# Prediction of Aortic Stenosis Progression Using Artificial Intelligence: A Machine Learning Model

## TL;DR

This study develops an AI model using echocardiograms to predict which patients with mild or moderate aortic stenosis will progress to severe disease, aiming to improve monitoring and treatment timing.

## Contribution

A novel AI model that reliably predicts progression of aortic stenosis using only echocardiography data, with strong performance metrics and interpretability.

## Key findings

- The model achieved an area under the curve of 0.91 and 83% accuracy in predicting severe aortic stenosis progression.
- 1,625 patients (47%) progressed to severe AS during follow-up, and the model successfully identified high-risk individuals.
- The model demonstrated robust calibration and generalizability through cross-validation.

## Abstract

Current guidelines for monitoring aortic stenosis (AS) progression focus on serial echocardiographic assessment, which is resource-intensive and subject to variability. Artificial intelligence may offer an opportunity to enhance the early identification of patients at risk of developing severe AS.

The objective of this study was to create an echo-based model that can predict whether a patient will deteriorate from mild/moderate AS to severe AS.

We retrospectively analyzed a single-center database of 529,751 echo exams and identified 9,330 echocardiograms of patients initially diagnosed with mild or moderate AS, 56% of which progressed to severe AS within 5 years. We developed a model agnostic to any patient data outside the scope of the echocardiography report, and the reports were obtained from a large database of a tertiary medical center. Performance was assessed for accuracy, area under the curve–receiver operating characteristic, and calibration SHapley Additive exPlanations values provided interpretability for the model's predictions.

During the follow-up, 1,625 (47%) patients developed severe AS. The model demonstrated strong predictive performance—an area under the curve–receiver operating characteristic of 0.91, an accuracy of 83%, and an Integrated Calibration Index = 0.0576. The model successfully identified patients at high risk of progression, with robust calibration and generalizability confirmed through cross-validation.

Our novel, echocardiography-focused artificial intelligence model is a reliable tool for the early identification of patients at risk of progression to severe AS. Pending future, multicenter, prospective validation, such models may facilitate personalized follow-up strategies and timely interventions, ultimately leading to improved patient outcomes and resource utilization.