期刊名称：International Journal of Computer Science and Network
印刷版ISSN：2277-5420
出版年度：2016
卷号：5
期号：4
页码：615-621
出版社：IJCSN publisher
摘要：Ophthalmic diseases like diabetic retinopathy,macular degeneration, glaucoma, etc. may cause gradual lossof eyesight and are some of the reasons behind blindness.Hence, blood vessel assessment and segmentation play a keyrole in the diagnosis of retinal disorders. The manualdetection of narrow blood vessels of retinal images is timeconsuming and may result in erroneous output. Therefore,computer aided, robust and performance oriented algorithmis required to detect and segment the blood vessels efficiently.The present work proposes an algorithm that usesthresholding technique, basic morphological operations andKirsch’s edge detection operator to detect the blood vesselsand segment the hard exudates efficiently. The detectedexudates regions are then compared with the ground truthexudate regions. Based on the correctly identified exudateregions between these two images, different performanceparameters like accuracy, specificity, sensitivity, PPV, PLRand misclassified proportions have been measured toevaluate overall performance of the proposed algorithm. Thisalgorithm successfully detects exudates with an average of98.47% accuracy, 54.67% sensitivity, 99.82% specificity,88.62% PPV, 303.73 PLR and 0.17% misclassifiedproportions.
