Computer detection of premature ventricular complexes: a modified approach.

The accuracy of a data reduction system for arrhythmia detection in identifying premature ventricular complexes was evaluated in continuous tape records of 30 patients in a coronary care unit. Computer analysis was performed with a Honeywell 316 digital computer. Threshold values for dominant complexes were automatically determined and recognition of premature ventricular complexes was based on differences in QRS configuration, timing and T wave configuration from the dominant complexes. Verification of the computer accuracy in detecting premature ventricular complexes was made with visual beat by beat inspection using a two channel strip chart recorder with simultaneous recording of the electrocardiogram and computer signal. This procedure allowed for exact beat to beat correlation and, thus, absolute determination of false positive and false negative identification. From 0.5 to 6 continuous hours of monitoring per patient (average 3.5 hours) were analyzed for a total of 105 monitoring hours. The basic cardiac rhythms noted were normal sinus rhythm, sinue arrhythmia, sinus tachycardia, demand pacemaker rhythm, atrial fibrillation and atrioventricular (A-V) dissociation with junctional rhythm. Premature ventricular complexes were evident in 28 tapes (93 percent) including 12 (43% with multifocal premature ventricular complexes and 3 (11 percent) with ventricular tachycardia. The visual count of premature ventricular complexes totaled 7,921. Of these, 7,542 (95 percent) were properly classified by the computer. The total computer count was 8.717, representing a 13 percent false positive and 5 percent false negative identification rate. The false positive identifications of premature ventricular complexes occurred during periods of 10 seconds or more of continuous noise artifact and in the presence of atrial premature complexes conducted aberrantly. When these sections of tape were excluded, the computer had a less than 2 percent false negative and 3 percent false positive rate of identification of premature ventricular complexes.