Authors: Mohammed El Amine Lazouni; Amel Feroui; Saïd Mahmoudi
Addresses: Biomedical Engineering Laboratory, Abou Bekr Belkaid Tlemcen University, Tlemcen, Algeria ' Biomedical Engineering Laboratory, Abou Bekr Belkaid Tlemcen University, Tlemcen, Algeria ' Computer Science Department, Faculty of Engineering, University of Mons, Belgium
Abstract: Glaucoma is a redundant disease and a major cause of blindness resulting from damage in the optic nerve. This disease generally spreads very slowly and does not show any symptom at the beginning. The research presented in this paper is both a clinical and a technological aid for diagnosis of early glaucoma based on four different artificial intelligence classification techniques, which are: multi-layer perceptron, support vector machine, K-nearest neighbour and decision tree. A majority vote system was applied to these techniques in order to optimize the performances of the proposed system. As far as the ratio cup to disc, which is one of the descriptors of the collected database, in this paper we detect automatically the cup by the k-means algorithm and the disc using mathematical morphology method. Moreover, we proposed a contour adjustment technique (Ellipse Fitting). The obtained results are satisfying, promising, and prove the efficiency and the coherence of our new database.
Keywords: glaucoma; classification; support vector machine; SVM; MLP; RBF; K-NN; majority voting; ReliefF; segmentation; LPE; K-means; ellipse fitting.
International Journal of Computer Aided Engineering and Technology, 2019 Vol.11 No.4/5, pp.613 - 633
Received: 03 Mar 2017
Accepted: 07 Sep 2017
Published online: 29 Mar 2019 *