Prediction of Ablation Index and Lesion Size Index for Local Impedance Drop-Guided Ablation

(1) <b>Background</b>: The effectiveness of RF ablation for PVI depends on the lesion location and size to achieve continuous and durable lesion lines. AI and LSI are widely accepted lesion metrics for guiding the ablation procedure. LI dynamics is another parameter that guides PVI and does not rely on input variables. Limited data are available on a direct comparison between lesion metrics. Our study aims to compare RF application durations and influencing factors during index-guided (AI and LSI) and LI-guided approaches by predicting lesion metrics using machine learning. (2) <b>Methods</b>: While the coefficients in AI and LSI formulas are not disclosed, we trained custom machine-learning models based on Random Forest and Gradient Boosting Regressors to predict AI and LSI metrics for LI-guided ablations. (3) <b>Results</b>: The median RF application durations differed significantly between the lesion metrics, with 7.32, 19.91, and 11.92 s for AI-, LSI-, and LI-guided procedures, respectively. Mean CF was found to be an important predictor of RF application duration for the AI- and LSI-guided approaches. (4) <b>Conclusions</b>: Depending on the metric used, the significant differences in RF application durations suggest that an AI-guide approach may allow for shorter RF application durations, followed by LSI-guided and LI-guided procedures. Further studies are needed to evaluate the safety and efficacy of these results in a clinical setting.