Segmentation scheme for a content-based retinal image retrieval system using electromagnetism like optimisation algorithm. (2016)

• Record Type: Journal Article
• Authors: Sivakamasundari, Jothi; Natarajan, Varadharajan
• Is Part Of: International journal of image mining. Volume 2:Number 2(2016)
• Page Start: 140
• Page End: 158
• Publication Date: 2016
• Subjects: content-based image retrieval -- diabetic retinopathy -- electromagnetism like optimisation algorithm -- multilevel thresholding -- retinal fundus image -- segmentation; Image processing -- Periodicals; Data mining -- Periodicals; 006.42
• Languages: English
• ISSNs: 2055-6039

Abstract

Content-based image retrieval (CBIR) system is used to assist physicians in the diagnosis of diabetic retinopathy (DR). Segmentation of retinal images to identify the anatomical landmarks and pathology bearing regions is a key process in early detection of DR. In this work, a new application of an electromagnetism like optimisation algorithm (EMOA) with Otsu multilevel thresholding is experimented on the normal and abnormal retinal images for the segmentation of blood vessels to identify DR. This procedure is used in the segmentation scheme of a content-based retinal image retrieval system and the performances are evaluated. This method provides comparatively better segmentation accuracy of 0.974 and 0.979 than other existing methods. The retrieval performance of EMOA segmentation-based CBIR system shows 97% of mean precision value. Hence, this CBIR system could be recommended in computer-assisted diagnosis for a better screening of the DR.