Longitudinal Validation of a Deep Learning Index for Aortic Stenosis Progression.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Aortic stenosis (AS) is a progressive cardiovascular disease requiring diligent monitoring.
• Transthoracic echocardiography is the standard, but deep learning offers advanced tracking potential.
• A deep learning-derived index for AS continuum (DLi-ASc) was previously developed.

Purpose of the Study:

• To examine longitudinal changes in the DLi-ASc.
• To assess the prognostic association of DLi-ASc with progression to severe AS.

Main Methods:

• Retrospective analysis of 2373 patients with 7371 echocardiographies.
• Longitudinal tracking of DLi-ASc (scaled 0-100) derived from echocardiographic views.
• Median follow-up of 42.8 months.

Main Results:

• DLi-ASc increased with worsening AS stages and correlated strongly with velocity and pressure gradients.
• Higher baseline DLi-ASc predicted faster AS progression.
• Annualized DLi-ASc change correlated with annualized velocity and gradient progression.
• Baseline DLi-ASc independently predicted severe AS progression.

Conclusions:

• DLi-ASc increases with AS progression and is independently associated with progression to severe AS.
• DLi-ASc serves as a noninvasive, imaging-based digital marker.
• Findings support DLi-ASc for longitudinal AS monitoring and risk stratification.