Differentiating atrial fibrillation mechanisms by signal processing of the electrocardiogram

Abstract

A total subcutaneous (SQ) implantable cardioverter defibrillator (ICD) will need to be as sensitive and specific at detecting and treating ventricular arrhythmias as a transvenous ICD. We induced a library of supraventricular and ventricular arrhythmias during ICD implantation that collected simultaneous electrogram data using intracardiac and left chest cutaneous electrodes representing surrogate SQ electrodes and vectors. Methods: A multicenter prospective study registered 143 patients, where 132 patients had usable data for inclusion into the library. Inclusion required a left chest dual-chamber ICD initial or replacement implant with dual-coil high-voltage electrodes. Integrated (INT) and dedicated (DED) bipolar ICD leads were used. Cable connections were attached to all electrodes including intravascular, muscular device pocket tissue, and cutaneous and then were connected to a TEAC digital recorder. At least 4 minutes of baseline rhythm was obtained. Bursts of electrocautery delivered to the pocket were recorded. Programmed stimulation via the atrial and ventricular leads was performed for induction. Induced rhythms were categorized as shockable or nonshockable. An episode inclusion heart rate was greater than 170 beats per minute for at least 20 seconds. Isuprel or atropine were used at the operator's discretion to obtain atrial rates in the target zone. Episodes were divided as RV lead INT, DED, or BOTH. Valid episodes were qualified as equivalent simultaneous SQ and transvenous data. Results: One hundred fourteen episodes in 76 patients met the rate inclusion for nonshockable rhythms, whereas 190 episodes in 110 patients met inclusion for shockable rhythms. Careful analysis found 81 valid nonshockable episodes and 104 valid shockable episodes. The nonshockable valid episodes were distributed as 37 (46%) INT, 3 (4%) DED, and 41 (50%) BOTH. The shockable valid episodes were distributed as 39 (38%) INT, 14 (13%) DED, and 51 (49%) BOTH. Nonshockable episode rates were concentrated between 165 and 190 beats per minute, whereas shockable episode rates were clustered in 3 rate patterns, 180, 240, and 300 beats per minute, with the major concentration at 300 beats per minute. Conclusions: A robust collection of valid shockable and nonshockable episodes were simultaneously and digitally collected from intracardiac and SQ surrogates in an effort to compare the sensitivity and specificity of all available ICD sensing vectors. The sensing algorithm comparative analysis will use the valid episodes of this library.