Search

Purpose To determine whether eyes with pathologic myopia can be identified and whether each type of myopic maculopathy lesion on fundus photographs can be diagnosed by deep learning (DL) algorithms. Design A DL algorithm was developed to recognize myopic maculopathy features and to categorize the myopic maculopathy automatically. Participants We examined 7020 fundus images from 4432 highly myopic eyes obtained from the Advanced Clinical Center for Myopia. Methods Deep learning (DL) algorithms were developed to recognize the key features of myopic maculopathy with 5176 fundus images. These algorithms were also used to develop a Meta-analysis for Pathologic Myopia (META-PM) study categorizing system (CS) by adding a specific processing layer. Models and the system were evaluated by 1844 fundus image. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were used to determine the performance of each DL algorithm. The rate of correct predictions was used to determine the performance of the META-PM study CS. Main Outcome Measures Four trained DL models were able to recognize the lesions of myopic maculopathy accurately with high sensitivity and specificity. The META-PM study CS also showed a high accuracy and was qualified to be used in a semiautomated way during screening for myopic maculopathy in highly myopic eyes. Results The sensitivity of the DL models was 84.44% for diffuse atrophy, 87.22% for patchy atrophy, 85.10% for macular atrophy, and 37.07% for choroidal neovascularization, and the AUC values were 0.970, 0.978, 0.982, and 0.881, respectively. The rate of total correct predictions from the META-PM study CS was 87.53%, with rates of 90.18%, 95.28%, 97.50%, and 91.14%, respectively, for each type of lesion. The META-PM study CS showed an overall rate of 92.08% in detecting pathologic myopia correctly, which was defined as having myopic maculopathy equal to or more serious than diffuse atrophy. Conclusions The novel DL models and system can achieve high sensitivity and specificity in identifying the different types of lesions of myopic maculopathy. These results will assist in the screening for pathologic myopia and subsequent protection of patients against low vision and blindness caused by myopic maculopathy.

Purpose: In this work, we propose a DnCNN-based method to enhance the contrast of images. Fundus images are widely used in clinical practice for detecting retinal dark spots. These spots trouble op...

Although unprecedented sensitivity and specificity values are reported, recent glaucoma detection deep learning models lack in decision transparency. Here, we propose a methodology that advances explainable deep learning in the field of glaucoma detection and vertical cup-disc ratio (VCDR), an important risk factor. We trained and evaluated deep learning models using fundus images that underwent a certain cropping policy. We defined the crop radius as a percentage of image size, centered on the optic nerve head (ONH), with an equidistant spaced range from 10-60% (ONH crop policy). The inverse of the cropping mask was also applied (periphery crop policy). Trained models using original images resulted in an area under the curve (AUC) of 0.94 [95% CI 0.92-0.96] for glaucoma detection, and a coefficient of determination (R-2) equal to 77% [95% CI 0.77-0.79] for VCDR estimation. Models that were trained on images with absence of the ONH are still able to obtain significant performance (0.88 [95% CI 0.85-0.90] AUC for glaucoma detection and 37% [95% CI 0.35-0.40] R-2 score for VCDR estimation in the most extreme setup of 60% ONH crop). Our findings provide the first irrefutable evidence that deep learning can detect glaucoma from fundus image regions outside the ONH. Research Group Ophthalmology, KU Leuven; VITO NV; Flemish Government; European Commission

In the human eye, Damage in the retina may cause ophthalmic diseases like cataracts, AMD, Hypertensive retinopathy, myopia, etc. To cure these diseases, many ophthalmologists use retinal fundus images as an important information source to find out ophthalmic diseases. Multiple techniques have been introduced for the screening of ocular diseases. Today’s world is in great demand to find out ocular diseases by using deep learning and machine learning techniques. This paper uses pre-trained deep neural networks to determine five categories of ophthalmic diseases such as cataract, AMD, Hypertensive retinopathy, myopia, and normal. Dataset is created into binary and multiclass, then trained on Resnet-101 of convolutional neural network (CNN) and evaluated. The accuracy of this model is found to be 90.38% and 88.5% for binary and multiclass respectively. Keywords: Retinal fundus image, Ocular diseases, CNN, ResNet, deep learning. Image processing, Ensemble classifier

