Forewarned Is Forearmed: Machine Learning Algorithms for the Prediction of Catheter-Induced Coronary and Aortic...

Area of Science:

• Cardiovascular Medicine
• Interventional Cardiology
• Health Informatics

Background:

• Catheter-induced dissections (CID) are severe complications of percutaneous coronary procedures.
• Existing data on CID risk factors are largely anecdotal, necessitating robust predictive models.

Purpose of the Study:

• To identify significant predictors of CID using advanced machine learning techniques.
• To develop a predictive model for clinical decision support in interventional cardiology.

Main Methods:

• Analysis of 84,223 coronary procedures (2000-2022) to identify 124 CID cases.
• Application of logistic regression and five machine learning algorithms, including Extreme Gradient Boosting (XGBoost).
• Evaluation of model performance using f1-score for feature importance ranking.

Main Results:

• XGBoost demonstrated optimal performance in predicting CID.
• Key predictors identified include guiding catheter use, small/stenotic ostium, radial access, hypertension, acute myocardial infarction, prior angioplasty, female gender, chronic renal failure, atypical coronary origin, and COPD.
• A 'perfect dissection candidate' profile was defined based on predictor clustering.

Conclusions:

• Machine learning algorithms significantly enhance risk prediction for CID.
• The developed model offers valuable clinical decision support for interventional cardiologists.
• Identifying patients with clustered risk factors can guide procedural modifications to minimize dissection risk.