CDT Postgraduate Studentship in Healthcare Innovation (RCUK Digital Economy Programme grant number EP/G036861/1)
Cardiac segmentation from Cardiac MRI.
Cardiac segmentation from clinical images is an on-going area of research. Accurate and practical segmentation enables better quantification of existing clinical parameters, and has the potential to lead to better understanding of physiology, as well as pathology via new biomarkers of disease. The vision for the project is to define computational models of electromechanical activity of the heart, using MRI image analysis and personalised modelling techniques, to investigate recovery from myocardial infraction, particularly in the acute phase.
The current stage of the research focuses on the definition and refinement of an anatomical model, based on a probabilistic approach to cardiac segmentation; strengths and weaknesses of such an approach are investigated with respect to the existing literature. An approach to pure statistical classification using 2D or higher feature space for classification was attempted. This approach, involving pixel intensity and relevant distance metrics as features, with the addition of Markov Random Fields as spatial priors, showed a high number of false positive results for cardiac tissue. Further approaches will employ more feature-guided and anatomically-driven approaches, while retaining the probabilistic nature of the segmentation (such as probabilistic level sets), in order to exploit partial volume effects to build more accurate models from MRI.