Background: High-density mapping for scar-related ventricular tachycardia (VT) is increasingly used. In the setting, manual annotation of abnormal potentials/target for ablation is challenging.
Objective: We sought to assess the accuracy of the automatic annotation of abnormal potentials with the Lumipoint algorithm (Rhythmia, Boston Scientific).
Methods: Hundred consecutive patients undergoing scar-related VT ablation in 2 tertiary centers have been included. Local abnormal ventricular activities (LAVAs) and ablation points were manually annotated during the procedure. Offline, these points were compared with automatic annotation using the Lumipoint algorithms for late potential (LP, after QRS), fragmented potential (FP, >7 spikes), and double potential (DP).
Results: Maps were acquired during sinus rhythm, right ventricular pacing, and left ventricular pacing in 34 cases, 63 cases, and 3 cases, respectively. The average mapping points were 9976 ± 7477. Automatic annotation identified the same area as a manual annotation (<10% difference) in 64 cases (64%), it missed certain areas with manual annotation in 18 cases (18%), and it highlighted an area of abnormal potentials not recognized during manual annotation in 15 cases (15%). No correlation was observed between manual and automatic annotation in 3 cases (3%). During sinus rhythm, FP achieved the highest proportion of true LAVA area in annotated area, then LP achieved the higher proportion compared to DP (FP, 86% ± 18%; LP, 72% ± 25%; DP, 50% ± 23%: FP vs. LP or DP, P < 0.001; LP vs. DP, P = 0.046). During right ventricular pacing, although FP still had the highest LAVA proportion, LP had the lower proportion compared to DP (FP, 86% ± 13%; LP, 47% ± 33%; DP, 60% ± 22%: FP vs. LP or DP, P < 0.001; LP vs. DP, P = 0.007). Notably, LP achieved a higher LAVA proportion in cases with the scar on the left ventricular lateral wall compared to cases without, even during right ventricular pacing (lateral scar, 79% ± 18%; no lateral scar, 49 ± 32%: P < 0.001).
Conclusion: Lumipoint algorithms allowed correct automatic identification of abnormal substrate of scar-related VT. FP showed high accuracy for the detection of LAVA regardless of underlying rhythm, whereas accuracies of LP and DP were affected by the rhythm during mapping.