Esophageal image segmentation for guidance of posterior wall lesions during atrial fibrillation ablation

Background Despite luminal esophageal temperature (LET) monitoring, esophageal injury remains a risk which impacts decision making during atrial fibrillation (AF) ablation. We sought to compare procedural characteristics including radiofrequency (RF) power, duration, and LET, among ablation procedures with and without image segmentation for esophageal visualization (EV). Methods The retrospective cohort included 73 patients (mean age 65.2 ± 8.6 years, 36% female, 55% paroxysmal AF) who underwent pre-procedural cardiac magnetic resonance or computed tomography and LET monitoring. Of all patients, 35 were historical patients that underwent standard AF ablation without EV, and 38 were contemporary patients, 28 of whom underwent AF ablation with EV and 10 that underwent AF ablation without EV. Results Total RF time was similar between the groups. The distribution of ablation power delivery was skewed toward higher power in the contemporary patients. However, among patients in the contemporary group, the proportion of > 35 Watts lesions was lower with EV ( P < 0.001). There was no difference between the max or mean LET. The standard deviation of LET change within patient during posterior wall ablation was lower in those with esophageal visualization compared to historical controls, but no change was seen compared to a smaller group of contemporary controls. No long-term clinical esophageal injury was observed. Conclusions In a retrospective analysis, EV was successfully performed in 28 patients. EV impacted RF power delivery decisions but was unassociated with RF time, changes in LET, or long-term safety.