Coronary artery disease (CAD) is one of the most common causes of death in the world. Diagnosis is based on imaging of the artery, either by CT or conventional angiography. Conventional angiography is an invasive technique which involves the introduction of a system of guide-wires, catheters and radiopaque contrast agent into the patient’s coronary arteries. Fractional Flow Reserve (FFR) is widely considered to be the gold-standard assessment of the physiological significance of CAD. FFR is measured invasively by the passage of a pressure wire through the diseased artery.
It is hypothesised that a computational model can be employed to characterise the haemodynamics of blood flow in patient-specific coronary arteries in order to compute clinical indices of interest, including FFR, in an effective and reliable way. The aims of this project are to combine a coronary artery reconstruction tool with Computational Fluid Dynamics (CFD) and Reduced Order Modelling (ROM) techniques to estimate the pressure drop and FFR in patient-specific coronary arteries in a fast and accurate way.
This thesis comprises two parts, both associated with the effective computation of FFR from angiographic data:
i. The first part addresses the problem of accurate reconstruction of coronary artery anatomy from multiple, single-plane coronary angiography (MSPCA), to underpin the creation of a computational model. A segmentation tool with a user-friendly graphical user interface (GUI) was developed in MATLAB to generate the surfaces meshes required for the CFD studies and to obtain other clinically-relevant coronary parameters.
ii. The second part focuses on the effective and accurate computation of the pressure gradient and the FFR using ROMs built from CFD solutions in ANSYS-Fluent, exploiting the ANSYS ROMBuilder suite. The methods were applied to compute pressure profiles along the length of the artery, and FFR, in representations of coronary stenosis. The study includes the identification of an appropriate parameterisation of the artery shape to support the effective construction and operation of a ROM, as well as an evaluation of the sources of error and a comparison between results from Bernoulli estimates, from 1D models, from ROM, from CFD and from clinical measurement. Sequential increases in complexity of the anatomical representation are made, from axisymmetric with idealised stenoses to realistic radius variations from a coronary artery dataset and finally to curved arteries. In all cases each arterial cross-section is assumed to be circular. The study includes analysis of the interaction between idealised serial stenoses, and of a dataset of 140 patient-specific arteries characterised by angiography
It was demonstrated that ROM applied to idealised coronary geometries achieved accuracies comparable with CFD results, and better than other approaches, in dramatically reduced timescales (order 900 times reduction relative to CFD). Limitations and opportunities for improvement include more accurate reconstruction of the cross-sectional profiles, more comprehensive representation of 3D curvature in the ROM and improved automation of segmentation, but the ROM approach shows great promise for this application in the delivery of solutions of sufficient accuracy in timescales consistent with the clinical process.
