Abstract
The authors evaluate a frequency-domain algorithm for use by the CREI-GARD computerized arrhythmia system (F. M. Nolle et al., 1986) to detect ventricular fibrillation (VF) in the surface ECG. This was accomplished using (1) The American Heart Association (AHA) database of 80 ECG tape recordings (channel 1), including 10 recordings with 11 documented VF episodes, and (2) The CREI-GARD database of 35 single-channel recordings of 35 documented VF episodes. The recordings were analyzed using sequential 4.096-s segments. All 46 episodes of VF were correctly identified. The average time between the onset of VF and the first detection was 3 s (range 0 to 9 s). VF episodes occurring in segments in association with ventricular tachycardia (VT) were all correctly identified. There are 19 false VF detections during other episodes of VT out of a total of 684 VT episodes (2.8%). There were 5 false VF detections during other rhythms. None occurred during artifact. The algorithm is described in detail.