Ophthalmology (Eye & Ear Hospital) - Theses
Permanent URI for this collection
Search Results
1 - 10 of 72
-
ItemCorrecting co-localization error between infrared scanning laser ophthalmoscopy and optical coherence tomography using deep learning( 2023-12)Multimodal imaging offers unique insights into ophthalmic pathologies and is poised to become the preferred diagnostic and prognostic method for predictive artificial intelligence (AI) systems utilised by clinicians. Spatially attentive convolutional neural networks are emerging as the preferred approach for predictive systems using large, complex image datasets, and accurate co-localization of hybrid image data will be required to boost computational efficiency at scale. In the largest study of its kind to date (English language papers, literature review performed December 2023), this thesis characterizes in detail co-localization error between optical coherence tomography (OCT) and infrared scanning laser ophthalmoscopy (IR-SLO) imaging in real-world settings, highlighting the intrinsic and extrinsic factors contributing to error magnitude. An original computational pipeline for co-localization error correction is developed, using AI approaches based on a ground truth dataset consisting of manually corrected B-scan segments and their corresponding IR data. A novel, first of its kind methodology is developed and validated in order to extract the ground truth dataset. To improve AI performance, large-volume data simulations are utilised to explore AI ‘best practice’ in tackling this problem, including the model architecture, loss function, and optimal hyperparameters that are most robust in the context of variable, noisy infrared reflectance data. Learnings from these simulations are utilised to build a best-case model, which is tested on a large real-world dataset. The final deep learning algorithm developed significantly outperforms existing methods for correcting co-localization error between OCT and IR-SLO, establishing a foundation for future AI-driven solutions in this area. In the long-term, this work will potentially improve the diagnosis of AMD and other retinal conditions that rely on a combination of OCT and IR-SLO data.
-
ItemThe renin-angiotensin system and body weight regulation : influence on weight gain and body composition in rats(University of Melbourne, 2008)
-
ItemDevelopment of a bioengineered corneal surface replacement(University of Melbourne, 2006)
-
ItemDevelopment of a bioengineered corneal surface replacement(University of Melbourne, 2006)
-
ItemReticular Pseudodrusen and their Impact in Age-Related Macular Degeneration( 2023-09)Background: Age-related macular degeneration (AMD) is a leading cause of vision loss, with no treatments available to slow or stop the development of late, sight-threatening complications in the early stages of the disease. Reticular pseudodrusen (RPD), which are distinctive subretinal drusenoid deposits, have become increasingly appreciated as a potentially critical phenotype driving vision loss in AMD. There is thus a need to better understand RPD, and to develop tools to facilitate such work. Aims: To explore the association between RPD and the health and function of eyes with non-late AMD by characterising their relationship with photoreceptor function, and to better understand the association of RPD with late AMD development. Finally, to develop an automated method to segment RPD on optical coherence tomography (OCT) imaging. Methods: The association between RPD extent and location (determined on combined OCT and near-infrared reflectance imaging) and photoreceptor function, determined using mesopic fundus-controlled perimetry and dark adaptation chromatic perimetry, was evaluated in eyes with large drusen and no evidence of late AMD. The association between both RPD volume and imaging characteristics (derived using texture analysis), and AMD disease progression, was investigated in a cohort of individuals with bilateral large drusen followed over 3 years. Finally, a deep learning (DL) model was developed to segment RPD on OCT B-scans and its performance for this task was compared to four retinal specialists. The performance of this model to detect eyes with RPD was also evaluated in four external cohorts against two retinal specialists. Results: The increasing global, but not local, extent of RPD was significantly associated with local cone and rod photoreceptor dysfunction in eyes with large drusen, even after adjusting for other key confounders. No significant association was observed between increasing two- or three-dimensional RPD extent and an increased rate of developing late AMD. Five textural features of RPD were extracted to characterise the pattern and distribution of RPD deposits within an eye, and none of these features were found to be significantly associated with late AMD development. Finally, the DL segmentation model for RPD showed a comparable level of agreement with four retinal specialists as the level of inter-reader agreement. The DL model also showed comparable performance to two retinal specialists for the detection of RPD in 559 eyes from 503 participants in the four external cohorts. Conclusions: These findings show that there may be generalized pathogenic changes in eyes with RPD that account for the observed localised cone- and rod-mediated visual dysfunction that has not been previously appreciated. Further work to understand these potential pathogenic changes could help uncover the disease mechanisms underlying RPD. Future work to gain a further understanding of RPD could be facilitated by the DL model for RPD segmentation developed, which was shown to be robust and comparable with human experts. Uncovering the disease mechanisms driving the formation of RPD and how they drive vision loss in AMD could help identify new therapeutic targets for this phenotype and ultimately aid in preventing irreversible vision loss in AMD.
