Thesis
Stereo vision technologies for retinal imaging
- Creator
- Rights statement
- Awarding institution
- University of Strathclyde
- Date of award
- 2021
- Thesis identifier
- T15885
- Person Identifier (Local)
- 201655713
- Qualification Level
- Qualification Name
- Department, School or Faculty
- Abstract
- Early detection and treatment are crucial in limiting vision loss induced by primary open-angle glaucoma, a highly prevalent subtype of glaucoma, which is the second leading cause of blindness globally. Subjective assessment of the optic nerve head (ONH), which shows morphological change prior to the development of symptoms, is the cornerstone of early detection but suffers from poor intra- and inter-observer variability. Quantitative three-dimensional (3D) imaging technologies, which objectively quantify such change, have excellent repeatability and reproducibility, and have been shown to have higher diagnostic ability than general ophthalmologists. These could boost diagnostic ability in primary/community care if they were available at an affordable cost. In this work, two novel low-cost quantitative 3D imaging solutions, based on computer stereo vision, for 3D surface reconstruction of the ONH were developed. The first solution featured low-cost cameras and an indirect ophthalmoscopy lens fitted to a slit lamp. Later versions of the system also featured a projector, used to artificially add texture to the ONH to enhance stereo matching performance. Using eye phantoms, the systems were calibrated, and acquired images of ONH targets in the phantoms were processed using developed stereo vision pipelines to yield reconstructions. Such reconstructions were evaluated against ground truth data. Initial testing was also carried out with porcine eyes. Aiming for a miniaturised handheld system, the second solution featured two small cameras in an indirect ophthalmoscopy configuration, with a folded optical path provided by an appropriate set of mirrors, and a miniature projection system. This solution was tested in a simple eye phantom only. This work demonstrated ONH 3D reconstruction using both systems, with promising levels of precision and accuracy. Further work is needed, particularly with biological ONHs, to gauge their clinical potential. If sufficient, these systems have the potential to broaden access to 3D ONH imaging in future.
- Advisor / supervisor
- Giardini, Mario Ettore
- Resource Type
- Note
- This thesis was previously held under moratorium from 28th June 2021 until 28th June 2026.
- DOI
- Date Created
- 2021
- Former identifier
- 9912990093502996
Relations
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
PDF of thesis T15885 | 2021-07-15 | Public | Download |