Automatic Severity Classification of Diabetic Retinopathy Based on DenseNet and Convolutional Block Attention Module

Abstract

Diabetic Retinopathy (DR) -a complication developed as a result of heightened blood glucose levels-is deemed as one of the most sight-threatening diseases. Unfortunately, DR screening is manually acquired by an ophthalmologist, a process that can be considered erroneous and timeconsuming. Accordingly, in recent years, automated DR diagnostics have become a focus of research due to the tremendous increase in diabetic patients. Moreover, the recent accomplishments demonstrated by Convolutional Neural Networks (CNN) settle them as state-of-the-art for DR stage identification. This paper proposes a new automatic deep-learning-based approach for severity detection by utilizing a single Color Fundus photograph (CFP). The proposed technique employs DenseNet169's encoder to construct a visual embedding. Furthermore, Convolutional Block Attention Module (CBAM) is introduced on top of the encoder to reinforce its discriminative power. The model is trained using cross-entropy loss on the Kaggle Asia Pacific Tele-Ophthalmology Society's (APTOS) dataset. On the binary classification task, we accomplished (97% accuracy -97% sensitivity -98.3% specificity -0.9455, Quadratic Weighted Kappa score (QWK)) compared to the state-of-the-art. Moreover, Our network showed high competency (82% accuracy -0.888 (QWK)) for severity grading. The major contribution of the proposed framework is that it efficiently grades the severity level of diabetic retinopathy, while reducing the time and space complexity required, which demonstrates it as a promising candidate for autonomous diagnosis.

RDB-CSCA: A deep learning method to identify the spectral information for tracing the egg quality

“…Compared with the CNN decision-making network, the classification accuracy is increased by 1.00%, the precision rate is increased by 2.45%, the recall rate is increased by 0.90%, and the F1 score is increased by 1.67%. Further, to prove the effectiveness of CSCA, this paper compares the effects of different attention mechanisms of squeeze-and-excitation (SE) (Hu et al, 2020), efficient-channel-attention (ECA) (Wang et al, 2020), and convolutional-block-attention-module (CBAM) (Farag et al, 2022) on the decision performance. The embedding method and decision framework of the attention mechanism are consistent with CSCA.…”

Diabetic Retinopathy Detection from Fundus Images of the Eye Using Hybrid Deep Learning Features

“…The maximum average accuracy achieved in this paper using hybrid features vector and SVM classifier is 97.80% for binary classification and 89.29% for multiclass classification. The closest average accuracy for binary classification is reported in [ 29 ] at 97.00%, whereas for multiclass classification, the closest is reported in the article [ 23 ] with a value of 82.18%. When performing classification using the features extracted from GoogleNet only (1000 features), the MDR class provides better performance compared to the ResNet-18 model.…”

“…The authors of [ 29 ] propose a deep-learning-based model that uses the DenseNet encoder and convolutional attention module block for DR severity detection. The encoder is used to extract the features from the input fundus images from the APTOS dataset and the attention block is used for refining the features.…”

A Deep Learning Based Approach for Grading of Diabetic Retinopathy Using Large Fundus Image Dataset

“…Various classifiers, i.e., Random Forest, K-Nearest Neighbour, and Support Vector Machines, were used to test the accuracy patterns on Messidor and IDRiD [ 21 ], resulting in an accuracy of 0.975 and 0.955, respectively. Mohamed et al [ 22 ] devised an algorithm for diabetic retinopathy classification using DenseNet-169 as a baseline model followed by CNN to enhance the classification capability of the model. The model was trained and tested on the APTOS dataset, yielding an accuracy of 97% for binary classification, i.e., healthy or Diabetic retinopathy.…”