Abstract 4139637: Artificial Intelligence ECG Mapping of Biventricular Pacing QRS Morphology Predicts Response to Cardiac Resynchronization Therapy

Background: Cardiac resynchronization therapy (CRT) is an important treatment modality for patients with heart failure and electrical dyssynchrony. A current limitation is that approximately 30% of patients do not respond, and identification of patients at greater risk for suboptimal outcome remains challenging. Hypothesis: We hypothesized that an artificial intelligence (AI) ECG mapping system would identify maladaptive ECG activation patterns in non-responders to CRT-defibrillator (CRT-D) therapy. Aims: We endeavored to analyze the biventricular-paced 12-lead ECG QRS complex using a novel AI tool and compare the projected segment of origin between responders and non-responders. Methods: We enrolled patients with heart failure 6 months or more after CRT-D implantation. ECG samples obtained during routine clinical follow-up during LV, RV, and BiV pacing were analyzed with an AI-based ECG mapping system. The projected segments of origin for each pacing setting were noted using a validated 30 segment ventricular model. CRT responders were defined as ≥5% increase in LVEF at 6 to 12 months post-implantation. Results: Of 49 study patients, 35 were responders (age 74±9 y, LVEF 28±9%, 31% female) and 14 were non-responders (age 68±12 y, LVEF 32±5%, 20% female). LVEF increased in responders by 15±7% and decreased in non-responders by 3±5% (p<0.001). Biventricular-paced ventricular activation patterns were significantly different between responders and non-responders: in non-responders, 8 of 14 patients (57.1%) had an activation pattern identified as mid to apical anterior or anteroseptal (Figure A) versus 6 of 35 responders (Figure B, 17.1%, p= 0.012). Left ventricular and right ventricular activation mapping results were not predictive. Conclusions: A maladaptive biventricular-paced electrical activation pattern, localized to the mid- or apical anteroseptal region of the LV, is associated with non-response to CRT. Studies of intraprocedural noninvasive LV activation mapping to improve target LV pacing site selection and response to CRT are underway.