Transfer Learning-Based Model for Diabetic Retinopathy Diagnosis Using Retinal Images

Abstract

Diabetic retinopathy (DR) is a visual obstacle caused by diabetic disease, which forms because of long-standing diabetes mellitus, which damages the retinal blood vessels. This disease is considered one of the principal causes of sightlessness and accounts for more than 158 million cases all over the world. Since early detection and classification could diminish the visual impairment, it is significant to develop an automated DR diagnosis method. Although deep learning models provide automatic feature extraction and classification, training such models from scratch requires a larger annotated dataset. The availability of annotated training datasets is considered a core issue for implementing deep learning in the classification of medical images. The models based on transfer learning are widely adopted by the researchers to overcome annotated data insufficiency problems and computational overhead. In the proposed study, features are extracted from fundus images using the pre-trained network VGGNet and combined with the concept of transfer learning to improve classification performance. To deal with data insufficiency and unbalancing problems, we employed various data augmentation operations differently on each grade of DR. The results of the experiment indicate that the proposed framework (which is evaluated on the benchmark dataset) outperformed advanced methods in terms of accurateness. Our technique, in combination with handcrafted features, could be used to improve classification accuracy.

Development of preprocessing methods and revised EfficientNet for diabetic retinopathy detection

“…Thus, it is very important to recognize and treat DR in the early stage. Many recent studies have applied DL and image recognition technology to the recognition of DR 5–11 …”

“…Jabbar et al 10 proposed the different data augmentation operations on each grade of DR to overcome the data misbalancing issues. Then, they adopted the pre‐trained network VGGNet (VGG16) to extract the features from the fundus images and combined with the transfer learning strategy to improve the classification performance of DR.…”

“…20,21 Thus, it is very important to recognize and treat DR in the early stage. Many recent studies have applied DL and image recognition technology to the recognition of DR. [5][6][7][8][9][10][11] Khan et al 5 proposed the VGG16, spatial pyramid pooling layer (SPP), and network-in-network (NiN) (VGG-NiN) VGG-NiN with the lowest parameters (45 486 280 parameters) to speed up the training and model convergence. The VGG-NiN was trained and tested on a GPU (NVIDIA Tesla k40) and adopted the Kaggle dataset to demonstrate the performance of the proposed method.…”

“…The evolution of deep learning (DL) has made artificial intelligence (AI) image recognition a mature technology 4 . Many recent studies have applied DL and image recognition technology to the recognition of DR 5–11 . The results of the APTOS 2019 Blindness Detection study (https://www.kaggle.com/c/aptos2019-blindness-detection) revealed that the recent methods obtained an accuracy of 75%–85% for grading the DR level.…”

“…3 The evolution of deep learning (DL) has made artificial intelligence (AI) image recognition a mature technology. 4 Many recent studies have applied DL and image recognition technology to the recognition of DR. [5][6][7][8][9][10][11] The results of the APTOS 2019 Blindness Detection study (https://www.kaggle.com/c/aptos2019-blindness-detection) revealed that the recent methods obtained an accuracy of 75%-85% for grading the DR level. Fundus photographs need to be taken with a professional fundus camera and the quality of fundus images varies considerably because of inconsistent environments, equipment, and photographers.…”