Diabetic Retinopathy Detection Using Transfer and Reinforcement Learning with Effective Image Preprocessing and Data Augmentation Techniques

Maria Tariq, Vasile Palade, YingLiang Ma, Abdulrahman Altahhan

Research output: Chapter in Book/Report/Conference proceedingChapter (peer-reviewed)peer-review

6 Citations (Scopus)

Abstract

Diabetic retinopathy is the consequence of advanced stages of diabetes, which can ultimately lead to permanent blindness. An early detection of diabetic retinopathy is extremely important to avoid blindness and to recover from it as soon as possible. This chapter discusses the application of recent deep and transfer learning models for medical image analysis, with the focus on diabetic retinopathy detection. The chapter presents an extensive discussion on the publicly available datasets with diabetic retinopathy images, and the Kaggle dataset is used for training and testing of our proposed model. The main challenges to handle noisy and not large enough datasets are discussed in this chapter as well, where image preprocessing techniques and data augmentation play a significant role. An extensive overview of recent data augmentation techniques is also given to tackle the problem of imbalanced nature of diabetic retinopathy datasets. The proposed model integrates deep learning and reinforcement learning to perform detection and imbalanced classification on the Kaggle dataset.
Original languageEnglish
Title of host publicationFusion of Machine Learning Paradigms
EditorsIoannis K. Hatzilygeroudis, George A. Tsihrintzis, Lakhmi C. Jain
PublisherSpringer
Pages33-61
Number of pages29
Volume236
ISBN (Electronic)978-3-031-22371-6
ISBN (Print)978-3-031-22370-9
DOIs
Publication statusPublished - 7 Feb 2023

Publication series

NameIntelligent Systems Reference Library
PublisherSpringer
Volume236
ISSN (Print)1868-4394
ISSN (Electronic)1868-4408

Keywords

  • Deep learning
  • Diabetic retinopathy
  • Reinforcement learning
  • Transfer learning

Cite this