AI-Enhanced Prediction of Aortic Stenosis Progression: Insights From the PROGRESSA Study

Area of Science:

• Cardiology
• Medical Informatics
• Artificial Intelligence in Medicine

Background:

• Aortic valve stenosis (AS) is a chronic, progressive condition with unpredictable individual progression rates.
• Accurate prediction of AS progression is crucial for timely intervention and patient management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting the progression of aortic valve stenosis using longitudinal patient data.
• To compare the performance of artificial intelligence models against traditional clinical prediction methods.

Main Methods:

• Trained machine and deep learning algorithms on data from 303 patients in the PROGRESSA study with annual clinical and echocardiographic follow-up.
• Evaluated model performance in predicting AS progression at 2- and 5-year intervals.
• Compared AI model performance against a standard logistic regression-based clinical model.

Main Results:

• LightGBM demonstrated superior predictive performance for AS progression at both 2- and 5-year terms, achieving high area under the curve values (0.85 and 0.83, respectively).
• Recurrent neural networks (GRU, LSTM) and XGBoost also showed strong predictive capabilities.
• The traditional clinical model exhibited the lowest predictive performance.

Conclusions:

• Artificial intelligence-guided approaches can significantly enhance risk stratification for patients with aortic valve stenosis.
• Multisource comprehensive data and advanced AI models offer improved prediction of disease progression and clinical outcomes in mild-to-moderate AS.
• AI integration into clinical routines holds promise for personalized management of AS.