An external automatic device to detect ventricular fibrillation

An analysis of a micro-processor-based system to detect cardiacarrest rhythms was made in a series of 84 cardiac arrests in78 patients. The ECG was sensed using 2 external ECGjdefibrillatorpads applied to the chest wall. In 5 arrests, the initial rhythmwas continuous electrical pacing in addition to the cardiacarrest and these were not included in the analysis. Of the remaining79 arrests, there were 15 in which the initial rhythm was ventricularfibrillation and 14 were correctly detected by the system (sensitivity93%). In the remaining 64 arrests the initial rhythm was notventricular fibrillation and 62 were correctly identified (specificity97%). A high specificity is required in any device employingautomatic detection of ventricular fibrillation. Continuouselectrocardiographic recordings of the cardiac arrests wereanalysed every 8 to 18 s: of 223 sections showing ventricularfibrillation, 165 were correctly determined by the system, i.e.sensitivity 74%, where the rhythm was not ventricular fibrillation,of the 5002 sections 4953 were correctly detected by the system,i.e. specificity 99%. The percentage accuracy of detection ofventricular fibrillation varied from 25–100% (mean 81%)and for detection of non-ventricular fibrillation from 57–100%(mean 99%) in every 8 to 18 s section of ECG tracing per cardiacarrest episode. Thus the automatic detection of cardiac arrestrhythms is feasible with a high degree of accuracy.