Abstract:
Glaucoma is among one of the top two eye-related diseases which results in blindness. This disease lacks early symptoms and as a result of that the patients of glaucoma remain unaware about the disease until it enters the advanced stage. The diagnosis of glaucoma is usually done manually from the 2D fundus images and the decision about the glaucoma presence is made on the basis of the cup to disc ratio which is derived from the size of the optic cup and optic disc. This project aimed towards the development of automatic glaucoma diagnosis system by the use of image processing techniques and machine learning approaches. The ultimate goal of the project is to drive state of the art approach for accurate glaucoma detection which can facilitate the general population for glaucoma screening. The early detection, would contribute towards the reduction of blindness cases in the world. This project is based on software which is easy to use and require fundus image as an input and the result is the detection of glaucoma even at its earlier stages. The project is divided into different modules to make it extensible for the diagnosis of other retinal diseases such as diabetic retinopathy, and macular edema. For the extraction of the optic disc, image processing techniques are used which produce good quality segmentation results while for the extraction of optic cup supervised learning approach is used in which multiclass based bagging method of ensemble classifier is used for training and classification. The results of the classifier are refined by the image processing techniques to achieve the best fitting of the cup elliptical boundary. The proposed method is tested on publically available image databases and evaluated against previous methods presented in the literature. The methods achieved much better accuracy in segmenting optic cup and optic cup and in the detection of glaucoma.
