A change-point method for multi-lead electrocardiogram monitoring using weighted multivariate functional principal...

Area of Science:

• Cardiology
• Biostatistics
• Signal Processing

Background:

• Cardiovascular diseases (CVDs) are a leading cause of global mortality, often linked to coronary artery issues.
• Electrocardiogram (ECG) signals are crucial for diagnosing cardiac conditions, with 12 leads monitoring heart electrical activity.
• Existing change-point detection methods for multi-channel ECG lack flexibility in weighting important channels.

Purpose of the Study:

• To develop an advanced change-point detection method for monitoring long-term cardiovascular treatment effectiveness.
• To enhance diagnostic accuracy by incorporating channel significance into ECG analysis.
• To address limitations in current methods for analyzing complex, multi-channel physiological signals.

Main Methods:

• Represented 12-lead ECG data using a third-order tensor (beats × samples × leads).
• Developed a novel Weighted Multivariate Functional Principal Component Analysis (WMFPCA) approach.
• Integrated WMFPCA with Hotelling's T² statistic for constructing monitoring schemes, considering intra-beat, inter-beat, and inter-lead correlations.

Main Results:

• The proposed WMFPCA-based method demonstrated superior performance in monitoring multi-channel processes compared to existing techniques.
• Simulation results confirmed the effectiveness of the novel approach.
• The model successfully validated on a real-world dataset of Myocardial Infarction (MI) patients.

Conclusions:

• The novel WMFPCA method offers a more flexible and accurate approach for change-point detection in multi-channel ECG data.
• This technique provides a valuable tool for monitoring long-term cardiovascular treatment and diagnosing conditions like MI.
• The study highlights the potential of tensor-based analysis and weighted functional principal components in cardiovascular signal processing.