Neural network classifier for glaucoma diagnosis

Early glaucoma diagnosis can prevent the irreversible damage to the eye. Computer aided diagnosis can help clinical specialist to evaluate the available data and assign them to a specific pathology. The purpose of this research is to find a classifier for the glaucoma diagnosis based on an original set of eleven visual functional and structural parameters collected from Ocular Response Analyzer and Optical Coherence Tomography. Data from 122 healthy eyes and 118 glaucomatous eyes compose the classifier database. Few configurations of feedforward neural network classifiers were investigated. The optimal classifier proves to be one with two hidden layers, with 22 neurons on the first layer and 5 on the second one. The classifier sensitivity is 100% and the specificity is 94.3%.