Patient-specific computational haemodynamics in the arteriovenous fistula and aorta of a chronic kidney disease cohort

End-stage renal disease (ESRD) patients require renal replacement therapy (RRT) for filtering wastes and excess fluids from blood, often due to kidney transplant waiting times. Haemodialysis, the most common RRT, requires a vascular access for an efficient and repeatable procedure. An arteriovenous fistula (AVF) is widely considered the ‘gold-standard’ vascular access. However, AVFs suffer from poor patency rates, and their non-maturation is a key issue in failing cases. Localised haemodynamics and flow patterns are hypothesised to be prevalent factors behind AVF success. In addition to this, the impact of AVF creation on other parts of the haemodynamic network requires further elucidation. For example, the presence of an AVF is known to increase cardiac load and alter resistances in the vascular network. CFD enables a robust investigation into haemodynamic metrics that are not easily measurable in-vivo, such as wall shear stress (WSS) and oscillatory shear index (OSI). CFD can therefore play a vital role in the pre-operative planning of AVFs, in addition to the postsurgical assessment. Despite the breadth of AVF research that has utilised CFD, there is still no clear consensus on the mechanisms of AVF failure that result in fistula re-intervention or abandonment. Neointimal hyperplasia, the gradual loss of luminal patency due to the thickening of the vascular wall, and inadequate outward remodelling, an insufficient increase in venous diameter and elasticity, are the two major mechanisms behind AVF failure. Considering the highly patient-specific nature of AVF, vascular access guidelines have proposed a patient-centred care approach for determining an AVF site and morphology. Evaluating the localised haemodynamics in such patient-specific anatomy requires the acquisition and segmentation of high-quality images, a traditional blocker in this research area due to the contraindicating nature of historic contrast agents in the ESRD community. In this research, high-quality images from ferumoxytol-enhanced magnetic resonance angiography (FeMRA) were segmented for a cohort of chronic kidney disease (CKD) patients. These segmentations were then subsequently used for three CFD investigations, which is the first time FeMRA and CFD have been coupled together. The principal aims of this research were to i) investigate the influence of a successful AVF on a patient’s proximal haemodynamics, ii) replicate haemodynamics within a cohort of patient-specific AVF vessels to examine the impact of several metrics on future outcomes, and iii) to examine the haemodynamic differences in the subclavian arteries between successful and unsuccessful AVF cases within a cohort. The novelty in this research is four-fold, i) the segmentation and dataset registration workflow used for generating a patient-specific model (from aorta to AVF), ii) the use of computational fluid dynamics (CFD) for investigating proximal haemodynamics to an AVF, iii) the use of ferumoxytol as the contrast agent for generating patient-specific 3D vasculature models for a set of juxta-anastomosis CFD simulations, and iv) aortic haemodynamic CFD investigations in an ESRD cohort with AVFs. The investigations of this research demonstrated that i) the changes in vascular resistances induced from the creation of the AVF were found to markedly increase the WSS in the proximal vasculature of a 19-year-old patient with a radiocephalic AVF (Chapter 3), ii) the influence of cross-sectional juxta-anastomosis vessel characteristics (notably the feeding artery curvature) are the most prominent factor in determining AVF future success (Chapter 4), and iii) the WSS in the subclavian artery of the arm used for the AVF is 3-4 times higher than the opposite subclavian artery (Chapter 5). Having demonstrated the feasibility of generating patient-specific anatomy (using FeMRA) for CFD analysis, the future possibilities of research area are greatly increased. It is proposed that segmenting the AVF of several more cohorts can contribute to a repository/library of reference AVF models with haemodynamic analysis. Additionally, segmentation of further cohorts can open the doors to statistical shape modelling (SSM). Completing the workflow within this research for multiple cohorts, or by implementing SSM for the generation of multiple patient-specific derived anatomies of successful and unsuccessful AVF anatomies, can improve pre-surgical planning, as the surgeon