Evaluation of an event classifier for pacemaker ventricular evoked response detection

In order to improve depolarization waveform classification, a new classifier for pacing evoked response vs. lead polarization was developed. The classifier was based on identification of polarization characteristics, considering deviations to be indicative of capture. Waveform classifier features were trained using saline tank and acute patient (pt) data. A total of 31 pts with 9 lead models were studied acutely prior to pacemaker implant. Following placement of lead(s), pts were paced at 1.0-4.8V (bipolar and unipolar), paced at specific AV delays to induce fusion, and paced to determine pacing threshold using an external pacemaker. The first 12 pts were part of the training set and the remaining 19 pts represented the test set. Over 24,000 ventricular pacing pulses were analyzed. Test set sensitivity was 99.5% and specificity was 97.2%. In conclusion, the capture event classifier showed improved fusion classification and a greater lead tolerance.