A Novel Approach in Testing the Accuracy of ECG Compression using Partial Percentage RMS Difference and Dynamic Time Warping

Compression of recorded ECG data has been a topic of interest since the introduction of computer storage and later Holter monitor applications of ECG recording technology. Compression is typically performed using an approximation algorithm belonging to one of three classes; direct data compression; transformation and parameter extraction. Some debate exists as to how the accuracy of each implementation should be measured and despite obvious limitations of a percentage RMS difference (PRD) calculation it has been used as a metric for comparing different compression techniques in terms of PRD versus compression ratio. This article questions the use of a percentage RMS difference (PRD) function as a sole measure of the fidelity of the compressed signal by demonstrating the possible error in the compressed signal that a total PRD fails to detect. A novel accuracy test procedure is presented which demonstrates how partial PRD calculations and more importantly dynamic time warping can be used to compare an annotated ECG signal and its reconstructed approximation identifying distortion of clinically significant information. Key-Words: ECG Compression, Dynamic Time Warping, Approximation Error, Dynamic Programming, Direct Data Compression, Pattern Recognition.