MICROPROCESSOR BASED ARRHYTHMIA DETECTION, EDIT, AND RECALL SYSTEM.

A microprocessor based adaptive template matching algorithm for ventricular premature beat detection has been developed. The system learns on the patient's dominant class and stores beats that vary from the dominant class. The recurring beats that are stored in this manner are considered physiological in nature and nonrecurring beats are considered artifact. An algorithm uses the result of the template matches, the prematurity and the frequency of the beats to alarm. The user can edit the data in real time by visual inspection of the stored matches to improve the detection efficiency. An off-line edit and recall terminal permits the user to review the waveform and trends of each stored beat. A cursor permits rapid review of the data. The off-line edit feature also permits the data to be reformatted to eliminate artifacts or unwanted background rhythms. Preliminary data indicate an unaided detection efficiency of 94. 3% true positive, 2. 8% false positive and 5. 7% false negative. The real time editing feature eliminates the background artifact and rhythms and reduces the false positive rate to nil.