A new computer program for comparative analysis of serial scalar electrocardiograms: description and performance of the 1976 IBM program.

Abstract An ECG analysis program capable of serial comparison has been written and its performance tested by the authors. The data base consisted of 772 comparisons from patients each having an average of 2.7 ECGs within a 2-month period. In 1939 contour statements, there were 3.3% statements omitted and 6.3% statements erroneously added. In 72 evolving infarctions there were 8.7% statements omitted and 12.5% erroneously inserted. There were 1249 P, Q, QRS, S-T, and T actual measurement changes of which 4.1% were missed. Twelve percent of the measurement change found by the program was not actually present. The program appears clinically acceptable for measurement and analysis of significant serial changes in the majority of ECGs and performs reasonably well in the more difficult task of defining evolution of myocardial infarction.