Decision Support System for Age-Related Macular Degeneration Using Convolutional Neural Networks

Langarizadeh, Mostafa and Maghsoudi, Banafshe and Nilforushan, Naveed (2017) Decision Support System for Age-Related Macular Degeneration Using Convolutional Neural Networks. Iranian Journal of Medical Physics, 14 (3). pp. 141-148.

[img]
Preview
Text
maghsoudi1.pdf

Download (1MB) | Preview
Official URL: http://ijmp.mums.ac.ir/article_8531.html

Abstract

Introduction: Age-related macular degeneration (AMD) is one of the major causes of visual loss among the elderly. It causes degeneration of cells in the macula. Early diagnosis can be helpful in preventing blindness. Drusen are the initial symptoms of AMD. Since drusen have a wide variety, locating them in screening images is difficult and time-consuming. An automated digital fundus photography-based screening system help overcome such drawbacks. The main objective of this study was to suggest a novel method to classify AMD and normal retinal fundus images. Materials and Methods: The suggested system was developed using convolutional neural networks. Several methods were adopted for increasing data such as horizontal reflection, random crop, as well as transfer and combination of such methods. The suggested system was evaluated using images obtained from STARE database and a local dataset. Results: The local dataset contained 3195 images (2070 images of AMD suspects and 1125 images of healthy retina) and the STARE dataset comprised of 201 images (105 images of AMD suspects and 96 images of healthy retina). According to the results, the accuracies of the local and standard datasets were 0.95 and 0.81, respectively. Conclusion: Diagnosis and screening of AMD is a time-consuming task for specialists. To overcome this limitation, we attempted to design an intelligent decision support system for the diagnosis of AMD fundus using retina images. The proposed system is an important step toward providing a reliable tool for supervising patients. Early diagnosis of AMD can lead to timely access to treatment.

Item Type: Article
Subjects: WN Radiology . Diagnostic Imaging
Divisions: Journals > Iranian J Medical Physics
Depositing User: ijmp ijmp
Date Deposited: 21 Jan 2018 15:24
Last Modified: 21 Jan 2018 15:24
URI: http://eprints.mums.ac.ir/id/eprint/8994

Actions (login required)

View Item View Item