Kubassova, Olga (2007) Analysis of dynamic contrast enhanced MRI datasets. PhD thesis, University of Leeds.
The purpose of this research is to perform automated analysis of 4D dynamic contrast enhanced MRI datasets (DCE-MRI) of the habd and wrist relating to rheumatoid arthritis (RA) studies. In DCE-MRI, sequences of images are acquired from the joints over time, during which a contrast agent pre-injected into a patient enhances disease affected tissues. Measurement of this enhancement, which is specific to voxels representing particular tissue types, allows assessment of the patient's condition. Currently, analysis of DCE-MRI data is performed using semi-automated or manula techniques, which are time-consuming and subjective. These approaches involve no pre-processing techniques that can compensate for patient motion and hardware instability, or locate the tissue of interest. In this thesis we present a solution for fully automated objective assessment of DCE-MRI data acquired from RA patients. Analysis begins with application of a registration technique that permits compensation for patient motion. Secondly, independent automatic algorithms for accurate segmentation of both bone interiors, joint exteriors, and blood vessels from data volumes of the metacarpophalangeal joints are introduced. Performance of the segmentation algorithms is evaluated with both state-of-the-art and novel techniques developed as a part of this thesis. We have utilised and enhanced a supervised approach and developed a family of unsupervised metrics for automated evaluation of segmentation outputs. Lastly, the datasets are interpreted using a model-based approach, which permits understanding of the behaviour of tissues undergoing the medical procedure, and allows for a robust and accurate extraction of various parameters that quantify the extent of inflammation in RA patients. The algorithms proposed have been demonstrated on datasets acquired with both low and high field scanners, from different joints, using various pulse sequences. They are user-independent, time efficient, and generate easily reproducible and objective results. Expert observers found our results promising for possibly future diagnosis and monitoring of RA patients.
|Item Type:||Thesis (PhD)|
|Additional Information:||Supplied directly by the School of Computing, University of Leeds.|
|Academic Units:||The University of Leeds > Faculty of Engineering (Leeds) > School of Computing (Leeds)|
|Depositing User:||Dr L G Proll|
|Date Deposited:||28 Mar 2011 10:15|
|Last Modified:||08 Aug 2013 08:46|