B-PO02-081 LEVERAGING A CLOUD BASED ARTIFICIAL INTELLIGENCE ENGINE TO ANALYZE DATA FROM REAL AF REGISTRY

Abstract

Real-AF is a multicenter prospective atrial fibrillation ablation (AFA) registry. CARTONET is a cloud-based solution embedding artificial intelligence (AI) algorithms that allow for procedural data aggregation and statistical analysis. To demonstrate the feasibility of using CARTONET to evaluate AFA characteristics from multiple sites in the REAL AF registry at a granular level. Procedural data of 187 subjects from 4 US sites in REAL-AF were analyzed using CARTONET. Wide antral circumferential ablation (WACA) segments were determined using a machine learning algorithm (anterior, inferior, roof, posterior). Average ablation duration, number of ablation sites, contact force (CF), stability, power, and ablation index for each segment were calculated. Baseline characteristics were as follows: mean age 66±11 years, BMI 30±6, CHADS2VASC 2.71±1.7, and HAS-BLED 1.68±1 with 52% male. Mean ablation duration was 21.9±13.3 mins with left WACA 6.5±3.1 and right WACA of 7.6±3 mins. Left WACA mean stability was 3.5±0.6 mm compared to right WACA stability of 3.3±0.8 mm. Mean ablation index ranged from 362 to 453 with the lowest being at the left posterior and the highest at the right anterior. Mean CF ranged from 11 g at left anterior to 15.9 g at right inferior. CARTONET enables analysis of ablation characteristics that will allow for a more granular approach to research in large multi-center studies and registries, such as REAL-AF. Future studies will be able to observe trends in ablation characteristics arising from new techniques and technologies.