Computer Analysis of Electrocardiograms Using a Linear Regression Technique

The paper describes some results of statistical analysis and classification of electrocardiograms (ECG) on a digital computer. The linear regression method is chosen for the analysis mainly because of its conceptual and mathematical simplicity. The analysis is based on the ECG waveform itself rather than the special criteria established by the experience of physicians. Three-lead ECGs for each subject were used. Each ECG was approximated efficiently by 12 orthonormal exponentials. Nine direction cosines were used to describe the spatial orientation of the vectorcardiographic loop. These 45 parameters (3×12+9=45) were regarded as coordinates in a 45-dimensional vector space, and the statistical analysis was aimed at relating certain portions of the vector space with the pathological categories. This procedure is intended primarily as a research tool. With ECGs of 53 subjects, the method has been quite promising. However, demonstration of clinical significance will require additional experimental investigation.