Visualization and quantification of HIV associated lipodystrophy from magnetic resonance images

Files in This Item:
File Description SizeFormat 
OSullivan_ucd_5090D_10052.pdf6.14 MBAdobe PDFDownload
Title: Visualization and quantification of HIV associated lipodystrophy from magnetic resonance images
Authors: O'Sullivan, Tadhg
Advisor: Rainford, Louise
Permanent link:
Date: 2012
Online since: 2015-08-25T14:31:44Z
Abstract: PurposeThis research aimed at developing a software tool for the purposes of quantifying andvisualizing HIV-associated lipodystrophy from full body magnetic resonance imaging(MRI) datasets. The primary goal for developing the software tool was to create andcompare the results gathered from MRI to those from the current gold standard, dualenergy X-ray absorptiometry (DEXA). A software based solution for this purpose isproposed and a full evaluation with the intention of future clinical use is presented. Theadditional aim of volume visualization in order to assess the external morphologicaleffects of HIV-associated lipodystrophy is also presented.MethodsThe data gathered for this study involved a cohort of HIV positive cases (n = 8) whichwere recruited in order to be scanned by both Dual Energy X-ray Absorptiometry(DEXA) and MRI techniques to facilitate comparison between the two modalities.The accurate identification and segmentation of adipose tissue from MRI datasets wasidentified as one the key components of this piece of research. A fully automaticsegmentation algorithm was implemented for this purpose. Quantification of segmentedadipose tissue and surface based volume visualization were implemented as the primary features of the software tool.The fully automatic segmentation algorithm was investigated in regards to accuracy andperformance. In order to evaluate the clinical relevance of the results of segmentation, acomparison of the results to those of the current gold standard (DEXA) was performed.Clinical feedback regarding the usefulness of the software tool in a clinical setting is alsopresented.Visualization of adipose distribution and external morphology from full body data werealso identified as an important component of this project. A surface based volumevisualization technique was implemented in order to allow users to view a patient’sexternal morphology. Application of a heat map to the surface in order to intuitivelyvisualize the distribution of adipose tissue was also implemented.ResultsThe findings of this study indicate that the results gathered by the software tooldeveloped compared well to those of the current gold standard. A strong correlationbetween the results of the two modalities was found with a correlation coefficient r of0.68 and significance level of p < 0.0001 with a very small 95% confidence interval. Areasonable level of agreement between the modalities was also recorded, the meandifference in fat measurements between the two was 5.62%. A panel of MR experts,Radiology (n = 2), MSc MRI Radiography Specialists (n = 3), PhD Medical ImagingMRI Specialists (n = 2) evaluated the segmentation technique used and it was found to be accurate and, due to the fact it was automatic, its results were 100% reproducible.ConclusionsIn this study the segmentation, quantification and visualization of adipose tissue from fullbody MRI dataset in place of the current gold standards was targeted and investigated.A proof of concept software tool was developed for this purpose and was evaluated foraccuracy and clinical relevance. The findings presented provide the evidence base that an appropriate tool was developed and could be used with MRI as an alternative to DEXA examination.
Type of material: Doctoral Thesis
Publisher: University College Dublin. School of Medicine and Medical Science
Qualification Name: Ph.D.
Copyright (published version): 2012 the author
Keywords: 3D VisualisationImage ProcessingMRIQuantificationSegmentation
Other versions:
Language: en
Status of Item: Peer reviewed
Appears in Collections:Medicine Theses

Show full item record

Download(s) 50

checked on May 25, 2018

Google ScholarTM


This item is available under the Attribution-NonCommercial-NoDerivs 3.0 Ireland. No item may be reproduced for commercial purposes. For other possible restrictions on use please refer to the publisher's URL where this is made available, or to notes contained in the item itself. Other terms may apply.