Abstract:
Advances in the area of computer science have a tremendous impact on the interpretation of
medical images. It enhanced the interpretation of medical images, and contributed to early
diagnosis. The typical architecture of a Computer Aided Diagnosis (CAD) system includes image
pre-processing, definition of region of interest, features mining and selection, and classification.
By the advancement and progress in digital imaging technology, one can easily capture the
retinal image using fundus camera which provides vital information about the sensory part of the
human eye which helps to diagnose eye disease. Glaucoma has been recognized as a major
cause of blindness. Early detection and treatment of glaucoma is the key to preventing
permanent vision loss. In this thesis, we have designed and implemented a system for the
automatic detection of glaucoma. First step of proposed system is acquiring the retinal image
database. Second step is the preprocessing of images to remove noisy area to lower the
processing time. After preprocessing, next step is to extract the features. Four features are used
to classify glaucoma, which includes Cup to Disc Ratio (CDR), Neuroretinal Rim ratio in
Inferior, Superior, Nasal and Temporal (ISNT) quadrants, ratio of blood vessels in ISNT
quadrants and displacement of blood vessels from center of optic disc. The first key feature cup
to disc ratio is calculated by two methods which are compared with clinical cup to disc ratios.
i.e. K-Means Clustering and Mean threshold morphological analysis (Proposed Method).
The developed method is tested on 50 images taken from three different publicly available
databases to extract all features. Three parameters are used to check the validity of proposed
algorithms i.e. visual inspection, accuracy, and computational time. We have achieved an
average accuracy of 94%. The mean computational time calculated for the proposed method is
5.37 seconds.