-
ItemPredicting progression in age-related macular degeneration( 2023-08)Background: Predicting which individuals will develop vision-threatening complications of age-related macular degeneration (AMD) is a challenging task, as traditional models based on colour fundus photography (CFP) correctly identify less than half of those who subsequently develop late AMD at 95% specificity. Thus, there is a great need to improve risk stratification for individuals with the early stages of AMD. Aims: To examine the prognostic significance of novel pathological characteristics related to drusen phenotypes and pigmentary abnormalities in the early stages of AMD. Methods: Retinal imaging data from 280 eyes from 140 individuals with bilateral large drusen enrolled in a longitudinal observational study was evaluated. Individuals underwent multimodal imaging (MMI) and microperimetry at baseline and then 6-monthly for up to 3-years. Disease progression was primarily evaluated based on the development of MMI-defined late AMD, and secondarily based on the rate of visual sensitivity decline on microperimetry prior to late AMD development. Four retinal specialists assessed the likelihood that each eye at baseline would progress with CFP, and then with MMI, to determine if MMI improves their ability to predict late AMD development. Baseline images were assessed for the: (i) presence of cuticular drusen, and extent of (ii) hyporeflective cores within drusen (HCD), (iii) hyperpigmentary abnormalities (HPAs), and (iv) hyperreflective foci (HRF) that do not spatially correspond to HPAs [HRF(OCT+/CFP-)]. The association with progression and impact on visual sensitivity of each feature was examined, including adjustments for well-established risk factors for progression (drusen volume from optical coherence tomography, presence of pigmentary abnormalities on CFP, and age). Results: The prediction of late AMD development by retinal specialists was improved when using MMI compared to CFP. However, a basic prediction model (age, presence of pigmentary abnormalities, and drusen volume) outperformed clinicians. In this cohort, neither the presence of cuticular drusen nor extent of HCD were significantly associated with an increased rate of progression to late AMD, reduced mean visual sensitivity at baseline, or an increased rate of visual sensitivity decline, after adjusting for well-established risk factors of progression. The quantification of HPA extent did not significantly improve the prediction of late AMD development compared to HPA presence, and the addition of HRF(OCT+/CFP-) extent to HPA extent also did not improve performance. Both HPA and HRF(OCT+/CFP-) extent were independently associated with reduced sector-based visual sensitivity, with the latter also associated with a significantly faster rate of visual sensitivity decline. Conclusions: Accounting for drusen phenotypes such as cuticular drusen and HCD, or the quantity of HPAs and HRF(OCT+/CFP-), did not significantly improve the prediction of late AMD development above what could be achieved by well-established risk factors. However, a basic prediction model using these parameters – drusen volume, presence of pigmentary abnormalities and age – outperformed retinal specialists, suggesting that such a model could improve counselling and monitoring of individuals in clinical practice. Such a model could also be used to better identify an enriched cohort to improve feasibility of future interventional trials, and thus help expedite the discovery of preventative treatments in the early stages of AMD.
