Left ventricular mechanics in human heart failure

Abstract

Heart failure (HF) is a global health threat affecting at 26 million people worldwide (ac-cording to a study done in 2014) and its prevalence is increasing along with the aging of the global population. It is a heterogeneous disease with a wide range of aetiologies. However, current clinical characterisation of HF relies on simplistic chamber-level measurements such as the ejection fraction (EF) of the left ventricle (LV) and LV pressure-volume relationships. Moreover, while therapies for HF with reduced EF (HFrEF) have proven effective, similar treatments have been lacking in efficacy in HF with preserved EF (HFpEF). This is largely due to our limited understanding of the di˙erent processes underlying the different forms of HF. Personalised biomechanical analyses of the LV are able to provide mechanistic insights into the ventricular dysfunction during HF on an individualised basis. In this thesis, a finite element modelling framework was developed by integrating cine magnetic resonance imaging (MRI) data with catheterisation data to analyse passive and contractile mechanical function of the LV. This framework was then applied to estimate stiffness and contractility of healthy and diseased hearts by matching model predicted global motion to that measured using cine MRI. First, a parameter estimation framework was applied to a high-resolution canine dataset including tagged MRI, diffusion tensor MRI, and haemodynamic data, and the identifiability of the stiffness and contractility parameters were investigated extensively. The construction of an objective function for parameter estimation was considered and improved, and the feasibility of simultaneously estimating the passive and contractile parameters was examined. This parameter estimation framework provided the basis for the development and applications in the subsequent studies. The parameter estimation framework was then applied to a clinical dataset provided by the St Francis Hospital in New York and the passive and contractile properties of the LV myocardium were quantified for groups of control, HFpEF, and HFrEF subjects. A novel method was developed to temporally align haemodynamic and imaging measurements, which were not recorded simultaneously. The identifiability of the biomechanical parameters was investigated and a novel method was developed for evaluating an identifiability threshold for passive myocardial stiffness. Group comparisons highlighted the heterogeneity in myocardial stiffness among HFpEF patients, which could help explain the lack of efficacy of treatments for the HFpEF disease group. This also motivates the development of more personalised treatment strategies in the future. HFpEF patients did not show any reduction in maximum contractility or maximum myofibre shortening when compared with control subjects; this affimed the typical clinical understanding of an absence of systolic dysfunction in this disease group. The HFrEF group had significantly elevated myocardial stiffness compared with control. The presence of diastolic dysfunction in HFrEF was surprising since the disease is typically associated with only systolic dysfunction. Myofibre shortening was significantly reduced in systole for the HFrEF group compared with control, however, the maximum contractility of HFrEF patients was not significantly different from control subjects. This suggested that, for these HFrEF patients, the increased myocardial stiffness may help to explain the reduced systolic myofibre shortening, instead of the reduced contractility of the myofibres. Finally, a novel method for simultaneously estimating the load-free LV geometry and myocardial stiffness was developed using principal component analysis. A series of approaches were proposed and their feasibilities were investigated using synthetic data. The optimal technique was applied to the same clinical dataset described above. The inclusion of the load-free geometry in the parameter estimation framework tended to decreased the estimated myocardial stiffness for HF patients compared to estimates derived from models that assumed the diastasis model to be the load-free state. The differences in myocardial stiffness estimates were more pronounced with larger diastasis cavity volumes. This highlighted the importance of load-free model estimation in clinical parameter estimation frameworks. This thesis demonstrated the use of biomechanical analyses to provide tissue-specific evaluations of LV mechanical function which could potentially elucidate HF aetiology. A large portion of the development in this thesis focused on improving the confidence of the estimated parameters by improving their identifiability and by developing a new method of incorporating the load-free geometry into the estimation framework. These investigations improved the clinical applicability of the parameter estimation framework and sets the stage for larger-scale clinical investigations using this method. Biomechanical model-based estimations of myocardial stiffness and contractility have the potential to provide a more personalised, and mechanism-driven treatment strategies for HF patients in the future.