Eye Diseases Classification Using Hierarchical MultiLabel Artificial Neural Network

Several types of research adapted to classify or diagnosis eye diseases using different classification techniques with different feature extraction techniques as well as with different eye data sets. The goal of this paper is that utilizing dataset of images of digital camera to classify external eye diseases. This dataset is categorized into four major classes depends on the part of the eye where the disease causes the problem. These classes were further subdivided into several subclasses equal to the number of diseases in that class. Colour histogram features utilized from Hue, Min, Max, Diff (HMMD) colour space were quantized non-uniformly as defined in MPEG-7 standard into 32-bins in addition to Laws texture features utilized from Sum channel of HMMD colour space extracted to be classified. The contribution of this paper is that a new classification technique which is Hierarchical Multi-label Classification (HMC) is applied. The overall prediction accuracy achieved is %75.7142.