-
ItemUsing genetic technologies to understand and treat retinal degeneration( 2023-04)Advances in genetic technologies, such as CRISPR/Cas9, have revolutionised the way we study and manipulate the genome. The first generation of CRISPR/Cas was used for genome editing; subsequent advances have led to modifications of CRISPR/Cas for additional uses, such as activation and repression of gene expression. This opens up the possibility to manipulate endogenous gene expression, demonstrating CRISPR as a useful tool to study gene functions and control cell identity. This thesis examines how advanced genetic technologies can be used to understand gene functions and develop therapies for retinal degenerative diseases, including retinitis pigmentosa (RP) and age-related macular degeneration (AMD). Chapter 1 provides a project overview and discusses the use of genetic technologies to study gene functions and develop treatments for retinal degenerative diseases. Part I of this thesis, which includes Chapter 2, 3, and 4, focuses on the development of a gene therapy for RP. In Chapter 2, I provided an overview of RP, its genetic contribution and recent therapeutic approaches using gene therapy, as well as an introduction to cellular reprogramming as an alternative method to treat RP. In Chapter 3, I described the development of a direct reprogramming technology to convert human Muller glia (MG) into induced rod photoreceptors (iRods) in vitro by activating the expression of selected transcription factors. Different combinations of these factors were tested on a human MG cell line, MIO-M1, and several combinations that promoted reprogramming of MG into iRods in vitro were identified. RT-qPCR and immunocytochemistry results demonstrated activation of the photoreceptor marker rhodopsin (RHO) in iRods. Also, multi-electrode array analysis showed that the iRods possessed functional electrophysiology. Finally, single-cell transcriptome analysis was performed to profile the iRods. The results highlighted that iRods expressed specific rod markers and found two different trajectories for iRod reprogramming. Subsequently, reprogramming of MG into iRods was tested in vivo by viral delivery of reprogramming factors into the P23H-3 rat model of autosomal dominant RP (Chapter 4). Treatment with reprogramming cocktails AAV-ANNr and AAV-Nr2P led to functional rescue, as well as localised changes to retinal morphology. Part II of this thesis, which includes two published journal articles (Chapter 5 and 6), explores how genetic technologies can be used to study functions of AMD-related genes. Chapter 5 serves as an introduction to Part II and discusses new technologies to understand functions of genes implicated in AMD. Chapter 6 explores the use of CRISPR technology to investigate a novel AMD-associated gene called POLDIP2. A POLDIP2 knockout human retinal pigment epithelial (RPE) cell line was generated using CRISPR/Cas, which displayed normal levels of cell proliferation, cell viability, phagocytosis and autophagy. RNA sequencing of the POLDIP2 knockout cell line highlighted changes in genes related to immune response, complement activation, vascular development and oxidative damage, which are biological processes relevant to AMD. This study also reveals a novel link between POLDIP2 and the mitochondrial superoxide dismutase SOD2, suggesting a potential role of POLDIP2 in oxidative stress regulation in AMD pathology. Taken together, these results contribute to our understanding of the genetic factors involved in retinal degenerative diseases and the advancement of novel therapeutic approaches to restore vision in diseased eyes.
-
ItemEmbracing the Future: An Examination of the Potential Role of Artificial Intelligence in Ophthalmology( 2023-01)Background: Substantial developments in artificial intelligence (AI) hold promise for screening and diagnosing ophthalmic diseases. However significant technological advances can be disruptive. Similarly, technology development and adoption do not necessarily occur hand-in-hand. The successful adoption of AI technology in eye health will necessitate education of healthcare professionals as well as appropriate organisational and sector support. Furthermore, it is necessary to make sure that eye health professionals are actively engaged in the implementation of AI applications in the clinical setting, to ensure that they are safe, effective, and used appropriately. Aim: This body of research aims to explore the implementation challenges of AI within ophthalmology and to inform the development of educational frameworks for ophthalmologists and trainees. The aims include: (1) Exploring knowledge and expectations of clinicians in ophthalmology and related medical specialties about AI. (2) Evaluating the impact of AI assistance on the grading performance of eye health professionals engaged in diabetic retinopathy screening. (3) Informing the development of an AI educational curriculum framework for ophthalmology trainees. Methods: A survey of ophthalmologists, dermatologists, radiologists/radiation oncologists and their trainees in Australia and New Zealand was carried out to understand perceptions of AI within these fields of medicine. Furthermore, the impact of a deep learning program as a tool used by optometrists, orthoptists and trainees for grading fundus photographs for diabetic retinopathy (DR) was assessed. Impacts on grading accuracy, confidence, and speed were evaluated. Finally the perspectives of ophthalmology trainees about AI education was explored using focus groups, to help guide the development of a framework educational AI curriculum for ophthalmology trainees. Results: Artificial intelligence was acknowledged by ophthalmologists, dermatologists and radiologists/radiation oncologists to represent a significant advance in healthcare technology that will have a broad-ranging positive influence. Reducing time spent on repetitive tasks and improving access to disease screening were seen as major potential benefits of the use of AI, with education identified as an important factor in the proper preparation of clinicians. In keeping with this, the study of AI assistance in identifying referable DR indicated that AI support was associated with increased accuracy, speed and confidence of optometrists, orthoptists, and their trainees, validating the potential utility of the technology. Furthermore, the focus group study demonstrated that ophthalmology trainees were keen for their College to include instruction on the clinical use of AI in their training curriculum and that trainees were interested in gaining a basic understanding of the technology and its potential implications. Conclusion: Artificial intelligence is perceived by ophthalmologists to likely have a significant and positive effect on their specialty. This perspective was reinforced by ophthalmology trainees who outlined how best to include an approach to AI in their educational curriculum. One AI adjunct was found to assist with DR screening and this may prove to have broader implications for the treatment of eye disease and easing workflow pressures. Preparing for the implementation of AI and understanding how it may affect clinicians plays an important role in underpinning the success of the technology within the field of ophthalmology.
