Deriving Explainable Metrics of Left Ventricular Flow by Reduced-Order Modeling and Classification

Area of Science:

• Cardiovascular Imaging
• Biomedical Engineering
• Machine Learning in Medicine

Background:

• Extracting explainable cardiac flow metrics is challenging for clinical applications.
• Reduced-order models (ROMs) offer a strategy for deriving interpretable clinical metrics from advanced cardiac flow imaging.
• Integrating ROMs with machine learning (ML) can provide insights for diagnosing and risk-stratifying cardiac patients.

Conclusions:

• Modal decomposition of cardiac flow can generate ROMs representing normal and pathological flow patterns.
• This approach uncovers simple, interpretable flow metrics capable of discriminating between disease states.
• The developed ROMs and derived metrics are suitable for further analysis using ML for clinical applications.