Peer Reviewed

1

DOI

10.17077/omia.1056

Conference Location

Athens, Greece

Publication Date

October 2016

Abstract

Optic cup(OC) segmentation on color fundus image is essential for the calculation of cup-to-disk ratio and fundus morphological analysis, which are very important references in the diagnosis of glaucoma. In this paper we proposed an OC segmentation method using convolutional neural networks(CNNs) to learn from big size patch belong to each pixel. The segmentation result is achieved by classification of each pixel patch and postprocessing. With large pixel patch, the network could learn more global information around each pixel and make a better judgement during classification. We tested this method on public dataset Drishti-GS and achieved average F-Score of 93.73% and average overlapping error of 12.25%, which is better than state-of-the-art algorithms. This method could be used for fundus morphological analysis, and could also be employed to other medical image segmentation works which the boundary of the target area is fuzzy.

Rights

Copyright © 2016 the authors

Included in

Ophthalmology Commons

Share

COinS
 
Oct 21st, 12:00 AM Oct 21st, 12:00 AM

Optic Cup Segmentation Using Large Pixel Patch Based CNNs

Athens, Greece

Optic cup(OC) segmentation on color fundus image is essential for the calculation of cup-to-disk ratio and fundus morphological analysis, which are very important references in the diagnosis of glaucoma. In this paper we proposed an OC segmentation method using convolutional neural networks(CNNs) to learn from big size patch belong to each pixel. The segmentation result is achieved by classification of each pixel patch and postprocessing. With large pixel patch, the network could learn more global information around each pixel and make a better judgement during classification. We tested this method on public dataset Drishti-GS and achieved average F-Score of 93.73% and average overlapping error of 12.25%, which is better than state-of-the-art algorithms. This method could be used for fundus morphological analysis, and could also be employed to other medical image segmentation works which the boundary of the target area is fuzzy.