End-to-end diabetic retinopathy grading based on fundus fluorescein angiography images using deep learning

Gao, Zhiyuan; Jin, Kai; Yan, Yan; Liu, Xindi; Shi, Yan; Ge, Yanni; Pan, Xiangji; Lu, Yifei; Wu, Jian; Wang, Yao; Ye, Juan

doi:10.1007/s00417-021-05503-7

Cited by 24 publications

(13 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 35 Additionally, deep learning segmentation has been used for the quantification and staging of capillary non-perfusion 36 and for evaluation of FA images in DR patients. 37 , 38 Apart from segmentation of FA images for capillary network morphology, advances have also been performed in automated detection of vascular leakage. 39 …”

Section: Discussionmentioning

confidence: 99%

Assessment of a Novel Semi-Automated Algorithm for the Quantification of the Parafoveal Capillary Network

Kapsala¹,

Pallikaris²,

Tsilimbaris³

2023

OPTH

View full text Add to dashboard Cite

Introduction We present a novel semi-automated computerized method for the detection and quantification of parafoveal capillary network (PCN) in fluorescein angiography (FA) images. Material and Methods An algorithm detecting the superficial parafoveal capillary bed in high-resolution grayscale FA images and creating a one-pixel-wide PCN skeleton was developed using MatLab software. In addition to PCN detection, capillary density and branch point density in two circular areas centered on the center of the foveal avascular zone of 500μm and 750μm radius was calculated by the algorithm. Three consecutive FA images with distinguishable PCN from 56 eyes from 56 subjects were used for analysis. Both manual and semi-automated detection of the PCN and branch points was performed and compared. Three different intensity thresholds were used for the PCN detection to optimize the method defined as mean(I)+0.05*SD(I), mean(I) and mean(I)-0.05*SD(I), where I is the grayscale intensity of each image and SD the standard deviation. Limits of agreement (LoA), intraclass correlation coefficient (ICC) and Pearson’s correlation coefficient (r) were calculated. Results Using mean(I)-0.05*SD(I) as threshold the average difference in PCN density between semi-automated and manual method was 0.197 (0.316) deg −1 at 500μm radius and 0.409 (0.562) deg −1 at 750μm radius. The LoA were −0.421 to 0.817 and −0.693 to 1.510 deg −1 , respectively. The average difference of branch point density between semi-automated and manual method was zero for both areas; LoA were −0.001 to 0.002 and −0.001 to 0.001 branch points/degrees 2 , respectively. The other two intensity thresholds provided wider LoA for both metrics. The semi-automated algorithm showed great repeatability (ICC>0.91 in the 500μm radius and ICC>0.84 in the 750μm radius) for both metrics. Conclusion This semi-automated algorithm seems to provide readings in agreement with those of manual capillary tracing in FA. Larger prospective studies are needed to confirm the utility of the algorithm in clinical practice.

show abstract

Section: Discussionmentioning

confidence: 99%

Assessment of a Novel Semi-Automated Algorithm for the Quantification of the Parafoveal Capillary Network

Kapsala¹,

Pallikaris²,

Tsilimbaris³

2023

OPTH

View full text Add to dashboard Cite

show abstract

“…In this study, DL involves millions of trainable parameters to achieve the target task using objective functions. It aims to generate the diabetic retinopathy severity for retinal image data from patients, which do not require any manual annotations, rather than requiring large amounts of labeled data as previous time-consuming and cost-expensive DL systems have done ( 11 , 12 , 21 ). These works were developed from an early DL architecture of a convolutional neural network (CNN), which allows learning independent representations from a single image.…”

Section: Methodsmentioning

confidence: 99%

Diabetic Retinopathy Grading by Deep Graph Correlation Network on Retinal Images Without Manual Annotations

Zhang

Sun

Chen³

et al. 2022

Front. Med.

View full text Add to dashboard Cite

BackgroundDiabetic retinopathy, as a severe public health problem associated with vision loss, should be diagnosed early using an accurate screening tool. While many previous deep learning models have been proposed for this disease, they need sufficient professional annotation data to train the model, requiring more expensive and time-consuming screening skills.MethodThis study aims to economize manual power and proposes a deep graph correlation network (DGCN) to develop automated diabetic retinopathy grading without any professional annotations. DGCN involves the novel deep learning algorithm of a graph convolutional network to exploit inherent correlations from independent retinal image features learned by a convolutional neural network. Three designed loss functions of graph-center, pseudo-contrastive, and transformation-invariant constrain the optimisation and application of the DGCN model in an automated diabetic retinopathy grading task.ResultsTo evaluate the DGCN model, this study employed EyePACS-1 and Messidor-2 sets to perform grading results. It achieved an accuracy of 89.9% (91.8%), sensitivity of 88.2% (90.2%), and specificity of 91.3% (93.0%) on EyePACS-1 (Messidor-2) data set with a confidence index of 95% and commendable effectiveness on receiver operating characteristic (ROC) curve and t-SNE plots.ConclusionThe grading capability of this study is close to that of retina specialists, but superior to that of trained graders, which demonstrates that the proposed DGCN provides an innovative route for automated diabetic retinopathy grading and other computer-aided diagnostic systems.

show abstract

“…Using FFA images, Gao et al [ 180 ] graded DR by investigating three deep networks, i.e., VGG16, ResNet50, and DenseNet. VGG16 achieved the best performance, with an accuracy of 94.17%.…”

Section: The Role Of Ai In the Early Detection Diagnosis And Grading ...mentioning

confidence: 99%

The Role of Medical Image Modalities and AI in the Early Detection, Diagnosis and Grading of Retinal Diseases: A Survey

et al. 2022

View full text Add to dashboard Cite

Traditional dilated ophthalmoscopy can reveal diseases, such as age-related macular degeneration (AMD), diabetic retinopathy (DR), diabetic macular edema (DME), retinal tear, epiretinal membrane, macular hole, retinal detachment, retinitis pigmentosa, retinal vein occlusion (RVO), and retinal artery occlusion (RAO). Among these diseases, AMD and DR are the major causes of progressive vision loss, while the latter is recognized as a world-wide epidemic. Advances in retinal imaging have improved the diagnosis and management of DR and AMD. In this review article, we focus on the variable imaging modalities for accurate diagnosis, early detection, and staging of both AMD and DR. In addition, the role of artificial intelligence (AI) in providing automated detection, diagnosis, and staging of these diseases will be surveyed. Furthermore, current works are summarized and discussed. Finally, projected future trends are outlined. The work done on this survey indicates the effective role of AI in the early detection, diagnosis, and staging of DR and/or AMD. In the future, more AI solutions will be presented that hold promise for clinical applications.

show abstract

End-to-end diabetic retinopathy grading based on fundus fluorescein angiography images using deep learning

Cited by 24 publications

References 38 publications

Assessment of a Novel Semi-Automated Algorithm for the Quantification of the Parafoveal Capillary Network

Assessment of a Novel Semi-Automated Algorithm for the Quantification of the Parafoveal Capillary Network

Diabetic Retinopathy Grading by Deep Graph Correlation Network on Retinal Images Without Manual Annotations

The Role of Medical Image Modalities and AI in the Early Detection, Diagnosis and Grading of Retinal Diseases: A Survey

Contact Info

Product

Resources

About