Tachycardia Detection in Implantable Antitachycardia Devices: Pattern Recognition Algorithms for Reliable Classification.

Abstract

The development of implantable devices for the management of cardiac arrhythmias has revealed limitations in present technology for the recognition of tachycardias. Electrical intervention from implantable antitachycardia pacemakers and defibrillators is presently in the stage of clinical testing in hundreds of patients. Implantable drug pumps for pharmacologic management are being proposed. All of these systems rely on hardware circuitry for detection of pathological tachycardias. It is the hypothesis of this thesis that reliable detection schemes will require the application of modern computer technology. In particular, digital signal processing techniques and sophisticated pattern recognition schemes are appropriate for the next generation of these devices. To address these problems, I have developed several algorithms which employ new pattern recognition schemes to classify arrhythmias. Extensive testing and evaluation has been done not only with tape-recorded electrograms, but on patients in the clinical setting during studies with temporary (acute) leads, and also in a series of dogs with chronic leads. Two confounding arrhythmias emerge in the detection of tachycardias treated by implantable devices: 1:1 tachycardia in which a physiological response to exercise or stress is misdiagnosed as the clinical tachycardia based on rate alone; and atrial fibrillation in which loss of signal amplitude related to the severe decrease in amplitude during fibrillation creates a sensing problem which confounds the diagnostic software. An atrial extrastimulus technique has been designed for the discrimination of physiological tachycardia from 1:1 paroxysmal tachycardia. Clinical tests on seven patients demonstrated the success of this interventional technique. It is applicable to all but a small subset of these tachycardias. To address the problem of reliable tachycardia detection for the discrimination of atrial fibrillation from sinus rhythm, four major algorithms have been designed and tested on a series of dogs in long-term studies of chronic leads. Since algorithms may be affected by long-term change of electrograms and by introduction of drugs, basic studies of the effect of lead maturation and cardioactive drugs on electrogram amplitude and slew rate have been performed. Five drugs have been examined and quantitatively assessed for evidence of change in the electrogram. (Abstract shortened with permission of author.)