CRITERIA FOR COMPRESSION OF BIOSIGNALS AND METHODS FOR DATA COMPRESSION IN ELECTROCARDIOGRAPHY

Chr. Zywietz
Arbeitsbereich Biosignalverarbeitung, Medizinische Hochschule Hannover

Digitized raw-data of biosignals are often very voluminous. For example, a 10 s resting ECG record produces typically an amount of 80 KB, for a three lead Holter electrocardiogram with appropriate digitization 80 to 240 MB of data are generated. Processing this large data volume might result in a small number of measurements and interpretative statements which cover not more than a few hundred bytes.

Since the beginning of biosignal processing data reduction and data compression for storage and retrieval has been subject to research and development and has been discussed in many publications. Various methods, like transformations, linear and non-linear signal approximation, utilization of inter-sample correlation, utilization of stationarity of the signal and other methods have been applied to compress data for transmission and storage.

A major issue is quality assurance and the definition of error tolerances for compressed data. Sometimes only the pre-reservation of diagnostic interpretative statements is necessary, sometimes - particularly in the research environment - the conservation of the original signal is of importance if - with newer, more advanced signal processing algorithms - a re-evaluation of the "expensive" original data is desired.

An overview on recently applied methods for biosignal data compression will be given and problems of typically applied quality indicators like root mean square errors, absolute amplitude differences etc. will be presented and discussed.

With regard to practically implemented algorithms and their application the compression of resting ECGs and the development of the Standard Communications Protocol for Electrocardiography will be discussed in detail. This example will demonstrate how data compression on quasi-stationary periodic signals can be applied: how with consideration of signal frequency content, sample correlation and also with consideration of the clinical needs a quasi optimal compression method has been developed. Also, the method of quality assurance will be described in detail. As a result of this research project an internationally accepted interchange format has been developed and standardized. This interchange format makes possible real-time telephone transmission of the conventional 12 lead ECG and also storage of more than 200 resting ECG on a single 1.44 MB floppy disk.