Proposed PhD Project
Background:
Multiple sclerosis (MS), an immune-mediated, demyelinating disease of the central
nervous system, is associated with multiple pathological changes within the retina. Optical coherence
tomography (OCT) investigations have revealed a reduction in the thickness of retinal nerve fiber and
ganglionic cell layers, and studies based on fundus photography have shown impaired retinal blood
circulation among MS patients. However, these findings are not specific to MS, and can be observed in
other neurodegenerative and ocular disorders as well. This project provides the first study aimed to
investigate whether any MS-specific feature can be found within infrared reflectance scanning laser
ophthalmoscopy (IR-SLO) images, i.e., monochromatic fundus-like images often taken along with OCT.
Aim:
The retinal changes occurring in MS can be effectively tracked through the analysis of IR-SLO
images. The novel biomarkers of MS proposed in this study, primarily related to retinal vessels, could
serve as additional or even substitute diagnostic markers to those currently obtained from time-consuming,
costly, and invasive procedures.
Plan:
An age-matching subject-wise stratified k-fold cross-validation data splitting approach will be
applied for constructing the training, validation and test datasets, minimizing the risk of model
overfitting. Machine learning models will be applied to the resulting features obtained from the
anatomical structures of fundus for classifying MS. Finally, the contribution of each feature to the
prediction of the best-performing model is calculated using feature importance methods.
