An improved method of discrimination of 1:1 tachycardias using a variance based model of electrical conduction of the heart.

Sudden cardiac death (SCD), which is responsible for 50% of all cardiac fatalities, causes an estimated 300,000 deaths each year in the United States alone. The underlying mechanism for most cases of SCD is a ventricular tachyarrhythmia. This is a condition in which the ventricles of the heart depolarize at an abnormally high and inefficient rate and thus interfere with efficient distribution of oxygenated blood to the body. To address this potentially lethal situation, implantable cardioverter defibrillators (ICDs) have become a widely used treatment. The success of these devices in saving lives has been phenomenal given their high sensitivity for detection of life-threatening arrhythmias, however their lack of accuracy for appropriately discriminating and treating different arrhythmias has been a major cause of concern in ICD technology. The largest problem with accurate detection in ICD devices is their inability to distinguish between supraventricular arrhythmias and ventricular arrhythmias. One particularly difficult subset of these problem arrhythmias is one-to-one (1:1) tachycardias, which demonstrate a matching high rate in both the upper and lower chambers of the heart. For improved discrimination of this subset of 1:1 tachycardias, a novel detection method based on a new feature derived from a variance model of electrical conduction in the heart was developed. The variance model theory suggests that in a supraventricular tachycardia the atrioventricular (AV) conduction intervals will be more stable than ventriculoatrial (VA) intervals. Likewise, in a ventricular tachycardia, VA intervals are expected to be more stable than AV intervals. This interval variability feature has been combined with some of the more traditional detection techniques in an algorithm which performs discrimination of 1:1 arrhythmias. The algorithm was trained with a data set of 45 patients giving a sensitivity to ventricular tachycardia of 100% and a specificity to supraventricular tachycardias of 97%, and subsequently tested on an independent data set of 45 patients resulting in 96% sensitivity and 90% specificity. Overall results for the entire data set were 97% sensitivity and 93% specificity. Performance of the algorithm demonstrates this method is superior to any comparable rate-based discrimination methods available.Ph.D.Applied SciencesBiomedical engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/132903/2/9990999.pd