Periprocedural evaluation of patients with BAV stenosis undergoing TAVR: a machine learning-based study.

Area of Science:

• Cardiovascular Medicine
• Medical Informatics
• Interventional Cardiology

Background:

• Transcatheter aortic valve replacement (TAVR) is a primary treatment for severe bicuspid aortic valve (BAV) stenosis.
• Patients with BAV stenosis undergoing TAVR face significant procedural challenges.
• Accurate risk assessment for periprocedural adverse events (PAEs) is crucial for this population.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting PAEs in patients with BAV stenosis undergoing TAVR.
• To identify key predictors of PAEs in this specific patient group.

Main Methods:

• Retrospective analysis of 1266 patients with BAV stenosis undergoing TAVR.
• Development of an ML prediction model using clinical characteristics and imaging data.
• Definition of PAE encompassing mortality, stroke, hemorrhage, kidney injury, vascular complications, and reoperation.

Main Results:

• Five key predictive factors identified: Type 0 BAV, aortic root calcification volume, horizontal aorta, annular ellipticity, and prior atrial fibrillation.
• The ML model demonstrated robust performance (AUC=0.801) in predicting PAEs.
• A significant graded relationship was observed between risk score quartiles and PAE incidence (0.6% to 9.6%).

Conclusions:

• The developed ML model effectively predicts PAEs in TAVR patients with BAV stenosis.
• This tool facilitates individualized procedural planning and enhances in-hospital management strategies.
• The model's findings were validated in an independent dataset, confirming its generalizability.