New Threshold for Defining Mild Aortic Stenosis Derived From Velocity-Encoded MRI in 60,000 Individuals.

Summary

Using velocity-encoded CMR and deep learning in 62,902 participants, the authors derived population reference ranges for aortic valve hemodynamics and identified a “mild AS” threshold (>95th percentile) that separates normal from abnormal function. This threshold was prognostically adverse and externally validated in 365,870 people from NEDA, supporting a data-driven redefinition of mild AS.

Key Findings

• Deep learning quantified AV area, peak velocity, and mean gradient from velocity‑encoded CMR in 62,902 UK Biobank participants.
• A natural boundary (>95th percentile) in AV hemodynamics was defined as ‘mild AS’ and associated with adverse outcomes.
• External validation in 365,870 individuals from NEDA confirmed prognostic relevance and generalizability.

Clinical Implications

Clinicians may consider earlier surveillance for patients exceeding the data-driven ‘mild AS’ thresholds and integrate AI-derived CMR or validated echo surrogates into risk stratification, while awaiting guideline updates.

Limitations

• Observational design with potential residual confounding
• CMR-based thresholds may require translation and calibration to echocardiography in routine care

Future Directions

Prospective validation of risk-guided surveillance based on the new thresholds, harmonization with echocardiographic parameters, and assessment of intervention timing.