Predicting Early recurrence of atrial fibrilation post-catheter ablation using machine learning techniques

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence

Background:

• Catheter ablation is a standard treatment for atrial fibrillation (AF), yet outcomes vary.
• Predicting ablation success is vital for patient selection and personalized management strategies.
• This study aimed to develop a predictive model for early AF recurrence post-ablation.

Purpose of the Study:

• To develop and evaluate a machine learning model for forecasting early atrial fibrillation recurrence after catheter ablation.
• To identify key predictors of AF recurrence to refine patient selection and post-procedural care.
• To assess the efficacy of the CatBoost model in predicting ablation outcomes.

Main Methods:

• A prospective longitudinal study analyzed data from 402 Iranian AF patients undergoing radiofrequency catheter ablation.
• Machine learning models were developed and evaluated, with CatBoost selected as the best performer.
• Key features influencing recurrence prediction included AF type, renal function, age, and cardiac conditions.

Main Results:

• The CatBoost model achieved 92.5% accuracy in predicting AF recurrence within 3 months.
• High sensitivity (88.6%) and specificity (94.0%) were observed, with an AUC of 0.96.
• Significant predictors included paroxysmal AF, BUN, creatinine, age, mitral regurgitation, and valvular heart disease.

Conclusions:

• Machine learning, specifically the CatBoost model, accurately predicts early AF recurrence post-catheter ablation.
• The model offers potential to enhance patient care by identifying individuals likely to benefit from ablation.
• Further validation with larger cohorts is recommended to confirm generalizability for ablation outcome prediction.