Diabetic Retinopathy Diagnosis in Retinal Images Using Hopfield Neural Network. Issue 6 (1st November 2016)

• Record Type: Journal Article
• Authors: Hemanth, D. Jude; Anitha, J.; Indumathy, A.
• Is Part Of: IETE journal of research. Volume 62:Issue 6(2016)
• Page Start: 893
• Page End: 900
• Publication Date: 2016-11-01
• Subjects: Hopfield neural network -- Image processing -- Diabetic Retinopathy & Accuracy; Electronics -- Periodicals; Telecommunication -- Periodicals; Electronics; Telecommunication; Periodicals; 621.38
• DOI: 10.1080/03772063.2016.1221745
• Languages: English
• ISSNs: 0377-2063

Abstract

ABSTRACT: The main objective of this work is to develop an artificial neural network (ANN) for automated detection of diabetic retinopathy (DR) in retinal images. Three hundred and thirty images from DR category and two hundred and ten images from normal category are used in this work. These images are taken from patients with ages in the range of (58.2 18.7) years who were under treatment at the Lotus Eye Care Hospital, Coimbatore, India. A single-layer Hopfield neural network (HNN) is designed to categorize the normal images and the abnormal DR images. The ANN is trained with 165 images from the abnormal DR category and 105 images from the normal category. The testing images are 165 from DR and 105 from normal category respectively. Specific numerical features obtained from these images are used as inputs for the ANN. The performance measures used for the study are accuracy, sensitivity and specificity. It has been observed that the proposed approach has provided accuracy in the range of 97%–98%. Thus, the experimental results suggest the practical feasibility of the proposed approach. A comparative study with existing works for screening of abnormal retinal images is also reported in this work.