can have a library of prior AVF in-silico trials to refer to.End-stage renal disease (ESRD) patients require renal replacement therapy (RRT) for filtering wastes and excess fluids from blood, often due to kidney transplant waiting times. Haemodialysis, the most common RRT, requires a vascular access for an efficient and repeatable procedure. An arteriovenous fistula (AVF) is widely considered the ‘gold-standard’ vascular access. However, AVFs suffer from poor patency rates, and their non-maturation is a key issue in failing cases. Localised haemodynamics and flow patterns are hypothesised to be prevalent factors behind AVF success. In addition to this, the impact of AVF creation on other parts of the haemodynamic network requires further elucidation. For example, the presence of an AVF is known to increase cardiac load and alter resistances in the vascular network. CFD enables a robust investigation into haemodynamic metrics that are not easily measurable in-vivo, such as wall shear stress (WSS) and oscillatory shear index (OSI). CFD can therefore play a vital role in the pre-operative planning of AVFs, in addition to the postsurgical assessment. Despite the breadth of AVF research that has utilised CFD, there is still no clear consensus on the mechanisms of AVF failure that result in fistula re-intervention or abandonment. Neointimal hyperplasia, the gradual loss of luminal patency due to the thickening of the vascular wall, and inadequate outward remodelling, an insufficient increase in venous diameter and elasticity, are the two major mechanisms behind AVF failure. Considering the highly patient-specific nature of AVF, vascular access guidelines have proposed a patient-centred care approach for determining an AVF site and morphology. Evaluating the localised haemodynamics in such patient-specific anatomy requires the acquisition and segmentation of high-quality images, a traditional blocker in this research area due to the contraindicating nature of historic contrast agents in the ESRD community. In this research, high-quality images from ferumoxytol-enhanced magnetic resonance angiography (FeMRA) were segmented for a cohort of chronic kidney disease (CKD) patients. These segmentations were then subsequently used for three CFD investigations, which is the first time FeMRA and CFD have been coupled together. The principal aims of this research were to i) investigate the influence of a successful AVF on a patient’s proximal haemodynamics, ii) replicate haemodynamics within a cohort of patient-specific AVF vessels to examine the impact of several metrics on future outcomes, and iii) to examine the haemodynamic differences in the subclavian arteries between successful and unsuccessful AVF cases within a cohort. The novelty in this research is four-fold, i) the segmentation and dataset registration workflow used for generating a patient-specific model (from aorta to AVF), ii) the use of computational fluid dynamics (CFD) for investigating proximal haemodynamics to an AVF, iii) the use of ferumoxytol as the contrast agent for generating patient-specific 3D vasculature models for a set of juxta-anastomosis CFD simulations, and iv) aortic haemodynamic CFD investigations in an ESRD cohort with AVFs. The investigations of this research demonstrated that i) the changes in vascular resistances induced from the creation of the AVF were found to markedly increase the WSS in the proximal vasculature of a 19-year-old patient with a radiocephalic AVF (Chapter 3), ii) the influence of cross-sectional juxta-anastomosis vessel characteristics (notably the feeding artery curvature) are the most prominent factor in determining AVF future success (Chapter 4), and iii) the WSS in the subclavian artery of the arm used for the AVF is 3-4 times higher than the opposite subclavian artery (Chapter 5). Having demonstrated the feasibility of generating patient-specific anatomy (using FeMRA) for CFD analysis, the future possibilities of research area are greatly increased. It is proposed that segmenting the AVF of several more cohorts can contribute to a repository/library of reference AVF models with haemodynamic analysis. Additionally, segmentation of further cohorts can open the doors to statistical shape modelling (SSM). Completing the workflow within this research for multiple cohorts, or by implementing SSM for the generation of multiple patient-specific derived anatomies of successful and unsuccessful AVF anatomies, can improve pre-surgical planning, as the surgeon can have a library of prior AVF in-silico trials to refer to