Computer classification of ST and T in averaged ECGs at rest and exercise

A computer program for ST and T classification at rest and exercise is described. The automatic ST classification was based on the modification of the Minnesota code suggested by Punsar et al. The T wave classification was based on the Minnesota code of 1968. Recordings at rest, during and after exercise from 217 exercise tests were visually classified by two observers, and their result was compared to the computer's. Computer-observer comparisons yielded no more disagreements than between-observer comparisons. There were computer-observer disagreements caused by computer errors in 3% of the records. The repeatability of the computer's ST classification on records taken during exercise was better than that of one visual observer. Within-computer and within-observer repeatability appeared to be negatively affected by remaining noise in the average.