A novel approach for estimating blood flow dynamics factors of eccentric stenotic arteries based on ML

Abstract

Reliable and rapid estimation of blood flow dynamics factors in eccentric stenotic arteries could significantly improve clinical treatments. Numerical simulation methods such as FSI and CFD are widely used to investigate blood flow conditions. However, both FSI and CFD are computationally expensive and not suitable for large-scale research. This work proposes an effective approach for estimating the blood flow dynamics factors of eccentric stenotic arteries based on ML. The estimation approach includes three steps: (a) blood flow conditions of idealized eccentric stenotic arteries modeled with different geometric parameters were simulated by CFD and FSI, and the error between CFD and FSI results were evaluated, (b) datasets of nonlinear relationships between discretized geometric parameters and blood flow dynamics factors were created using CFD, and (c) blood flow dynamics factors of eccentric stenotic arteries were estimated by carefully designed ML model. The accuracy validation was conducted by different representative cases, which have different combinations of geometric parameters including stenosis severity, eccentricity, and stenosis plaque length. The ML model can output the blood flow dynamics factors including peak wall shear stress and peak systole velocity accurately within 1 s, which shows proposed approach not only achieves accurate estimation of blood flow dynamics factors but also bypasses the expensive computational cost.