A novel multifidelity machine learning framework accurately predicts hemodynamics in stenosed vessels, potentially reducing the need for extensive experimental data and improving patient-specific risk assessment.
Coronary artery disease in presence of lesions in the coronary arteries may restrict blood flow to the heart. Thus, a better understanding of the hemodynamics involved and development of rapid and accurate methodology will enable clinicians to assess patient-specific risk in order to administer timely interventions. In this study, it was proposed to develop and validate a new multifidelity machine learning framework to address limitations of the computational fluid dynamics approaches widely used to model hemodynamics in coronary and peripheral arteries. Proposed method significantly reduces the amount of experimental (including in-vivo) data required for training and will make the arterial pressure predictions less dependent on the availability of specific boundary conditions. Integration of various data sources allowed to introduce a set of physiology-related features (e.g., viscosity, blood flow, resistance, elasticity, etc.) and to achieve, in particular, an excellent agreement with mean values of in vitro pressure measurements: 1.2 ± 3.5 mmHg.
Velikorodny et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: