Show simple item record

AuthorAlSaad R.
AuthorAl-Maadeed S.
AuthorAl Mamun M.A.
AuthorBoughorbel S.
Available date2020-03-29T12:15:10Z
Publication Date2018
Publication NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ResourceScopus
ISSN3029743
URIhttp://dx.doi.org/10.1007/978-3-319-96136-1_6
URIhttp://hdl.handle.net/10576/13542
AbstractAutomated Diabetic Retinopathy (DR) screening methods with high accuracy have the strong potential to assist doctors in evaluating more patients and quickly routing those who need help to a specialist. In this work, we used Deep Convolutional Neural Network architecture to diagnosing DR from digital fundus images and accurately classifying its severity. We train this network using a graphics processor unit (GPU) on the publicly available Kaggle dataset. We used Theano, Lasagne, and cuDNN libraries on two Amazon EC2 p2.xlarge instances and demonstrated impressive results, particularly for a high-level classification task. On the dataset of 30,262 training images and 4864 testing images, our model achieves an accuracy of 72%. Our experimental results showed that increasing the batch size does not necessarily speed up the convergence of the gradient computations. Also, it demonstrated that the number and size of fully connected layers do not have a significant impact on the performance of the model.
SponsorThis work was supported by Sidra Medicine (authors RA and SB), as well as a grant from the Qatar National Research Fund through National Priority Research Program (NPRP) No. 6-249-1-053 (authors SA and MA).
Languageen
PublisherSpringer Verlag
SubjectConvolutional Neural Networks
Deep learning
Diabetic retinopathy
Medical imaging
TitleA Deep Learning Based Automatic Severity Detector for Diabetic Retinopathy
TypeConference Paper
Pagination64-76
Volume Number10934 LNAI


Files in this item

FilesSizeFormatView

There are no files associated with this item.

This item appears in the following Collection(s)

Show simple item record