Diabetic retinopathy, initially symptomless medical condition of diabetes, is one of the significant reasons of vision impairment all over the world. The early detection and diagnosis can reduce the incidence of severe vision loss and improve the efficiency of treatment. Fundus imaging, a non-invasive diagnosis method, is initial and widely used mode for analysing diabetic retinopathy. However, the precision of fundus imaging-based diagnosis of retinal disease is vastly dependent on experience and knowledge of ophthalmologists. Computer-aided diagnostic systems designed for retinal fundus images aid quick diagnosis, offer an external viewpoint during decision making, and serves as an important way of assessing treatment response to retinal diseases. In this paper, firstly the issues faced by ophthalmologists in characterization of various landmark structures and retinal lesions related to diabetic retinopathy by ophthalmologists are stated. Secondly, a comprehensive review of the state-of-the-art methods on landmark structures detection and segmentation, retinal lesions segmentation and diabetic retinopathy screening methods with retinal fundus images is presented. A concise tabular summary of each section comparing various methods, related retinal image databases, performance parameters, advantages, and disadvantages of the published methods is also exhibited. Finally, the main findings with focus to current challenges and ways for further improvement with respect to research gaps are discussed and concluded.

Diabetic retinopathy (DR) is one among the common disease associated with the human eye that can cause blindness. Detection of DR is very important as the disease will damage the eye with the passage of time. A computer-aided diagnosis-based system is used nowadays to assist the medical practitioner to correctly detect DR during the early stages. In this work, a retinal image’s classification is proposed, which is composed of three major blocks. Initially, the images are preprocessed using CLAHE and DNCNN neural networks, which will reduce the induced noise in the images. Preprocessed images then segmented using morphological and K-mean algorithms. The enhanced images have shown a better peak signal-to-noise ratio. The segmented images are then fed to the proposed EyeNet, which is a transfer learning-based model. The architecture of EyeNet is based on ResNet-18. The network is trained on more than 1500 images from clinical, DRIVE, and STARE databases and has shown an accuracy of 99.76%.

Aim: Flow-mediated dilatation (FMD) and retinal vascular analysis (RVA) may assist in predicting cardiovascular disease (CVD) but are poorly characterised in South Africa. We recorded baseline FMD and retinal vascular widths in healthy participants, and investigated associations with cardiovascular risk factors. Methods: Endothelial function (measured with FMD), microvascular structure (evaluated via fundus image analysis) and major CVD risk factors were assessed in 66 participants from Cape Town. Results: Median FMD% was 9.6%, with higher values in females. Mean retinal arteriolar and venular widths were similar to 156 and similar to 250 mu m, respectively. FMD was not associated with CVD risk factors. Hypertension was associated with narrower retinal arterioles and venules. Conclusions: We report novel baseline FMD data in healthy South African adults from the Western Cape, and show that retinal microvascular calibres are associated with blood pressure. Our baseline FMD and RVA data could serve as a reference for future studies in South Africa.

Medical image segmentation can be implemented using Deep Learning methods with fast and efficient segmentation networks. Single-board computers (SBCs) are difficult to use to train deep networks due to their memory and processing limitations. Specific hardware such as Google's Edge TPU makes them suitable for real time predictions using complex pre-trained networks. In this work, we study the performance of two SBCs, with and without hardware acceleration for fundus image segmentation, though the conclusions of this study can be applied to the segmentation by deep neural networks of other types of medical images. To test the benefits of hardware acceleration, we use networks and datasets from a previous published work and generalize them by testing with a dataset with ultrasound thyroid images. We measure prediction times in both SBCs and compare them with a cloud based TPU system. The results show the feasibility of Machine Learning accelerated SBCs for optic disc and cup segmentation obtaining times below 25 milliseconds per image using Edge TPUs., Preprint of paper published in Engineering Applications of Artificial Intelligence

The purpose of this study was to evaluate the use of Fiji Image J application for digital fundus image analysis of retinal vessel diameter in diabetic retinopathy rat model. Male Sprague-Dawley rats, weighing 200-250 grams, were divided into two groups: normal and diabetic. The diabetes was induced by intraperitoneal (IP) injection of streptozotocin (STZ, 55 mg/kg body weight). Normal rats received IP citrate buffer. Fundus images were captured at week 0, 6 and 12 post-induction to observe changes in retinal veins and arteries. Images obtained were then analyzed using Fiji Image J software. Retinal venous diameter was increased in both groups at week 6 and 12 compared to baseline (p

