Medical Imaging

EPSRC Centre for Doctoral Training

Research

Projects

  • 105 - A machine learning approach to solving the SAR problem for ultrahigh field MRI

    Ultrahigh field (UHF) MRI has the potential to provide very high resolution anatomical images with new types of contrast compared with more conventional lower field MRI. However the required high-frequency RF fields are highly spatially variable inside the human body, and in order to image outside the brain multiple transmitters (so called parallel transmit, pTx) are required1. An issue with this approach is that quantification of potential RF heating effects is not straightforward. Currently UHF-MRI facilities use conservative safety limits that ensure safety but stop UHF-MRI from reaching its true potential. UHF-MRI also often requires specialist operators, limiting workflow and clinical applicability. More...

  • 108 - Quantitative MRI of the fetus

    Fetal MRI has been established in clinical practice for investigation of pathology, and more recently in research into brain development. As the fetus develops the brain undergoes dramatic morphological and microstructural changes, that can be visualised using MRI (see Fig. 1). During the second half of pregnancy important developmental processes take place, in particular neuronal migration, formation of white matter tracts, beginnings of myelination and appearance of cortical folds. These changes can be observed on standard MR images, both as changes in morphology (folding for example) or the change of apparent contrast. Contrast is determined both by the underlying tissue properties (T1, T2, T2*, …) and the properties of the imaging sequence. More...

  • 115 - Unified multi-dataset and multi-parametric synergistic PET-MR image reconstruction with deep learning

    Positron emission tomography (PET) is a powerful medical imaging modality for the brain, for cancer and for the heart, and the way the noisy raw data are processed in order to deliver an image (image reconstruction) is a crucial component to the success of PET. Similarly, for the immensely successful modality of magnetic resonance imaging (MRI), the way that fast acquisitions (often requiring undersampled data) are reconstructed has a significant impact on their end point image quality. More...

  • 301 - Explaining the predictions of deep learning models in cardiology

    The use of machine learning techniques in cardiology is an active research area and recently state-of-the-art results have been achieved by applying the latest ‘deep learning’ techniques to tasks such as segmentation of the myocardium and blood pool [1]. Machine learning has also been used to identify and exploit the useful information carried by the motion and deformation of the heart as it beats [2,3]. More...

  • 302 - Image-based optimisation of life-saving implanted cardiac defibrillation devices using advanced patient-specific computational whole-torso models

    Sudden cardiac death remains one of the largest killers in Western Society. Implanted cardioverter defibrillator (ICD) devices are a highly effective means of primary and secondary prevention of sudden cardiac death; however, they have a number of significant limitations. Successful defibrillation via ICDs requires very high shock energies, which limits ICD battery life, and causes significant pain and resulting psychological disorders for recipients receiving inappropriate shocks. Complications also arise from standard ICD electrodes being implanted inside the heart cavity itself. More...

  • 303 - Predicting Optimal Heart Failure Treatment with Patient Specific Models

    The LV pumps blood around the body and the location of lead that paces the LV plays a crucial role in determining CRT outcome. Evaluation of different pacing sites have found that on average pacing the LV (lateral) free wall returns better outcomes and currently represents best practice. However, a consistent finding in lead position studies is that in a significant patient cohort the LV free wall is not the optimum site and indeed some studies found it to be the worst pacing site for specific patients. More...

  • 304 - Non-invasive assessment of pulmonary hypertension using imaging and pulse wave analysis

    PH is a life-threatening condition of the pulmonary circulation clinically defined by a sustained elevation of blood pressure in the pulmonary arteries. The main vascular changes in the initiation and progression of PH are vasoconstriction, remodelling, increased arterial stiffness, thrombosis, and vascular cell proliferation. These homeostatic imbalances produce profound haemodynamic changes, including a substantial increase in pulmonary vascular resistance, reduction of vessel distensibility (flexibility), and reduction of cardiac function, ultimately leading to right ventricular failure and death.1 More...

  • 305 - Spatio-temporal modelling of fetal and neonatal brain development using multi-modal MRI

    During the second half of pregnancy the brain undergoes rapid development. Important changes include neuronal migration, formation of white matter tracts, the onset of myelination, cortical folding and elaboration of cortical microstructure. Disruptions to these processes result in life-long neurodisability. In particular, infants who are born preterm have altered cortical and white matter development that can be identified around the time of normal birth (40 weeks gestation). However, the critical time of onset of these developmental abnormalities is not clear. More...

  • 306 - Developing an fMRI-diagnostic framework for characterising cognitive control impairments based on machine learning analyses of dynamic network states

    Recent neuroimaging research has demonstrated that the brain has a highly adaptable functional network structure. Different networks are specialised to different functions, but the diversity of tasks that we face in everyday life can tap these functions in different combinations. This leads to the observations of ‘dynamic network states’. Simply put, the vast array of possible tasks that we can perform is reflected by the diverse conjunctions of network configurations that the brain can transiently express. These are visible to fMRI as transient patterns of correlated activity across brain regions. More...

  • 307 - Development of a decision support tool for neuroimaging using explainable ML

    The purpose of this project is to develop an ML-based decision support tool for brain Magnetic Resonance Images (MRI) with application to neurological disorders. A particular focus will be on the development of a decision support tool that will not only be able to diagnose patients based on clinical imaging and non-imaging information, but will also be able explain how it has reached its decision. For this, we will use a large database of multi-modal brain MRI from patients with different forms of dementia. More...

12 Next