Diabetic retinopathy (DR) is a major complication of diabetes and a leading cause of blindness among the working-age population. However, the complex distribution and variability of lesion characteristics within the dataset present significant challenges for achieving high-precision classification of DR images.
Objective
We propose an automatic classification method for DR images, named DR-ConvNeXt, which aims to achieve accurate diagnosis of lesion types.
Methods
The method involves designing a dual-branch addition convolution structure and appropriately increasing the number of stacked ConvNeXt Block convolution layers. Additionally, a unique primary-auxiliary loss function is introduced, contributing to a significant enhancement in DR classification accuracy within the DR-ConvNeXt model.
Results
The model achieved an accuracy of 91.8%,sensitivity of 81.6%, and specificity of 97.9% on the APTOS dataset. On the Messidor-2 dataset, the model achieved an accuracy of 83.6%, sensitivity of 74.0%, and specificity of 94.6%.
Conclusions
The DR-ConvNeXt model's classification results on the two publicly available datasets illustrate the significant advantages in all evaluation indexes for DR classification.
Hoo-ChangSHolgerRRMingchenG, et al.Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning. Comput Res Reposit2016; 35: 1285–1298.
5.
SimonyanKZissermanA. Very deep convolutional networks for large-scale image recognition. International conference on learning representations, 2014, abs/1409.1556.
6.
kaimingHXiangyuZShaoqingR, et al.Deep residual learning for image recognition. Computer vision and pattern recognition, 2016, abs/1512.03385(1), pp. 770–778.
TanMLeQV. Efficientnet: rethinking model scaling for convolutional neural networks. arXiv: Learning2019; abs/1905.11946: 6105–6114.
9.
ZhuangLHanziMChao-YuanW, et al.A ConvNet for the 2020s. Computer vision and pattern recognition, 2022, pp. 11966–11976.
10.
ZhangHHeZ. Automatic cataract grading methods based on deep learning. Comp Methods Program Biomed2019; 182: 104978.
11.
FaragMMFouadMAbdel-HamidAT. Automatic severity classification of diabetic retinopathy based on DenseNet and convolutional block attention module. IEEE Access2022; 10: 38299–38308.
12.
MdRILwayFAMdN, et al.Applying supervised contrastive learning for the detection of diabetic retinopathy and its severity levels from fundus images. Comp Biol Med2022; 146: 105602.
13.
GuanghuiYYuanLTianweiZ, et al.Attention-Driven cascaded network for diabetic retinopathy grading from Fundus images. Biomed Signal Process Control2023; 80: 104370.
14.
XiangLYuchenJJiusiZ, et al.Lesion-attention pyramid network for diabetic retinopathy grading. Artif Intell Med2022; 126: 102259.
15.
HosseinSSinaRBehnamP, et al.Dual branch deep learning network for detection and stage grading of diabetic retinopathy. Biomed Signal Process Control2024; 93. doi: 10.1016/j.bspc.2024.106168
16.
PeimingZJieZQiaohongL, et al.Fundus image generation and classification of diabetic retinopathy based on convolutional neural network. Electronics2024; 13: 3603–3603.
17.
ZhikeHBinYShuiguangD, et al.Category weighted network and relation weighted label for diabetic retinopathy screening. Comput Biol Med2022; 152: 106408.
18.
HaiyanLXiaofangDWeiS, et al.Resampling-based cost loss attention network for explainable imbalanced diabetic retinopathy grading. Comput Biol Med2022; 149: 105970.
19.
de La TorreJVallsAPuigD. A deep learning interpretable classifier for diabetic retinopathy disease grading. Neurocomputing2020; 396: 465–476.
20.
DasPKPumrinS. Diabetic retinopathy classification: performance evaluation of pre-trained lightweight CNN using imbalance dataset. Eng J-Thailand2024; 28: 13–25.
MichaelDAJamesCFDennisPH, et al.Automated analysis of retinal images for detection of referable diabetic retinopathy. JAMA Ophthalmol2013; 131: 351.0–357.0.
24.
Van GrinsvenMJvan GinnekenBHoyngCB, et al.Fast convolutional neural network training using selective data sampling: application to hemorrhage detection in color fundus images. IEEE Trans Med Imaging2016; 35: 1273–1284.
25.
JoséIOElenaPMarianaDF, et al.Learning to detect red lesions in fundus photographs: an ensemble approach based on deep learning. Computer methods and programs in biomedicine2017; 153:115–127.
26.
KrizhevskyASutskeverIHintonGE. Imagenet classification with deep convolutional neural networks. Adv Neural Inform Process Syst2012; 25: 1106–1114.
27.
AlexeyDLucasBAlexanderK, et al.An image is worth 16(16 Words: transformers for image recognition at scale. International conference on learning representations, 2021.
28.
ChristianSWeiLYangqingJ, et al.Going deeper with convolutions. Computer vision and pattern recognition, 2015, abs/1409.4842(1), pp. 1–9.
29.
AndrewGHMenglongZBoC, et al.MobileNets: efficient convolutional neural networks for mobile vision applications. Comput Res Reposit2017: abs/1704.04861. doi: 10.48550/arxiv.1704.04861.
30.
BodapatiJDVeeranjaneyuluNShareefSN, et al.Blended multi-modal deep ConvNet features for diabetic retinopathy severity prediction. Electronics2020; 9: 914–914.
31.
SaraHKPeymanHKRezaK, et al.Diabetic retinopathy classification using a modified Xception architecture. Int Symp Signal Process Inform Technol2019: 1–6. doi: 10.1109/ISSPIT47144.2019.9001846.
32.
ShaikNSCherukuriTK. Hinge attention network: a joint model for diabetic retinopathy severity grading. Appl Intell2022; 52: 15105–15121.
33.
MeilingFJingyiWKaiW, et al.Grading of diabetic retinopathy images based on graph neural network. IEEE Access2023; 11: 98391–98401.
34.
ShanthiTSabeenianRS. Modified Alexnet architecture for classification of diabetic retinopathy images. Comp Electr Eng2019; 76: 56–64.
