A Deep Learning Based Automatic Severity Detector for Diabetic Retinopathy

Rawan T. Al Saad, Somaya Al-Maadeed, Md Abdullah Al Mamun, Sabri Boughorbel

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Automated 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.

Original languageEnglish
Title of host publicationMachine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings
PublisherSpringer Verlag
Pages64-76
Number of pages13
ISBN (Print)9783319961354
DOIs
Publication statusPublished - 1 Jan 2018
Event14th International Conference on Machine Learning and Data Mining in Pattern Recognition, MLDM 2018 - New York, United States
Duration: 15 Jul 201819 Jul 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10934 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other14th International Conference on Machine Learning and Data Mining in Pattern Recognition, MLDM 2018
CountryUnited States
CityNew York
Period15/7/1819/7/18

    Fingerprint

Keywords

  • Convolutional Neural Networks
  • Deep learning
  • Diabetic retinopathy
  • Medical imaging

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Al Saad, R. T., Al-Maadeed, S., Al Mamun, M. A., & Boughorbel, S. (2018). A Deep Learning Based Automatic Severity Detector for Diabetic Retinopathy. In Machine Learning and Data Mining in Pattern Recognition - 14th International Conference, MLDM 2018, Proceedings (pp. 64-76). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10934 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-319-96136-1_6