PurposeThe abnormalities of glaucoma have high impact on deciding and representing the causes that effects severity of blindness in human beings. The simulation experimental results would help the ophthalmologist in diagnosing of glaucoma abnormality accurately. The significant effect of glaucoma has a huge impact on the quality of human life, and its growth rate in world population tremendously increases. Glaucoma is considered as second largest cause for the blindness in the world; hence identification of it marks the importance of its detection at the earliest.Design/methodology/approachThe prime objective of the work proposed is to build up a human intervention free image preparing framework for glaucoma screening. The disc calculation is assessed on retinal image dataset called retinal Image for glaucoma Analysis. The proposed method briefs a novel optic disc division calculation depending on applying a level-set strategy on a confined optic disc image. In the instance of low quality image, a twofold level set is designed, in which the principal level set is viewed as restriction for the optic disc. To keep the veins from meddling with the level-set procedure, an inpainting strategy has been applied. Also a significant commitment is to include the varieties in notion adopted by the ophthalmologists in distinguishing the disc localization and diagnosing the glaucoma. Most of the past investigations are prepared and tested depending on just a single feature, which can be thought to be one-sided for the ophthalmologist.FindingsIn continuation, the correctness has been determined depending on the quantity of image that matched with the investigation pattern adopted by the ophthalmologist. The 175 retinal images were utilized to test the results of proposed work with the manual markings of ophthalmologists. The error-free calculation in marking the optic disc region and centroid was 98.95% in comparison with the existing result of 87.34%.Originality/valueIn continuation, the correctness has been determined depending on the quantity of image that matched with the investigation pattern adopted by the ophthalmologist. The 175 retinal images were utilized to test the results of proposed work with the manual markings of ophthalmologists. The error-free calculation in marking the optic disc region and centroid was 98.95% in comparison with the existing result of 87.34%.

Objective: Branch Retinal Artery Occlusion (BRAO) has become a serious disorder causing permanent loss of vision. It is mainly caused due to the rupture of blood vessels which is too small in size ...

Convolutional neural network (CNN) is a method of supervised deep learning. The architectures including AlexNet, VGG16, VGG19, ResNet 50, ResNet101, GoogleNet, Inception-V3, Inception ResNet-V2, and Squeezenet that have 25 to 825 layers. This study aims to simplify layers of CNN architectures and increased accuracy for fundus patches classification. Fundus patches classify two categories: normal and neovascularization. Data used for classification is MESSIDOR and Retina Image Bank that have 2,080 patches. Results show the best accuracy of 93.17% for original data and 99,33% for augmentation data using CNN 31 layers. It consists input layer, 7 convolutional layers, 7 batch normalization, 7 rectified linear unit, 6 max-pooling, fully connected layer, softmax, and output layer.

The localization of the optic disc and fovea is crucial in the automated diagnosis of various retinal diseases. We propose a novel deep learning driven heatmap regression model based on the encoder-decoder architecture for the joint detection of optic disc and fovea centers in color fundus images. To train the regression model, we transform the ground-truth center coordinates of optic disc and fovea of the IDRiD dataset to heatmaps using a 2D-Gaussian equation. The model is capable of pinpointing any single pixel in a vast 2D image space. The model is tested on IDRiD test dataset, Messidor, and G1020 datasets. The model outperforms the state-of-the-art methods on these datasets. The model is very robust and generic, which can be trained and used for the simultaneous localization of multiple landmarks in different medical image datasets. The full implementation code and the trained model with weights (based on Keras) are available for reuse at https://github.com/bhargav-jb/FundusPosNet.

