Algorithms for improved detection of supraventricular arrhythmias

A postprocessing system used for the detection of a subset of supraventricular arrhythmias (SVA), supraventricular tachyarrhythmias (SVTA) and atrial premature beats (APBs), is described. A particularly difficult set of records is selected from a newly created SVA database SVA detection strategies which incorporate criteria based on examining R-R interval statistics in a wider context ('retrospective analysis') in order to identify isolated APBs more accurately are studied. Other strategies which use explicit 'look-ahead' criteria to find runs of SVTA are also studied. The 'retrospective analysis' algorithm yields a modest increase in APC sensitivity (to 44%), offset by a decrease in positive predictivity (to 91%). Applying the 'look ahead' algorithm to the output of the 'retrospective analysis' algorithm makes the APC sensitivity 78%, with 91% positive predictivity.<<ETX>>

[1]  G. Moody,et al.  QRS morphology representation and noise estimation using the Karhunen-Loeve transform , 1989, [1989] Proceedings. Computers in Cardiology.