Computer‐based ECG Analysis in Acute Myocardial Infarction

A routine bipolar ECG has been recorded on magnetic tape in 12 patients with acute myocardial infarction. The recording was started 2–8 h after the onset of symptoms and lasted for 12 h. Minute‐by‐minute classification from a paper chart write‐out revealed ventricular ectopic beats (VBs) in all the patients. Ventricular tachycardia was seen in 6 patients and 2 had ventricular fibrillation, one of whom died. The tape recordings were analysed by a non‐interactive arrhythmia monitoring system. The accuracies of two different algorithms for detecting ventricular arrhythmias were compared. Program I was based on the comparison of all beats with a running average complex and the correlation of abnormal beats with a typical VB waveform. Program II implied the separation of complexes with different morphology on the basis of a five‐parameter description. The diagnosis of a VB was based on discriminant functions and rhythm criteria. A total of 2775 VBs were diagnosed, out of which programs I and II correctly diagnosed 48 and 60%, respectively. With program I only 34% of the alarm situations due to VBs were recognized, whereas 50% were correctly identified by program II. False positive VBs amounted to 0.21 and 0.07%, respectively, out of a total of 631 200 complexes and gave rise to false alarms corresponding to 3/4 of all alarms with program I and about 1/3 with program II—misdiagnosed alarm situations are not included in these figures. However, one patient with a technically bad recording contributed to more than 50% of the false alarms with both programs. Following this study and after minor modifications, an arrhythmia monitoring system based on program II has been taken into routine use in our Coronary Care Unit.