Title: Fine-tuning of pre-trained convolutional neural networks for diabetic retinopathy screening: a clinical study

Authors: Saboora M. Roshan; Ali Karsaz; Amir Hossein Vejdani; Yaser M. Roshan

Addresses: Khorasan Institute of Higher Education, No. 77, Moalem Blvd., Mashhad, Iran ' Khorasan Institute of Higher Education, No. 77, Moalem Blvd., Mashhad, Iran ' Navid-e-Didegan Clinic, Between 7&9 Mollasadra St., Ahmad Abad Blvd., Mashhad, Iran ' University of British Columbia, 6190 Agronomy Road, Suite 301, Vancouver, BC V6T 1Z3, Canada

Abstract: Diabetic retinopathy is a serious complication of diabetes, and if not controlled, may cause blindness. Automated screening of diabetic retinopathy helps physicians to diagnose and control the disease in early stages. In this paper, two case studies are proposed, each on a different dataset. Firstly, automatic screening of diabetic retinopathy utilising pre-trained convolutional neural networks was employed on the Kaggle dataset. The reason for using pre-trained networks is to save time and resources during training compared to fully training a convolutional neural network. The proposed networks were fine-tuned for the pre-processed dataset, and the selectable parameters of the fine-tuning approach were optimised. At the end, the performance of the fine-tuned network was evaluated using a clinical dataset comprising 101 images. The clinical dataset is completely independent from the fine-tuning dataset and is taken by a different device with different image quality and size.

Keywords: deep learning; convolutional neural network; diabetic retinopathy; inception model; clinical study.

DOI: 10.1504/IJCSE.2020.10028621

International Journal of Computational Science and Engineering, 2020 Vol.21 No.4, pp.564 - 573

Received: 05 May 2018
Accepted: 10 Apr 2019

Published online: 24 Apr 2020 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article