is a recognized method for deriving respiratory signals from electrocardiograms (ECGs).
• Existing methods may not fully capture complex, nonlinear relationships within ECG data for accurate respiration estimation.

Purpose of the Study:

• To introduce and evaluate an improved ECG-derived respiration (EDR) algorithm utilizing kernel Principal Component Analysis (kPCA).
• To enhance the accuracy of respiration signal extraction from single-lead ECGs by accounting for data nonlinearities.

Main Methods:

• Employed kernel Principal Component Analysis (kPCA), a generalization of PCA that uses kernel functions to map data into a higher-dimensional space.
• Compared various kernels, identifying the radial basis function (RBF) kernel as optimal for EDR signal derivation.
• Tuned the variance parameter (σ^2) of the RBF kernel for further performance optimization.

Main Results:

• The kPCA-derived EDR signals demonstrated statistically significant improvements in correlation and magnitude squared coherence coefficients compared to PCA and R peak amplitude methods (p<0.0001).
• The tuned RBF kernel within kPCA provided superior performance in extracting respiratory information from ECG signals.

Conclusions:

• Kernel Principal Component Analysis (kPCA) offers a significant advancement over traditional PCA and R peak amplitude methods for ECG-derived respiration estimation.
• The proposed kPCA algorithm, particularly with a tuned RBF kernel, provides a more accurate and robust method for extracting respiratory signals from single-lead ECGs.