Learning curve of a novel integrated ICE system withAI-assist auto-contours

Background/Introduction Adoption of novel medical technology can potentially improve procedure efficiency and safety but may be hindered by early challenges in physician familiarity and workflow. Here, we report a sub-analysis of learning curve results from a multicenter experience with a novel integrated ICE system using a new AI-Assist Auto-Contours feature. Purpose Evaluate how procedural effectiveness, efficiency, and safety evolved over the course of physician learning curve using the novel Auto-Contours tool in the ViewFlex X First-in-Human (FIH) study. Methods The ViewFlex X FIH study (NCT06772493) is a first-in-human, non-significant risk study of the ViewFlex™ X ICE Catheter, Sensor Enabled™ integrated with the EnSite™ X EP System. In this cohort, intracardiac echocardiography (ICE) was used for visualization of the chamber of interest and generating a 3D model using contours drawn using Auto-Contours, followed by the creation of a voltage-based model to support ablation strategy. In this sub-analysis, cases were split by operator into early procedures (EaP; Cases 1-3), mid-term procedures (MP; Cases 4-6), and late procedures (LP; Cases 7 and beyond), then grouped for multicenter analysis. Only operators that completed ≥4 cases (e.g., enough cases to reach the MP level) were included in this analysis. Results This sub-analysis included 63 cases. The distribution of cases by indication and learning curve stage can be found in Table 1. Acute effectiveness and safety were consistent across the learning curve, with acute success achieved in 100% of patients in all 3 groups, except for 1 PVC/VT case in the MP group. There was only 1 safety event reported; an episode of post-operative vasoplegia that was not procedure or device related. Although not statistically significant, there was a downward trend in total procedure and ICE modeling time across indications (with the exception of PVC/VT) (Figure 1). Physician-rated accuracy of the ICE model did not differ from EaP to LP (P=0.1971), indicating a high degree of physician confidence in utilizing this novel feature independent of learning curve stage. Conclusion Cardiac ablation supported by a novel integrated ICE system using a new AI-Assist Auto-Contours feature has a high level of acute success and safety regardless of the stage in the learning curve, signaling that there is a low barrier to entry to adopt this new technology. Although not statistically significant, total procedure times and ICE modeling times trended downward, indicating that this feature can be integrated with minimal workflow disruption. As this AI-Assistant AutoContour feature is new and only available to a small number of physicians, further analyses evaluating the impact of physician learning curve on patient outcomes from a larger dataset are warranted.