Evaluation of an automatic cardiac activation detector for bipolar electrograms

The identification of local activation events in bipolar cardiac electrograms, the first step of isochronal map construction, is a time-consuming and difficult process. Owing to the variability among bipolar activation complexes and the lack of practical knowledge concerning the relationship of the bipolar waveform to action potential characteristics, a set of empirical rules to guide the assignment of local activation times have been adopted. A computer program, called AP, has been designed, which implements these rules in the form of a syntactic analyser. Canine epicardial recordings were used to evaluate AP by comparing local activation times, assigned by AP, with times assigned independently by three investigators. The Hermes-Cox model for detector evaluation and a bootstrap statistical method were used in conjunction with ROC analysis to evaluate the ability of AP to detect events. Analysis of discrepancies among investigator-assigned times showed that the reliabilities of AP event detection and AP-assigned times were comparable to those of the investigators. The methods used in system design and evaluation are applicable to a broad range of problems in the detection and localisation of waveform components.