Author ORCID Identifier

https://orcid.org/0009-0001-3350-1819

Date of Award

Summer 8-8-2023

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Rolando Estrada

Second Advisor

Jingyu Liu

Third Advisor

Rajshekhar Sunderraman

Abstract

Assessment of retinal fundus image is very informative and preventive in early ocular disease detection. This non-invasive assessment of fundus images also helps in the early diagnosis of vascular diseases. This unique combination help in the early diagnosis of diseases. Applying image enhancement techniques with advanced Deep learning techniques helps to overcome such a challenging problem. Most Deep learning models give a diagnosis without attention to underlying pathological abnormalities. In this thesis, we tried to solve the problem in the same way as ophthalmologists and experts in the field approach the problem. We created models that can detect an Optic disc, Optic cup, and vascular regions in the image. This work can be integrated into any ocular disease detection, such as glaucoma, and vascular disease detection, such as diabetes. Extensive work is applied for better sampling when all models were suffering from a lack of data in the medical imaging field. The entire work on the retinal fundus image was in 2d images. In the extension of this work, we applied our knowledge to 3d MRI-Brain images. We attempt to predict attention scores in children, which is a big factor in the detection of kids with ADHD. But both work on fundus images and brain MRI images are under the umbrella of medical imaging. We believe this advancement in this line of research can be very valuable for future researchers in the area of automated medical imaging, especially in automated retinal disease diagnosis.

DOI

https://doi.org/10.57709/35838868

File Upload Confirmation

1

Share

COinS