Adaptive Sampling of the Electrocardiogram Based on Generalized Perceptual Features

Area of Science:

• Biomedical Engineering
• Signal Processing
• Medical Informatics

Background:

• Non-uniform information distribution in electrocardiograms (ECG) is recognized, influencing data processing methods.
• Gaze tracking quantifies information density in ECG components by analyzing expert scanpaths.

Purpose of the Study:

• To apply generalized perceptual features for adaptive sampling control in digital ECG.
• To develop a method for non-uniform sampling of ECG signals based on information density.

Main Methods:

• Estimating local ECG bandwidth based on temporal information density.
• Calculating pointwise local sampling frequency and adaptively low-pass filtering channels.
• Interpolating sample values to create non-uniform time series.

Main Results:

• Reconstruction of regularly sampled ECG with 3-5% PRD error for compression ratios of 3.0-4.7.
• Demonstrated good reproducibility of ECG diagnostic features according to IEC 60601-2-25:2015 standards.
• Distortions were localized to less diagnostically relevant parts of the cardiac cycle.

Conclusions:

• Perceptual sampling offers an effective approach for adaptive ECG data acquisition.
• The method preserves essential diagnostic information while achieving data compression.
• This technique supports efficient and accurate digital ECG processing.