Classification eye disease (PTERYGIUM) using comparative method of deep learning

Zulkiflee, Anwar Syaakirin (2022) Classification eye disease (PTERYGIUM) using comparative method of deep learning. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Classification eye disease (PTERYGIUM) using comparative method of deep learning.pdf - Submitted Version Download (6MB)

Abstract

Pterygium is a common eye disease that affects workers who are regularly exposed to UV rays and dusty environments, such as construction workers, welders, and engineers. The pterygium is not lethal, however it does cause blurred vision and pain for the sufferer. The typical pterygium screening test is time consuming and repetitive; hence, an automated pterygium detection based on deep learning was developed to speed up the screening process. The project proposed an enhancement technique towards the original dataset which utilized the contrast stretching, gamma correction and unsharp masking. The result image showed a much more detailed representation of a pterygium lesion in an image. The best configuration of gamma, 2 and unsharp masking with 'Radius', 0.25 and 'Amount', 2 showed the best performance model thus also stabilize the most with the model performance of accuracy 97 .45%, sensitivity 98. 71 % and standard deviation of 0.4. Finally, the proposed enhancement technique is capable of producing an outstanding result even with a limited dataset, as well as reducing the time required for the model to train while obtaining a stable classification result from the model.

Actions (login required)

View Item