Long-duration electrocardiogram classification based on Subspace Search VMD and Fourier Pooling Broad Learning System

Area of Science:

• Biomedical Engineering
• Signal Processing
• Cardiovascular Health

Background:

• Detecting early cardiovascular disease from short Electrocardiogram (ECG) signals is difficult.
• Long-duration ECG data acquisition is prone to noise, complicating analysis.
• Existing methods struggle with high dimensionality and unclear spatial characteristics in long ECG recordings.

Purpose of the Study:

• To develop advanced signal processing techniques for accurate early cardiovascular disease detection.
• To address noise and data complexity challenges in long-duration ECG analysis.
• To enhance feature representation and reduce dimensionality for improved diagnostic accuracy.

Main Methods:

• Subspace Search Variational Mode Decomposition (SSVMD) was employed for noise reduction and optimal parameter selection in ECG signals.
• A Fourier Pooling Broad Learning System (FPBLS) was proposed, integrating Fourier features and broad pooling to manage data dimensionality and enhance feature clarity.
• The MIT-BIH arrhythmia database was utilized for empirical validation of the proposed methods.

Main Results:

• SSVMD effectively preprocesses ECG data by removing baseline drift and high-frequency noise modes.
• FPBLS successfully reduces data dimensions while preserving critical features, improving spatial characteristic representation.
• The combined approach demonstrated superior performance compared to the latest state-of-the-art methods in arrhythmia detection.

Conclusions:

• The proposed SSVMD and FPBLS methods offer a robust solution for early cardiovascular disease detection using long-duration ECG signals.
• These techniques effectively mitigate noise and data complexity, paving the way for more reliable diagnostic tools.
• The study validates the efficacy of the novel methods, showing significant improvements over existing approaches.