AN ALGORITHM TO DETECT LOW INCIDENCE ARRHYTHMIC EVENTS IN ELECTROCARDIOGRAPHIC RECORDS FROM AMBULATORY PATIENTS

An algorithm was devised to detect low incidence arrhythmic events in electrocardiograms obtained during ambulatory monitoring. The algorith m incorporated baseline correction and R wave detection. The RR interv al was used to identify tachycardia, bradycardia, and premature ventri cular beats. Only a few beats before and after the arrhythmic event we re stored. The software was evaluated on a prototype hardware system w hich consisted of an Intel 86/30 single board computer with a suitable analog pre-processor and an analog to digital converter. The algorith m was used to determine the incidence and type of arrhythmia in record s from an ambulatory electrocardiogram (EGG) database and from a cardi ac exercise laboratory. These results were compared to annotations on the records which were assumed to be correct. Standard criteria used p reviously to evaluate algorithms designed for arrhythmia detection wer e sensitivity, specificity, and diagnostic accuracy. Sensitivities ran ging from 77 to 100%, specificities from 94 to 100%, and diagnostic ac curacies from 92 to 100% were obtained on the different data sets. The se results compare favourably with published results based on more ela borate algorithms. By circumventing the need to make a continuous reco rd of the EGG, the algorithm could form the basis for a compact monito ring device for the detection of arrhythmic events which are so infreq uent that standard 24-h Holter monitoring is insufficient.