Quantification of the influence of detailed endocardial structures on human cardiac haemodynamics and electrophysiology using HPC

Abstract

In the last decade, computational modelling has been playing an important role as a non-invasive technology that may significantly impact the diagnosis as well as the treatment of cardiac diseases. As an example, the Food and Drugs Administration (FDA) has created new guidelines for in-silico trials for advancing new device pre-clinical testing and drug cardio-toxicity testing applications, since simulation studies have the potential to accelerate the discovery while reducing the need for expensive lab work and clinical trials. On the European side, the Avicenna Alliance aims to develop a road-map for in-silico clinical trials and establish the bases for the technology, methods and protocols required in order to make possible the use of computer simulations before the clinical trials.

A common characteristic of the existing human cardiac models is that personalised geometries usually come from in-vivo imaging and the majority of computational meshes consider simplified ventricular geometries with smoothed endocardial (internal) surfaces, due to a lack of highresolution, fast and safe in-vivo imaging techniques. Acquiring human high-resolution images would mean for the patient to undergo long, expensive and impractical scans, in the case of magnetic resonance images (MRI), or could present a risk for the patient's health, in the case of computed tomography (CT), since this process involves a considerable amount of radiation. Smoothed ventricular surfaces are indeed considered by the majority of existing human heart computational models, both when modelling blood flow dynamics and electrophysiology. However, the endocardial wall of human (and other mammals species) cardiac chambers is not smooth at all; it is instead characterised by endocardial sub-structures such as papillary muscles (PMs), trabeculations and false tendons (FTs). Additionally, fundamental anatomical gender differences can be found in cardiac sub-structural heart configuration as female hearts present less amount of FTs.

Since there is little information about the role of endocardial sub-structures in human cardiac function, considering them in the human in-silico cardiac simulations would present a first step towards the understanding of their function. Additionally, comparing simulations results including sub-structural anatomical information with those obtained when considering simplified human cardiac geometries (representing common existing models) would shed a light on the errors introduced when neglecting human endocardial sub-structures. Another important aspect which is often ignored in in-silico simulations and could influence their outcome is gender phenotype. Female hearts have reduced resources for repolarization due to differences in K+ channels as compared to male phenotypes, leading to longer action potential durations (APDs). Longer APDs are consistent with clinical observation that females have longer QT intervals (time the heart takes to depolarize and repolarize) than males. Gender specificity can lead then to arrythmogenesis differences and so it may be important to consider different gender phenotypes when running in-silico electrophysiological simulations, in order to obtain results which are of clinical relevance and that can be compared to the subject-specific clinical data.

In this thesis, therefore, we have created highly detailed human heart models from ex-vivo highresolution MRI data, to study the role of cardiac sub-structures and gender phenotype in human cardiac physiology, through computational fluid dynamics (CFD) and electrophysiological highperformance computing (HPC) simulations.