1st Supervisor: Martin Bishop, King’s College London
2nd Supervisor: Pablo Lamata, King’s College London
Clinical Champions: Aldo Rinaldi, King’s College London and Sanjad Prasad, Royal Brompton
Aim of the PhD Project:
The ultimate aim of this work is to build an enhanced risk prediction model to identify susceptibility to cardiac arrhythmias in dilated cardiomyopathy patients by identifying characteristic structural changes in ventricular anatomy and tissue (fibrosis), in combination with features identified by biophysically-detailed computational modelling.
Project Description / Background:
Clinical Motivation – Dilated cardiomyopathy (DCM) is a form of heart failure that is characterised by significant dilation and thinning of the ventricular cavities, along with tissue structural (fibrotic) remodelling. Unfortunately, this condition is associated with a high incidence of lethal ventricular arrhythmias. At-risk patients are given an implanted cardioverter defibrillation (ICD) device that can reliably protect against lethal arrhythmias. However, such devices have a significant number of limitations, including implant complications along with a high incidence of (painful) inappropriate shocks, which can lead to intolerable levels of distress and psychological issues with device recipients, along with long-term cardiac consequences of frequent strong shocks. These issues are more pertinent as many DCM patients are relatively young, due to the strong genetic component of the disease. Accurate risk-stratification for device need is therefore vital, but remains a significant clinical challenge with existing tools. Consequently, many patients receive an (expensive, unpleasant) device (with potential long-term consequences) that is never used, whilst other patients die suddenly in the community without a device due to misdiagnosed risk. Thus, being able to improve risk assessment of DCM patients is of utmost clinical need.
State-of-the-art – Current methods of risk stratification rely mainly on MR imaging, specifically looking at the degree of ventricular dilation (as a marker of disease progression) along with the presence of fibrosis, which is known to be a significant marker of risk in these patients. However, such assessment is relatively crude and relies largely on clinical judgement. Machine learning based risk prediction models are being increasingly used in different areas of cardiology. For arrhythmic risk prediction, such techniques have been employed that largely use features derived from structural (MR) imaging, although they have yet to be applied to the specific case of DCM, with its complex structural and tissue remodelling. Moreover, many machine learning models fail to realise full clinical uptake or enthusiasm due to lack of clinical interpretability.
Specific Goal – In this project, we aim to produce a refined risk prediction model that applies shape-based analysis tools to identify key morphological modes of variation that are associated with arrhythmic risk in the DCM population. Biophysically-based computational modelling will be used to provide a clinical and mechanistic understanding of the morphological arrhythmogenic features identified from the shape analysis. Furthermore, functional simulations using the derived models will also be used to determine additional quantitative features regarding the functional affects of such shape variations that will be used to augment the risk prediction model.
Candidate Background – This project would suit a candidate with a strong physical science background (applied maths, physics, computer science, engineering) with a particular interest in medicine and physiology.
CMR images (left) converted into 3D shape models (centre) which can be used to identify arrhythmogenic patterns of remodelling which may be mechanistically explained through biophysically-based computational simulation (right).