A special program has been developed for on-line processing of orthogonal electrocardiograms during exercise with a small computer system. Special measures were taken to obtain high measurement accuracy and to prevent errors due to baseline drift, skeletal muscle noise or premature beats. The program was tested on exercise electrocardiograms from 47 normal men and 50 patients with coronary artery disease. Single representative complexes were obtained by averaging of beats selected with the aid of the spatial velocity time function of the QRScomplex and the STsegment. The method used is comparable to, but faster than, cross correlation of the spatial velocity time functions. The onset and end of the QRS complex during exercise were determined using templates derived from the first record of the particular subject. This algorithm resulted in a lower measurement error than other methods for waveform analysis.Depression of the ST segment of the electrocardiogram (ECG) recorded during exercise in subjects with a normal ECG at rest is considered to be a specific in,dication for the presence of myocardial &hernia (1, 2, 3). It also has a prognostic value for the development of ischemic heart disease in otherwise asymptomatic subjects (4,5). However, the interobserver variation in the visual interpretation of exercise ECG's is large (6). This is caused in part by the amount of noise present in the signal and in part by the different criteria employed for classification. With digital computer techniques the signal to noise ratio of the ECG and thus the measurement accuracy can be improved. Also multivariate analysis of ECG measurements can now be performed. It is therefore to be expected that computer assisted interpretation will improve the diagnostic value of exercise ECG's (7).Averaging of a number of heart beats is often employed to reduce the amount of noise in the ECG caused by skeletal muscle contractions during exercise. However, averaging may cause considerable errors particularly when the time alignment of the beats is not accurate, or when premature beats are present during exercise. Also QRS complexes may be detected erroneously when sudden baseline shifts take place. Special techniques should therefore be employed to avoid these sources of artifacts.In this study a new method is presented for the detection of QRS complexes and for averaging of the exercise ECG. It is based on selection of those QRS complexes and ST segments which have a similar waveform. In addition an improved method

No abstract