Automatic and efficient R wave discrimination in the right atrium using a two-state hidden Markov model.

Abstract

Discrimination of far-field R waves from atrial events in atrial electrograms (EGMs) is problematic in present implantable pacing systems. Adjustments of atrial refractory periods and sensitivity settings are the only options, and they will not provide optimal performance in many patients. The reliable detection or rejection of R waves in atrial EGMs would avoid problems of atrial undersensing or oversensing, thus benefiting DDD patients by providing more reliable and specific atrial arrhythmia detection. In addition, detection of far-field R waves could allow a measurement of AV conduction time in AAI and aid in discrimination of supraventricular tachyarrhythmia from ventricular tachyarrhythmia. Both atrial and ventricular unipolar EGMs were collected from 25 patients undergoing pacemaker implant or replacement. An average of 141 seconds of intrinsic or VVI paced EGMs was recorded and post analyzed. A new two-state hidden Markov model (HMM) was developed specifically for far-field R wave and P wave discrimination in the atrium. The recorded patients' EGMs were analyzed using this model, and the sensitivity and positive predictivity of far-field R wave detection were evaluated. The collected atrial EGMs were visually examined and marked as the control for verification of the detection analysis. Far-field R wave detection using this model had an overall sensitivity of 94% +/- 9.4% and a positive predictivity of 98.3% +/- 4.4%; and the far-field R wave rejection using the same model had a sensitivity and a positive predictivity of 98.8% +/- 3.8% and 99.1% +/- 1.7%, respectively. Far-field R wave detection in the right atrium by the two-state HMM is reliable and accurate, and can significantly improve atrial arrhythmia management for patients.