Detecting Aortic Valve Anomaly From Induced Murmurs: Insights From Computational Hemodynamic Models

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Acoustics

Background:

• Transcatheter aortic valve replacement (TAVR) requires continuous monitoring for complications like leaflet thrombosis.
• Current monitoring methods (e.g., angiography) are costly, invasive, or use nephrotoxic agents, limiting routine use.
• Traditional heart sound auscultation is subjective and declining in proficiency, hindering reliable diagnosis.

Purpose of the Study:

• To develop and validate a novel, non-invasive, and practical auscultation-based technique for monitoring aortic valve function.
• To establish a computational model correlating hemodynamic phenomena and valve motion with acoustic signals.
• To enable early detection of anomalous valve function using machine learning for patient-centric monitoring.

Main Methods:

• Numerical simulations of blood flow and valve dynamics in healthy and stenotic aortic models.
• Modeling heart sound propagation as elastic waves through the thorax to the body surface.
• Extracting "acoustic signatures" using principal component analysis and training a linear discriminant classifier.

Main Results:

• Accurate prospective detection of anomalous valve function was demonstrated.
• Principal component-based acoustic signatures effectively captured key audible diagnostic features.
• The developed classifier correlated recorded heart sounds with valve status.

Conclusions:

• Auscultation-based monitoring using machine learning offers a practical, non-invasive, and non-toxic approach for aortic valve surveillance.
• This technology has the potential for inexpensive, safe, at-home monitoring of TAVR, surgical, and native valves.
• Further development can significantly improve post-procedural patient care and long-term valve health management.