Biomedical Signal Processing and Control 40 (2018) 366–377 Contents lists available at ScienceDirect Biomedical Signal Processing and Control jo ur nal homep age: www.elsevier.com/locate/bspc Research paper A new expert system based on fuzzy logic and image processing algorithms for early glaucoma diagnosis A. Soltani a,∗ , T. Battikh a , I. Jabri a , N. Lakhoua b a National Superior School of Engineering of Tunis, University of Tunis, Electrical Engineering Department b National Engineering School of Carthage, University of Carthage Tunis, Electrical Engineering Department a r t i c l e i n f o Article history: Received 2 March 2017 Received in revised form 23 July 2017 Accepted 15 October 2017 Keywords: Glaucoma diagnosis Decision-making system Ophthalmologic images Optic nerve head Image processing Fuzzy logic a b s t r a c t Decision-making systems based on images have increasingly become essential nowadays mostly in the medical field. Indeed, the image has become one of the most fundamental tools for both clinical research and sicknesses’ diagnosis. In this context, we treat glaucoma disease which can affect the optic nerve head (ONH), thus causing its destruction and leading to an irreversible vision loss. This paper presents a new glaucoma Fuzzy Expert System for early glaucoma diagnosis. Original ONH images are first pre- treated using appropriate filters to remove the noise. Canny detector algorithm is then used to detect the contours. Main parameters are then extracted, after having identified elliptical forms of both optic disc and excavation. This operation is performed by using Randomized Hough Transform. Finally, a classification algorithm, based on fuzzy logic approaches, is proposed to determine patients’ conditions. Our system is advantageous as far as it takes into consideration both instrumental parameters and risk factors (age, race, family history. . .) which make an important contribution to the valuable identification of cases suspected to have glaucoma. The proposed system is tested on a real dataset of ophthalmologic images of both normal and glauco- matous cases. Compared with other existing systems, the experimental results show the superiority of the proposed methods. The percentage of good predictions is more than 96%, reaching an improvement of 1–9% over earlier methods. © 2017 Elsevier Ltd. All rights reserved. 1. Introduction Regular monitoring and control of optic nerve’s head structure should become a part of routine clinical management of glaucoma since the early detection and treatment of such a retinal eye dis- ease is crucial to save the loss of vision. In fact, the patient may become blind if he does not receive any treatment [1]. The British Journal of Ophthalmology estimates that by 2020 about 79.6 mil- lion people worldwide will be diagnosed with glaucoma. However, being a chronic disease, glaucoma remains incurable. Indeed, as it is a “neurological” it affects nerve cells, which once destroyed, they cannot be replaced [2]. Therefore, its early diagnosis and treatment can prevent the vision loss [3]. The manual assessment of the optic nerve head via ophthalmoscopy or digital imaging, such as scan- ning laser tomography (SLT), scanning laser polarimetry (SLP) and optical coherence tomography (OCT), is recommended as clinical tools to detect glaucoma [4]. But, despite their benefit, they are not ∗ Corresponding author. E-mail address: amira soltani@yahoo.fr (A. Soltani). available to the great majority of physicians because of their cost [5]. The cup-to-disc ratio (CDR) and “Inferior, Superior, Nasal, and Temporal” (ISNT) rule are 2 key indicators to assess the ONH. The CDR is defined as the ratio of the diameter of the optic cup to the diameter of the optic disc. As glaucoma progresses, optic nerve fibers gradually disappear. Thus, the optic cup becomes larger with respect to the optic disc which increases the CDR [6]. In least developed countries such as Tunisia which is the subject of this study, the CDR is usually obtained via manual measurement by an ophthalmologist. However, observing ONH changes manually is a time-consuming process, and its accuracy varies according to the ophthalmologist’s experience. Since early detection of this disease is essential to prevent the permanent blindness, many efforts have been made on automatic detection of Glaucoma at an early stage. Bock et al. presented an automated glaucoma classification system that does not depend on the segmentation measure- ments [7]. They took a purely data-driven approach which is very useful in large-scale screening. The proposed algorithm under- takes a standard pattern recognition approach with a 2-stage https://doi.org/10.1016/j.bspc.2017.10.009 1746-8094/© 2017 Elsevier Ltd. All rights reserved.