Development and assessment of new post-processing methodologies in 3D contrast enhanced MRI

Brennan, David (2004) Development and assessment of new post-processing methodologies in 3D contrast enhanced MRI. PhD thesis, University of Glasgow.

Abstract

This thesis aims to investigate some of the methods currently used in contrast MR imaging. It specifically focuses on methods that require subtraction of noncontrast enhanced (pre) 3D imaging data sets from contrast-enhanced (post) data, collected within a single imaging session. Current methods assume that there is little or no intra-scan patient motion and thus do not attempt to correct for this. This thesis aims to determine if such motion does exist and if so what methods are best suited to correct it. The thesis begins by describing some of the relevant MR physics and history of contrast enhancement in chapter 1, and expands on this in chapter 2 by focusing on angiographic, and contrast-enhanced techniques. Chapter 2 continues by investigating an MP RAGE subtraction technique for producing venograms, which requires pre and post-contrast data subtraction. Data is collected for 20 patients and the effects of motion correction on the resulting venograms are investigated. Chapter 3 investigates a different type of pre and post-contrast enhanced study where it is used for tumour volume measurement. Examining the effects on tumour volumes measured with and without the realignment correction provides quantitative evidence that realignment is a requirement in this and similar types of study. To enable the significance of segmentation accuracy on realigmnent to be tested a phantom pre and post-contrast data set is developed in chapter 4. Chapter 5 uses this data set to test the effects of differing segmentation accuracies, with respect to the accurately segmented phantom data, on realignment accuracy where the pre and post-contrast data differ by known rotations and translations. This provides information on the effects of contrast enhancement on realignment accuracy, as well as providing information on the required brain segmentation accuracy required to accurately realign these data sets. Chapter 6 expands on this work by testing segmentation accuracy effects on two real patient data sets. The first patient data set differs from the phantom data in terms of its noise characteristics and the second has a space occupying lesion similar to those regularly encountered in the clinical setting. Chapter 7 aims to develop an automatic technique for segmenting, realigning and visualising venographic data using the venography technique described in chapter 2. It uses a histogram and morphological operations to ensure that all of the contrast enhanced data is removed from the data, whilst attempting to segment the brain to an acceptable accuracy. Although this algorithm is specifically designed for venograms visualisation, it would require only a small amount of adjustment enabling it to be applied to the tumour volume measurement technique described in chapter 3. Chapter 8 tests this algorithm using the data collected in chapter 2 and measures its performance in producing satisfactory brain segmentations, which is required for accurate realignment. This would also be required for accurate realignment in tumour volume measurement studies. Chapter 8 also measures the algorithms capabilities in correctly producing visualisation data sets for the purposes of venography. The algorithm has limited success in both brain segmentation and venous visualisation, nevertheless this is encouraging as a first attempt as the algorithm is being applied to real patient data sets reflecting a range of pathological conditions and not only to selected normal data sets. Chapter 8 suggests some modifications that could be applied to the algorithm that might improve its future success. This includes modifying it to become a semi-automated technique. (Abstract shortened by ProQuest.).

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Keywords: Medical imaging.
• Colleges/Schools: College of Medical Veterinary and Life Sciences
• Supervisor's Name: Condon, Dr. Barrie
• Date of Award: 2004
• Depositing User: Enlighten Team
• Unique ID: glathesis:2004-71091
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 10 May 2019 10:49
• Last Modified: 12 Jul 2021 15:41
• URI: https://theses.gla.ac.uk/id/eprint/71091