A Multimodal Video-Based AI Biomarker for Aortic Stenosis Development and Progression

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Aortic stenosis (AS) poses a significant public health challenge with limited current biomarkers for personalized patient management.
• Existing biomarkers for AS lack the precision needed for tailored screening and follow-up strategies.
• A novel video-based artificial intelligence (AI) biomarker, the Digital AS Severity index (DASSi), has shown potential in detecting severe AS from echocardiography.

Purpose of the Study:

• To evaluate the capability of the DASSi biomarker in identifying the development and progression of AS in patients with no or mild-to-moderate AS at baseline.
• To assess the association of DASSi with future aortic valve replacement (AVR) risk.
• To explore the cross-modal translation of DASSi from echocardiography to cardiac magnetic resonance (CMR) imaging.

Main Methods:

• A retrospective cohort study involving two large patient cohorts undergoing echocardiography (Yale New Haven Health System and Cedars-Sinai Medical Center).
• Development of a computational pipeline for translating DASSi analysis from echocardiography to CMR imaging using UK Biobank data.
• Analysis of annualized changes in peak aortic valve velocity (AV-Vmax) and AVR events, correlating them with baseline DASSi scores.

Main Results:

• Higher baseline DASSi scores were significantly associated with faster progression of AV-Vmax in both echocardiographic cohorts.
• DASSi scores of 0.2 or greater were linked to a 4- to 5-fold increased risk of AVR compared to scores below 0.2, independent of traditional risk factors.
• The association between DASSi and AS progression was validated in a large cohort undergoing CMR imaging, with DASSi ≥0.2 showing an 11-fold higher AVR risk.

Conclusions:

• The DASSi, an AI-based video biomarker, is independently associated with AS development and progression.
• DASSi enables opportunistic risk stratification across various cardiovascular imaging modalities, including echocardiography and CMR.
• The potential exists for DASSi's application on portable devices, enhancing accessibility for AS screening and monitoring.