Physics-based modeling and data-driven algorithm for prediction and diagnosis of atherosclerosis

Abstract

Atherosclerosis, a pattern of the disease arteriosclerosis in which the inner wall of the artery develops deposition of plaques, abnormalities, and lesions. Early screening of this type of vascular anomalies is pivotal to prevent patient’s vascular risk. However, current diagnostic routines based on standard risk factors (e.g., hypertension, diabetes, and dyslipidemia) scoring are not satisfactory. Furthermore, non-invasive imaging routines have substantial restraints such as cost, instrument proficiency as well as renal toxicity, and radiation hazard. In this work, we propose an efficient computational model that integrates data-driven algorithms and physics-based modeling to predict and diagnose atherosclerosis. The method is based on a reduced-order mathematical model to characterize the global and regional vascular apparent wall compliance in addition to the hemodynamic variables. More specifically, we propose to use the fractional-order method which offers an extra flexibility through the fractional differentiation order (FDO). This flexibility in the model allows for an accurate biomarker identification of arterial stiffness. In fact, the FDO inherently controls the interplay transition between the viscosity and elasticity levels of the vasculatures. Moreover, we incorporated machine learning (ML) algorithms to assess carotid-to-femoral pulse wave velocity, which is well-correlated with atherosclerosis. The results indicated that there exists a natural synergy between ML and the FDO mathematical modeling approach. Combining both strategies could provide a reliable and accurate platform for atherosclerosis diagnosis and prevention.

More information

Identifiers

ORCID for Zehor Belkhatir: orcid.org/0000-0001-7277-3895

Catalogue record

Export record

Altmetrics