Machine Learning-Based Algorithm to Predict Procedural Success in a Large European Cohort of Hybrid Chronic Total...

Area of Science:

• Cardiology
• Interventional Cardiology
• Machine Learning in Medicine

Background:

• Chronic total occlusion percutaneous coronary intervention (CTO-PCI) success rates are lower than non-CTO PCI.
• Existing scores for predicting CTO-PCI outcomes have suboptimal discriminatory performance.
• Accurate prediction of CTO-PCI success is crucial for procedural planning and patient management.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based model for predicting procedural success in CTO-PCI.
• To integrate clinical and angiographic data for improved prediction accuracy.
• To compare the performance of the ML model against established complexity scores.

Main Methods:

• A European multicenter cohort of 8904 patients undergoing CTO-PCI was used.
• A Light Gradient Boosting Machine (LightGBM) model was trained on 75% of the data and tested on 25%.
• Sixteen clinical, 16 angiographic variables, procedural volume, and 3 complexity scores (J-CTO, PROGRESS-CTO, RECHARGE) were utilized.

Main Results:

• The best ML model (LightGBM) achieved an AUC of 0.82 (training set) and 0.73 (test set) for procedural success prediction.
• The ML model significantly outperformed conventional scores: J-CTO (AUC 0.66), PROGRESS-CTO (AUC 0.62), and RECHARGE (AUC 0.64).
• The ML-based model demonstrated superior accuracy in predicting CTO-PCI procedural success.

Conclusions:

• Machine learning models can accurately predict procedural success in CTO-PCI.
• This ML-based approach offers potential for tailored patient management strategies.
• Further prospective validation in real-world settings is recommended to integrate this model into clinical decision-making.