Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Evaluation of the potentials for optical coherence tomography (OCT) to detect early signs of retinal neurodegeneration

Kulmaganbetov, Mukhit ORCID: https://orcid.org/0000-0003-2112-4208 2021. Evaluation of the potentials for optical coherence tomography (OCT) to detect early signs of retinal neurodegeneration. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of Thesis_of_Mukhit_Kulmaganbetov.pdf]
Preview
PDF - Accepted Post-Print Version
Download (6MB) | Preview
[thumbnail of Cardiff University Electronic Publication Form] PDF (Cardiff University Electronic Publication Form) - Supplemental Material
Restricted to Repository staff only

Download (180kB)

Abstract

Among neuroretinal degenerations, glaucoma and age-related macular degeneration (AMD) have become the most frequent reasons for irreversible blindness globally. Among the causes of the elderly and senile dementia, Alzheimer’s disease (AD) has the leading position, the early ocular symptoms of which can potentially be a prognostic factor. The aim of this thesis was the early in vivo ligand-free detection of degenerative changes in the inner and outer retinal layers, which was possible using high-resolution optical coherence tomography (OCT) with the machine learning (ML) algorithms: support vector machine (SVM) and principal component analysis (PCA). Prior to the application of SVM and PCA for the classification of human OCT images, evaluation of the classifiers was performed in the classification of optical phantoms, the accuracy of which was in the range of 82-100%. This was the first attempt to measure the textural properties of various polystyrene and silica beads optical phantoms. To identify optical changes that characterise early apoptosis, OCT imaging of axotomised retinal ganglion cells (RGCs) in ex vivo retinal murine explants was performed. Substantial optical alterations in RGC dendrites in the early stages of apoptosis (up to 2 hours) were detected. ML algorithms correctly classified the retinal texture of the inner plexiform layer (IPL) of transgenic AD mice in all cases, indicating the potential for further investigation in in vivo animal and human studies. Not only the optical signature but also the transparency of the dissected murine retinal explants was investigated. Moreover, ML classification of 3xTg mice IPL layer was studied in terms of optical changes due to the RGD dendritic atrophy. ML classifiers’ accuracy in the detection of early and neovascular AMD was 93-100% for the texture of retinal pigment epithelium, 69-67% for the outer nuclear layer, 70% for the inner segment and 60-90% for the outer segment of photoreceptors. Classification of AMD stages and comparison with the age-matched healthy controls was carried out in the outer retina and RPE. Grey-level co-occurrence, run-length matrices, local binary patterns features were extracted from the IPL of the macula to classify glaucoma OCT images. The accuracy of linear and non-linear SVMs, linear and quadratic discriminant analyses, decision tree and logistic regression was between 55-70%. Based on the classifiers’ precision, recall and F1-score, Gaussian SVM outperformed other ML techniques. In this study, the observation of early glaucomatous subtle optical changes of human IPL was conducted. Also, the significance of various supervised ML algorithms was investigated. Understanding the optical signature of cumulative inherent speckle of OCT scans arising from apoptotic retinal ganglion cells and photoreceptors may provide vital information for the prevention of retinal neurodegeneration.

Item Type: Thesis (PhD)
Date Type: Submission
Status: Unpublished
Schools: Optometry and Vision Sciences
Subjects: R Medicine > RE Ophthalmology
Uncontrolled Keywords: optical coherence tomography, retina, neuroretinal degeneration, age-related macular degeneration, Alzheimer’s disease, machine learning, support vector machine, principal component analysis
Date of First Compliant Deposit: 7 October 2021
Last Modified: 10 Dec 2022 02:55
URI: https://orca.cardiff.ac.uk/id/eprint/144577

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics