Signal Acquisition-Independent Lossless Electrocardiogram Compression Using Adaptive Linear Prediction.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Medical Informatics

Background:

• Electrocardiogram (ECG) data requires efficient compression for storage and transmission.
• Existing ECG compression methods may not always achieve optimal lossless compression ratios.
• Adaptive linear prediction and entropy coding are promising techniques for signal compression.

Purpose of the Study:

• To propose and evaluate a novel lossless ECG compression method.
• To assess the effectiveness of prediction error-based adaptive linear prediction combined with modified Golomb-Rice coding.
• To determine the optimal prediction order for ECG compression.

Main Methods:

• Implemented a lossless ECG compression algorithm utilizing adaptive linear prediction to minimize prediction errors.
• Employed modified Golomb-Rice coding to encode the prediction errors into binary code.
• Evaluated the method on the PTB Diagnostic ECG, European ST-T, and MIT-BIH Arrhythmia databases.
• Investigated prediction coefficients using the autocorrelation method to determine the optimal prediction order.

Main Results:

• Achieved average compression ratios of 3.16, 3.75, and 3.52 for single-lead ECG signals across the evaluated databases.
• Demonstrated consistent performance despite variations in signal acquisition setups.
• Confirmed that second-order linear prediction is suitable for ECG compression applications.

Conclusions:

• The proposed lossless ECG compression method effectively reduces data size while preserving signal integrity.
• The combination of adaptive linear prediction and modified Golomb-Rice coding offers a robust solution for ECG data compression.
• Second-order linear prediction is validated as an appropriate choice for ECG compression.