Retinal Lesion Segmentation for Early Detection of Diabetic Retinopathy Using CNNs, Vision Transformers, and Hybrid Deep Learning Models

Ali, Qasim (2025) Retinal Lesion Segmentation for Early Detection of Diabetic Retinopathy Using CNNs, Vision Transformers, and Hybrid Deep Learning Models. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (702kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Detecting the presence of the eye conditions such as diabetic retinopathy which may lead to blindness in case it is not treated early, are some of the main processes that entail retinal lesion segmentation. This report discusses the ability of four deep learning basics, including SimpleSegNet, Vision Transformer (ViT) in scratch, SegFormer, and UNet to detect lesions in retinal fundus images. To train and test these models we deployed a dataset of eye images and their binary masks. We have trained each model five times and checked its performance according to such measures as accuracy, precision, and F1 score. We did also do some pictures to depict how well the models fitted the lesions than the real masks. As the result, models based on advanced design and already learned information, such as SegFormer and UNet, are more effective than the rest due to their in-advance learning. This would be useful to enable the doctors in Pakistan to diagnose the eye diseases more quickly, in zones where eye doctors are less in numbers. The models are compared in the report to come up with the best model to be used in this job.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Vamadevan, Arundev UNSPECIFIED
Uncontrolled Keywords:	Diabetic Retinopathy; Retinal lesion segmentation; SimpleSegNet; Vision Transformer; SegFormer; UNet; medical imaging; fundus images
Subjects:	R Medicine > RE Ophthalmology Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence R Medicine > Healthcare Industry
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	11 Aug 2026 15:14
Last Modified:	11 Aug 2026 15:14
URI:	https://norma.ncirl.ie/id/eprint/9495

Actions (login required)

View Item