AbstractRecently, researchers have reported the potential value of CMR texture, especially for evaluation of ischemic and non-ischemic cardiomyopathy. However, these studies were mostly based on more advanced CMR sequences, and focused on myocardial disease at a later stage. Just one paper reports the application of TA using conventional CINE CMR sequences, and only one paper mentions healthy population myocardial TA with a small cohort.
There are three main aims of this thesis. Firstly, to explore an ‘easy to practice’ and reliable CINE CMR 2D image TA workflow; secondly, to investigate healthy population myocardial TA characteristics; and finally to validate the significance of this technique through different myocardial disease groups.
TA intra- and inter-observer repeatability work involved n=60 healthy volunteers, and a total of n=15 (at end-diastole) and n=13 (at end-systole) texture variables showed a high repeatability. These variables then were put forward in healthy population (n=600) analysis, where major differences were observed between ED and ES, and between female and male. Minor differences were also detected between a younger age group (40-45 years) and a medial age group (55-63 years), (p﹤0.05). A whole myocardial wall ROI and an ROI just including the septal wall represented very consistent TA results, whilst ROIs at the lateral wall showed less sensitive TA changes.
In all the three different cardiac disease groups (n=61, 64 and 63 respectively), a good discriminate power was observed using logistic regression, with AUC values from 0.74 to 0.91 that were more sensitive on the septal wall, and less sensitive on the lateral wall, (p ﹤0.0001).
The data presented in this thesis confirmed good repeatability of texture analysis in CMR 2D CINE images. The work has highlighted the variability of texture parameter values in different cardiac phases, genders, age groups, and anatomical regions. It has validated the value of texture analysis in myocardial disease.
|Date of Award
|Graeme Houston (Supervisor) & Stephen Gandy (Supervisor)
- Texture Analysis