Fusion of Pixel and Texture Features to Detect Pathological Myopia Beng-Hai Lee 1 , Damon Wing Kee Wong 1 , Ngan Meng Tan 1 , Zhuo Zhang 1 , Joo Hwee Lim 1 , Huiqi Li 1 , , Jiang Liu 1 , Weimin Huang 1 , Seang Mei Saw 2 , Louis Hak Tien Tong 3 , Tien Yin Wong 3 1 Institute for Infocomm Research, A*STAR, Singapore 2 National University of Singapore 3 Singapore Eye Research Institute Abstract— Myopia is a growing concern in many societies. In extremely high myopia, pathological myopia, which can cause visual loss, can occur. Pathological myopia is also accompanied by various visually perceivable symptoms on the retina, such as peripapillary atrophy. PAMELA is an automatic system for the detection of pathological myopia through the presence of peripapillary atrophy. In this paper, we describe two modules in the PAMELA system based on texture analysis and gray level analysis. A decision engine is then used to fuse the two individual results to obtain an overall analysis. From the results run on a sample batch of images from the Singapore Eye Research Institute, a sensitivity of 0.9 and a specificity of 0.94 with a total accuracy of up to 92.5% is obtained. The promising results indicate good potential for further development of PAMELA as a tool for mass screening for the detection of pathological myopia. This electronic document is a “live” template. The various components of your paper [title, text, heads, etc.] are already defined on the style sheet, as illustrated by the portions given in this document. Keywords- Pathological myopia, computer aided diagnosis I. INTRODUCTION Myopia, or nearsightedness, is a huge public health problem with the economic and medical costs of myopia enormous and often under-estimated. In simple refractive myopia, the images are focused in a location in front of the retina, leading to defocused images when projected onto the retina, resulting in blurred vision. In cases where the degree of myopia is extremely high, there is a possibility of the occurrence of pathological myopia. Pathological myopia differs from simple refractive myopia, and is defined as myopia caused by pathologic axial elongation [1]. Also known as “degenerative myopia”, pathological myopia can be accompanied by straightened and stretched vessels, temporal peripapillary atrophic crescent, tilting of the optic disc, posterior staphyloma, lacquer cracks in the Bruch’s membrane, geographic areas of atrophy of the retinal pigment epithelium and choroids, subretinal haemorrhage, and choroidal neovascularisation[2]. Myopia-related visual impairment may affect the productivity, mobility, quality of life and activities of daily living of individuals. Potentially blinding pathologic myopia is often irreversible in nature, especially if diagnosed late. In the Shihpai Eye Study of elderly Taiwanese Chinese aged 65 years or older[2], myopic macular degeneration was the second leading contributing cause of visual impairment (12.5%) after cataract (41.7%). The risks of visual loss in myopia are sufficiently high to warrant measures to prevent pathologic myopia. The current methods for assessment of a patient for pathologic myopia are still largely reliant on manual efforts. This limits the applicability of such methods for screening efforts, though much desired. A fully automatic method for the assessment of retinal images for pathologic myopia has yet to be reported and yet very much in need, due to the large social impact of myopia in many societies today. To our knowledge no other systems for pathological myopia detection has been described. In this paper, we present a method for the detection of pathological myopia through the PAMELA (Pathological Myopia dEtection through peripapillary atrophy) system. The PAMELA system detects the presence of pathological myopia through several parallel approaches, with the final decision based on a weighted committee of votes. In Section 1, a brief description of pathological myopia has been performed. Section 2 describes the methodology employed within the PAMELA system. Following this, a description of an experiment carried out with a sample set of images is described, together with some discussion of the results, in Section 3. Finally, Section 4 presents the overall conclusions obtained in this paper. II. METHODOLOGY As previously described, there are several typical features for pathological myopia (PM), of which peripapillary atrophy (PPA) is a good indicator. PPA appears as temporal choroidal or scleral crescents or rings around the optic disc. The PAMELA system principally employs the presence of the PPA crescent to detect for the presence of pathologic myopia. An overview of the system framework is shown in Figure 1. An initial step in the PAMELA system is to first obtain an ROI from the retinal fundus image which localizes the optic disc. This helps to save computing resources due to the comparatively smaller area of the optic disc (10%) compared to the retinal fundus image. The ROI was selected through a grid-based intensity analysis of the retinal image which we Fengshou Yin 1