T1 Mapping–derived Predictors of Cardiac Remodeling and Fibrosis in Athletes using Advanced Machine Learning...

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Athletes often exhibit physiological cardiac remodeling.
• Distinguishing physiological from pathological changes is crucial for cardiovascular health.
• Cardiac remodeling and myocardial fibrosis can be assessed using cardiovascular magnetic resonance (CMR) T1 mapping.

Purpose of the Study:

• To predict cardiac remodeling and/or myocardial fibrosis in athletes.
• To utilize machine learning models with T1 mapping data from CMR.
• To identify key imaging biomarkers for early detection of adverse cardiac events.

Main Methods:

• 104 athletes and 20 controls underwent 3.0T CMR scans.
• Measured cardiac function, native T1, and extracellular volume (ECV) in 16 left ventricular segments.
• Employed Gradient Boosting Machines (GBM), logistic regression, CART, and SVM for prediction models.

Main Results:

• Athletes showed higher ECV and lower native T1 values compared to controls.
• Athletes with cardiac remodeling had significantly higher native T1 and ECV values in specific segments.
• The GBM model achieved an AUC of 0.899, with native T1 (segment 10), ECV (segment 3), and body surface area as top predictors.

Conclusions:

• Increased native T1 and ECV values in athletes correlate with cardiac remodeling, suggesting a link to myocardial fibrosis.
• Early CMR monitoring of T1 and ECV can assess risk and guide management in athletes.
• A GBM model effectively predicts adverse cardiovascular events using T1 mapping parameters.