-
ItemUsing Cellular Reprogramming and CRISPR Technologies for the Study of Neurodegeneration and the Development of Cell-Based Therapies( 2023-05)The premise of my research is that gene expression can be used to modify cell identity to develop models for the study and treatment of neurodegeneration. Neurons are essential for a broad set of body processes, thus damage to neural systems generally results in debilitating and irreversible diseases that place a huge burden on societies. Therefore, there is a pressing need for the development of effective approaches to generate neural tissue, which can be used to study neurodegenerative disorders, test potential therapies and to ultimately treat neurodegeneration. Regenerative technologies such as cellular reprogramming and induced pluripotent stem cells (iPSC) offer hope for the treatment of degenerative and incurable diseases. I used a combinatorial approach of cellular reprogramming technologies and genetic engineering tools to enhance the potential of regenerative approaches. In this regard, the simplicity, programmability and specificity of CRISPR technology provides an excellent tool to regulate gene expression, which we harnessed to improve classical methods of differentiation and disease modelling. This thesis includes studies using a diverse range of technologies with great potential for retinal biology: 1) The use of iPSCs to develop a fast protocol for neuronal differentiation, which can provide in vitro models to study neurodegeneration; 2) Using cellular reprogramming to convert Mueller glia into cone photoreceptors; 3) Gene therapy to treat photoreceptor degeneration in vivo using viral vectors; 4) Genetic engineering with CRISPR technology to regulate gene expression and study gene function in age-related macular degeneration pathophysiology and cell fate modulation through cellular reprogramming.
-
ItemMultimodal Retinal Imaging in Alzheimer's Disease( 2023-02)Background There are over 44 million people living with dementia worldwide, and most have Alzheimer's disease (AD). The current diagnostic methods for AD include brain imaging modalities such as magnetic resonance imaging and positron emission tomography which require significant resources, cost, and time. In contrast, retinal imaging modalities are much more widely available, accessible, and are lower in cost. The retina is an extension of the brain and they share a common embryological origin, thus highlighting the potential role for retinal imaging in the diagnosis of AD. Aims The aims of this thesis are: firstly, to perform a systematic review and meta-analysis to understand and evaluate the current evidence regarding retinal imaging in AD. Secondly, to perform a cross-sectional study of retinal imaging in people with biomarker-defined AD compared to healthy controls by exploring the association between these biomarkers and retinal parameters measured by optical coherence tomography (OCT), optical coherence tomography-angiography (OCT-A), and hyperspectral imaging. Thirdly, to explore the role of multimodal imaging in AD by combining findings from various retinal imaging methods to assess the potential of a composite biomarker of AD. Methods A systematic review of PubMed, EMBASE and Scopus was performed in accordance with PRISMA guidelines. Random-effects meta-analyses of standardised mean difference, correlation and diagnostic accuracy were conducted. The findings of this review were used to guide a cross-sectional study, to examine retinal parameters from OCT, OCT-A, and hyperspectral imaging of 35 people with AD and 38 healthy controls. Finally, a literature review on multimodal imaging was performed, and a novel multimodal model was generated using a machine learning approach. Results Meta-analysis of previous studies demonstrated that in people with AD there was a trend towards thinning in most retinal layers on OCT, increased foveal avascular zone area on OCT-A, and reduced arteriole and venule fractal dimension on fundus photography. The cross-sectional study demonstrated that, when compared to healthy controls, AD patients had no significant differences in retinal thickening in most retinal layers on OCT, significantly higher numbers of branching arterioles and venules on OCT infrared en face images, increased foveal avascular zone area and vessel density changes on OCT-A, and increased hyperspectral scores on hyperspectral imaging. The literature review of multimodal imaging outlined current terminology, approaches and challenges in the field. Finally, a novel multimodal model was generated and its clinical utility and limitations were discussed. Conclusion Prior research has identified associations between a number of retinal imaging parameters and AD, however limitations in study design including small sample sizes and non-biological definition of AD cases combined with heterogeneity in imaging methods and reporting make it difficult to determine the utility of these changes as AD biomarkers. Additional retinal imaging biomarkers were found to be associated with AD in our cross-sectional study of a biologically defined AD cohort and the role of combining retinal imaging parameters from multiple imaging modalities in a multimodal approach was explored.