Development Of An Android-Based Eye Disease Diagnosis System

ABSTRACT

The human eye is a vital organ of vision which gives the sense of sight. However, there are many diseases that affect the human eye and can lead to partial or total blindness. The major causes of blindness include cataract, glaucoma, diabetic retinopathy and several other corneal and retinal infections but most of them can be prevented with an early diagnosis. Therefore, there is the need for a cheaper, easy-to-understand and readily available system to aid the diagnostic process. This project aims to develop a diagnostic system for selected eye diseases on the Android mobile phone platform.

The project was designed with four modules namely: image acquisition, image pre-processing, feature extraction and classification. Image acquisition was achieved through publicly available online fundus databases. The pre-processing techniques used included conversion to grayscale, histogram equalization and thresholding. Feature extraction and classification was done using the BRISK algorithm. This project was implemented to detect Glaucoma and Diabetic Retinopathy using the Android platform. Android studio was the platfonn of choice for the development as it had better preinstalled features compared to other development environments. To achieve real-time processing while minimizing processor requirements, Intel's OpenCV (Open Source Computer Vision) library was used for the computer vision operations. The performance analysis results show that the classifier attained an accuracy of 86.7% for healthy images, 100% for diabetic retinopathic images and 100% for glaucomic images. The average classification accuracy of the developed system was 95.6% across the healthy and diseased classes.

The developed Android based eye disease diagnosis system can greatly assist in the early detection of

eye related diseases especially in areas with limited health facilities. To improve the performance of the system, a handheld ophthalmoscope should be used for image acquisition, training and testing. This would greatly increase it perfonnance with real life samples. Also, it could be expanded to be able to diagnose other retinal diseases.