Independent component analysis and decision trees for ECG holter recording de-noising

Area of Science:

• Biomedical Engineering
• Signal Processing
• Medical Informatics

Background:

• Electrocardiogram (ECG) signals are crucial for diagnosing cardiac conditions.
• Noise and artifacts in ECG recordings can lead to misdiagnosis.
• Existing denoising methods may struggle with specific types of interference.

Purpose of the Study:

• To develop and evaluate a novel ECG denoising method using Independent Component Analysis (ICA).
• To compare the proposed ICA-based method against a wavelet-based denoising technique.
• To assess the algorithm's effectiveness in removing both standard and uncommon ECG artifacts.

Main Methods:

• Development of an ECG de-noising algorithm combining JADE source separation and a binary decision tree.
• Application of the algorithm to publicly available ECG data from Physionet.
• Comparative analysis using Root Mean Square Error (RMSE) against a wavelet-based filter.

Main Results:

• The proposed ICA method achieved comparable results to standard filtering for common noises (power line interference, baseline wander, EMG).
• Significantly better performance was observed in removing uncommon noise, specifically electrode cable movement artifacts.
• The algorithm demonstrated robust noise removal capabilities across different artifact types.

Conclusions:

• The developed ICA-based ECG denoising technique offers an effective solution for noise removal.
• This method shows particular promise in addressing challenging artifacts not well-handled by conventional filtering.
• The approach provides a valuable tool for improving the quality and diagnostic accuracy of ECG signals.