Machine learning based risk stratification of patients undergoing ventricular tachycardia ablation

Abstract Background Monomorphic ventricular tachycardia (VT) is a potentially life-threatening condition. Although radiofrequency catheter ablation represents an effective treatment method for many of these patients, significant variability observed in postprocedural mortality, which attributable to multiple factors, including the high burden comorbidities. Therefore, there great demand accurate risk stratification system. Purpose We sought implement machine learning pipeline predict 1-year all-cause mortality patients undergoing VT ablation. Methods For 265 consecutive who underwent at our center, we retrospectively collected demographics, medical history, cardiovascular laboratory results, echocardiographic measurements, and ablation-related parameters. To based on features, several supervised models were trained evaluated using 5-fold cross-validation. applied recursive elimination technique identify optimal subset input features. The area under receiver operating characteristic curve (AUC) with 95% confidence interval (CI) was calculated quantify models' performance. also identified most important predictors Shapley values. As final step, used topological data analysis discern visualize patient subgroups different risk. Results 57 (22%) died during follow-up period. In predicting best performance achieved by random forest model utilizing 18 features [AUC: 0.73 (95% CI: 0.68–0.78)]. This significantly outperformed other previously published scores such as I-VT 0.63 0.55–0.70), p<0.001 vs. forest] or PAINESD 0.55–0.71), p=0.009 forest]. mitral E-wave deceleration time, cardiac resynchronization therapy, age, electrical storm, hemoglobin concentration. network created best-performing model, could five subsets clinical characteristics rates (Figure 1). Conclusions Our efficiently Thus, it facilitate prompt identification high-risk personalization strategies, ultimately leading improved outcomes. Funding Acknowledgement Type funding sources: Public grant(s) – National budget only. Main source(s): Heart Program, part Research, Development Innovation Fund HungaryThematic Excellence Programme Ministry Technology Hungary