City Research Online

Abstract

Glaucoma is a leading cause of visual disability across the world and when diagnosed the glaucoma patient will spend the rest of their life receiving treatment in managed clinical care. In the glaucoma clinic, retinal and optic nerve head (ONH) imaging can be used to help the clinician to manage patient treatment appropriately. By providing high resolution images of the optic nerve head structures and identifying changes therein related to disease onset and progression, an objective measure can be obtained as to how well or badly treatment is preventing further disease damage. This thesis contributes to the field of glaucoma progression detection by the analysis of clinical imaging data using confocal scanning laser tomography (CSLT). Primarily it is an investigation of how best to appraise and optimise current algorithms which aim to detect these glaucomatous structural changes in the optic nerve head. This is done by addressing how the performance of these methods can be best assessed in the absence of a gold standard for glaucomatous structural progression.

Glaucoma expert assessment of photographs of the optic disc is the current clinical standard of assessing glaucomatous damage evident in the ONH. This is used in this thesis to act as a reference standard by which these algorithms can be compared. In addition, the statistical principles underpinning trend detection techniques are also investigated along with the performance of these techniques to detect trends in CSLT data in the presence of different types of measurement noise and image quality. A new computer model is developed and validated to simulate stable series of CSLT images, with realistic variability, which can be used to benchmark the false-positive rates of current and future progression algorithms. In conclusion, the main results reported in this thesis show that uncertainties involved in expert assessment of change in ONH photographs limits this as a reference standard for structural change in glaucoma. In addition, since stability in clinical datasets is uncertain, simulation using modelled series is shown to provide a new benchmark for comparing methods of progression detection.

Publication Type:	Thesis (Doctoral)
Additional Information:	© 2011 Neil O'Leary
Subjects:	R Medicine > RE Ophthalmology
Departments:	Doctoral Theses School of Health & Psychological Sciences > Optometry & Visual Sciences School of Health & Psychological Sciences > School of Health & Psychological Sciences Doctoral Theses

Export

Downloads