Medical Imaging

EPSRC Centre for Doctoral Training

Research

Projects

  • 301 - The Automated Neuroscientist a Bayesian Optimisation Approach

    An ongoing challenge in cognitive neuroscience is developing novel methods to explore the link between neural activity and cognition. In traditional functional MRI (fMRI) experiments, subjects are scanned while they perform specific tasks (e.g. a memory or visual task) to find out which brain regions the tasks are associated with. In this PhD project, we will develop a novel approach, the “automated neuroscientist”, which turns the typical fMRI design on its head: we will develop novel computational techniques for real-time fMRI to track brain activity, and identify a set of tasks/stimuli that activate a specific target neural state of interest, such as a brain connectivity network. State-of-the-art Bayesian optimisation algorithms will be developed to explore a large bank of pre-defined tasks in real-time by sequentially processing the acquired fMRI data. Potential clinical applications will be explored by assessing individual differences in cognitive dysfunction following focal stroke. More...

  • 302 - Explanatory cardiac motion models based on machine learning

    The motion and deformation of the heart as it beats can provide valuable information about cardiac health. In recent years techniques based on spatiotemporal atlases have emerged that have enabled disease to be diagnosed and characterised from image-derived measurements of motion and deformation. Such atlases establish a common space where the motions/deformations of different subjects can be compared, and machine learning approaches can be subsequently used for classification and regression. However, like most machine learning algorithms, whilst producing highly promising results in terms of accuracy, these techniques typically lack explanatory power. For example, they can tell you what disease is present but they cannot explain in simple terms how they reached that decision. Recent work in the machine learning literature is seeking to overcome this weakness. This project aims to build upon these recent developments to produce a computer-aided diagnosis tool that can analyse cardiac motion and deformation and provide interpretable explanations as well as accurate characterisations of disease. More...

  • 304: Understanding the dual epidemics of atrial fibrillation and heart failure: Image-based computational modelling approach

    The need of managing complex and interlinked cardiovascular epidemics of atrial fibrillation (AF) and heart failure (HF) in the same patient presents a unique therapeutic challenge, and an integrated approach to understanding AF-HF mechanisms is required to design efficient therapy. This project aims to explore the role of two major risk factors common in both AF and HF − fibrosis and adrenergic dysregulation, their contribution to mechanistic links between the two conditions and therapeutic interventions that can mitigate such mechanisms primarily in HF patients with AF. The project will combine medical imaging to quantify atrial structure and fibrosis in AF-HF patients and advanced biophysical modelling to address a fundamental lack of knowledge regarding functional electrophysiological and adrenergic mechanisms underlying the combined AF-HF state. The interdisciplinary approach will be used to produce an image-based computational framework for the stratification of AF-HF patient state, which will help tailor therapy to the need of a patient. More...

  • 306 - Modality propagation and machine learning for detection of the epileptogenic zone from [18F]FDG PET and MR imaging

    Investigate the potential of modality propagation and machine learning for the objective detection of the epileptogenic zone during the pre-surgical workup of patients with refractory focal epilepsy. More...

  • 307 - Development of computational dosimetric models for radiobiological assessment of radionuclide agents for molecular radio-therapy and imaging

    Rapid developments in clinical translation of new radionuclide applications in imaging and molecular radio-therapy (MRT) has led to new treatments in oncology while new chemistry is being translated for theranostics (therapy/diagnostics) involving alpha and beta emitting radionuclides. This requires development of micro-dosimetric models accounting for the sub-cellular dose distributions, currently not addressed by existing models, in order to fully assess the potential of an agent in imaging and therapy. This project aims to develop such computational models based on Monte Carlo methods and the use of in vitro (eg micro-autoradiography) and in vivo (eg small animal SPECT/CT and PET/CT) imaging data to help fully assess the radiobiological profile of new agents in view of clinical translation. More...

  • 309: Advanced image-based computational modelling for patient-specific optimisation of anti-arrhythmia electrotherapy

    Application of strong electrical shocks to the heart is the only currently reliable means of terminating otherwise lethal cardiac arrhythmias (defibrillation). However, the use of such high shock energies is highly detrimental to the patient’s long-term health and psychological well-being. Currently, there is therefore a strong urge to develop novel, lower energy defibrillation devices and shock protocols. This project aims to develop a detailed biophysical understanding of the interaction of electric fields with diseased hearts in order to optimise anti-arrhythmia shock therapy in a personalised manner, reducing required shock strengths. To do this, we will develop a novel computational modelling pipeline to generate high-resolution image-based models of patients with ischemic heart disease (scar). Simulations will be conducted to better predict which individuals may benefit from electrotherapy and optimise the configurations of the electrotherapy devices, depending on the specific nature of their scar in a personalized manner. More...

  • 311: Non-Invasive, MR-Based Assessment of Blood Pressure in the Aorta and Left Ventricle

    This project will create and validate a computational tool for calculating the aortic and left ventricular (LV) blood pressure (BP) waveforms from MR data using blood flow modelling. Aortic BP is an important predictor of cardiovascular events and enables calculation of LV BP. Combined with LV volume, LV BP enables calculation of pressure-volume loops which are of paramount importance to cardiologists for the assessment of cardiac performance and treatment of heart disease. Current measurement of aortic and LV pressures is by invasive procedures. The proposed tool will be tested using our MR-compatible cardiovascular simulator rig and by comparison against invasive BP measurements acquired in patients. The tool will be integrated as an additional module in the Philips Healthcare IntelliSpace Portal software platform to provide a major advance in the characterisation of cardiovascular mechanical function. More...

  • 312: Is endothelial function assessment feasible by non-invasive pulse wave analysis? A computational and in vivo study

    Dysfunction of endothelial cells lining blood vessels is a major risk factor for cardiovascular disease. The aim of the project is to investigate whether endothelial function can be assessed from pulse waves measured non-invasively. This will be achieved by studying haemodynamic mechanisms underlying changes in the shape of peripheral pulse waves produced by pharmacological drugs that alter endothelium function and by determining whether endothelium dysfunction alone is responsible for significant variations in the contour of pulse waves. The research will be carried out using computational modelling of blood flow in the larger systemic arteries of both the rabbit and human. Rabbit data is valuable since a complete anatomic and haemodynamic dataset is available to calibrate and validate numerical models and, hence, study the problem in detail. This research is necessary before non-invasive, cost-effective diagnosis of endothelial dysfunction based on pulse waves could enter routine use. More...

  • 315: Computational Modelling Cellular Variability

    Robust cardiac function depends on the billions of individual cells that make up the heart working together to contract synchronously and pump blood. While all cardiac cells are similar, experimental measurements show that each cell is also distinct. Conventionally experimental and clinical recordings are averaged to remove variability from measurements, however, this approach removes the measurement of potentially important physiological or pathological co-variation and variation. Combining advanced multi-scale biophysical computer simulations of cardiac function (Figure 1) with Bayesian statistical approaches, this PhD will develop a statistical-systems approach to quantify the degree and impact of variability and covariance in clinical and experimental measurements. More...

  • 316: The Connectome Avatar Project

    The project aims to embody our current generation of computational brain models, that are based on the anatomical Human Brain Connectome, into novel sensory-motor simulated Avatars. More...

12 Next