Real-time automatic detection of cardiac arrhythmias by rate and morphology analysis of intracardiac electrograms.

We apply digital signal analysis and pattern recognition to detection of abnormal cardiac rhythms from intracardiac electrograms (electrocardiograms recorded from catheters inside the heart). This addresses the problem of treatment of sudden cardiac death. Recent developments for this problem are Implantable Cardioverter Defibrillators (ICDs) and Pacing-Cardioverter Defibrillators (PCD$s\sp{TM}$), now becoming extremely popular methods of therapy with over 25,000 implantations. However, these devices utilize almost exclusively rate-only detection schema, and they have high incidences of false shocks due to incorrect diagnosis of cardiac arrhythmias. The main contribution of this work is to combine time-domain signal analysis schemes and rate analysis into an overall cardiac arrhythmia detection algorithm, and to implement this system for execution in real-time. This sequential procedure utilizes the least costly features (rate detection) as a first stage, and employs more complex signal analysis methods only for difficult rhythms which confound the rate-only diagnosis. A fast time-domain signal analysis algorithm, the Bin Area Method (a substitute for the classical correlation analysis) was incorporated into this algorithm, and the overall system was tested, yielding a success rate of 95% in a test set of 56 passages of cardiac arrhythmias. To address and solve a subset of the arrhythmia detection problems which remains, a 'sudden onset' criterion has been developed based upon median filtering principles to discriminate between benign and malignant arrhythmias which require different choices of therapy. Signal detection theory was utilized to statistically validate the performance of this criterion. A solution to competing cardiac rhythms that confound all current detection schemes has also been developed by employing statistical measures.PhDElectrical Engineering: SystemsUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/103385/1/9319503.pdfDescription of 9319503.pdf : Restricted to UM users only