Theoretical and experimental rate distortion performance in compression of ambulatory ECGs

The relationship between data rates and signal distortion in compression of ambulatory ECG signals is investigated. The rate-distortion calculations indicate that, for ECG compression with an average data rate of 200 b/s, which would allow 24-h Holter recording with 4 MB of memory, distortion in the compressed signal of almost 15 mu V RMS must be tolerated. The calculations indicate that the use of average beat subtraction should provide data rates near 200 b/s with a root-mean-square (RMS) distortion of about 11 mu V. It is demonstrated that achievable data rates without average beat subtraction are somewhat lower than theoretical estimates for ECG compression without average beat subtraction. It is shown that first difference compression results in data rates that are roughly 30-40 b/s lower than theoretical estimates. When average beat subtraction is used at lower distortion levels, achievable data rates are close to theoretical estimates, but at higher distortion levels achievable data rates are as much as 60 b/s higher than theoretical estimates.