Title: Robust feature selection algorithm based on transductive SVM wrapper and genetic algorithm: application on computer-aided glaucoma classification

Authors: Nawel Zemmal; Nabiha Azizi; Mokhtar Sellami; Djamel Zenakhra; Soraya Cheriguene; Nilanjan Dey; Amira S. Ashour

Addresses: Labged Laboratory, Computer Science Department, Badji Mokhtar University, Annaba 23000, Algeria ' Labged Laboratory, Computer Science Department, Badji Mokhtar University, Annaba 23000, Algeria ' Labged Laboratory, Computer Science Department, Badji Mokhtar University, Annaba 23000, Algeria ' Labged Laboratory, Computer Science Department, Badji Mokhtar University, Annaba 23000, Algeria ' Labged Laboratory, Computer Science Department, Badji Mokhtar University, Annaba 23000, Algeria ' Department of Information Technology, Techno India College of Technology, Rajarhat, Kolkata 700156, India ' Faculty of Engineering, Department of Electronics and Electrical Communications Engineering, Tanta University, Tanta, 31527, Egypt

Abstract: Glaucoma has become a devastating disease after cataract to cause blindness. Thus, early diagnoses for glaucoma can prevent the vision loss. Computer-aided diagnosis (CAD) systems, which automate the process of ocular disease detection, are urgently needed to alleviate the burden on the clinicians. In the current work, advanced machine-learning algorithms are investigated to propose a robust system for Glaucoma diagnosis based on retinal images. Three features extraction methods, namely the grey-level cooccurrence matrix (GLCM), Hu moments and central moments are combined to form the entry feature vector. To select the most relevant features and taking into account at the same time the unlabelled data existing in medical databases, a new scheme of feature selection algorithm is proposed. It is based on transductive support vector machine (SVM) wrapper and genetic algorithm, which automatically detect and classify the glaucoma disease using fundus images. The effectiveness of the proposed GA-TSVM is evaluated on a public retinal database RIM-ONE using the classification accuracy, sensitivity and specificity metrics. The experimental results established that with 16% of labelled data, the proposed system could easily distinguish between the normal and the affected glaucoma cases.

Keywords: glaucoma disease; computer-aided diagnosis; CAD; genetic algorithm; feature extraction and selection; SSL; semi-supervised learning; TSVM; transductive support vector machine.

DOI: 10.1504/IJISTA.2018.094018

International Journal of Intelligent Systems Technologies and Applications, 2018 Vol.17 No.3, pp.310 - 346

Available online: 10 Aug 2018 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article