Glaucoma has been credited to be the foremost cause of preventable loss of sight in the world second only to cataract. Its effect on the eye is usually irreversible and can only be prevented by early detection. In this paper, we developed a glaucoma detection technique. This technique includes a modified U-Net model called ‘U-Net lite’ and an extreme gradient boost (XGB) algorithm. The novel U-Net lite model is designed to have fewer parameters than the original U-Net model. The U-Net lite’s parameters are 40 times fewer than the original U-Net model which makes the proposed model faster and cheaper to train. The proposed model is utilized to segment both the optic cup and the optic disc from the fundus images. The extreme gradient boost algorithm is utilized to analyze extracted features from segmented optic cups and discs and hence detect glaucoma. The proposed U-Net lite model was both trained and tested on the DRIONS, DRISHTI-GS, RIM-ONE V2 and the RIM-ONE V3 databases. When tested for optic disc segmentation on the four databases, the model achieved the following average dice-scores: 0.96 on RIM-ONE V3, 0.97 on RIM-ONE V2, 0.96 on DRIONS, and 0.97 on DRISHTI-GS. The XGB algorithm achieved an accuracy of 88.6% and an AUC-ROC of 93.6 % in detecting glaucoma from the RIM-ONE V3 and DRISHTI-GS database. The proposed glaucoma detection technique achieves a state-of-the-art accuracy and is useful for observing structural changes in an optic cup and optic disc.

Small changes in retinal blood vessels may produce different pathological disorders which may further cause blindness. Therefore, accurate extraction of vasculature map of retinal fundus image has become a challenging task for analysis of different pathologies. The present study offers an unsupervised method for extraction of vasculature map from retinal fundus images. This paper presents the methodology for evolution of vessels using Modified Pixel Level Snake (MPLS) algorithm based on Black Top-Hat (BTH) transformation. In the proposed method, initially bimodal masking is used for extraction of the mask of the retinal fundus image. Then adaptive segmentation and global thresholding is applied on masked image to find the initial contour image. Finally, MPLS is used for evolution of contour in all four cardinal directions using external, internal and balloon potential. This proposed work is implemented using MATLAB software. DRIVE and STARE databases are used for checking the performance of the system. In the proposed work, various performance metrics such as sensitivity, specificity and accuracy are evaluated. The average sensitivity of 76.96%, average specificity of 98.34% and average accuracy of 96.30% is achieved for DRIVE database. This technique can also segment vessels of pathological images accurately; reaching the average sensitivity of 70.80%, average specificity of 96.40% and average accuracy of 94.41%. The present study provides a simple and accurate method for the detection of vasculature map for normal fundus images as well as pathological images. It can be helpful for the assessment of various retinal vascular attributes like length, diameter, width, tortuosity and branching angle.

Automated segmentation of the optic disc (OD) and optic cup (OC) from different datasets plays an important role in the diagnosis of glaucoma and greatly saves human resources in both data annotation and image segmentation. However, the domain shift between different datasets suppresses the generalization ability of the segmentation network, especially damaging the performance of segmentation in the target domain, which is unlabeled. Therefore, using a transfer learning algorithm or domain adaptation method to enhance the migration ability of segmentation models has become an essential step and has attracted the attention of many researchers. In this paper, we propose an unsupervised domain adaptation network, called the Minimizing-entropy and Fourier Domain Adaptation network (MeFDA), to narrow the discrepancy between the source and target domains and prevent the degradation of segmentation performance. First, we perform adversarial optimization on the entropy maps of the predicted segmentation results to alleviate the domain shift. Then, direct entropy-minimization optimization is applied to the unlabeled target domain data to improve the credibility of the prediction segmentation maps. To enhance the prediction consistency of the target domain data, we augment the target domain dataset through the Fourier transform by replacing the low-frequency part in the target images with that of the source images. Then, a semantic consistency constraint is imposed on the raw images and augmented images of the target domain to improve the prediction consistency of the segmentation model, thereby further narrowing the discrepancy between the source and target domains. Experiments on several public retinal fundus image datasets prove the superiority of MeFDA compared with state-of-the-art methods, and the ablation study analyzes the importance of the different proposed components.

Glaucoma is a turmoil where serious harm to the optic nerve prompts vision misfortune. Glaucoma distinguishing proof includes the estimation of the shape and size of the optic cup. Because of the interweavement of the optic cup with veins, the optic cup division is somewhat dull errand. Pre-preparing followed by division is utilized for optic cup division which is additionally handled to discover its measurement. In view of the reality the fractal measurement is utilized to discover the element of unpredictable articles, a novel methodology is proposed for glaucoma location utilizing the edge strategy for fractal examination. The Normal and glaucoma absconded picture is arranged and recognized by computerized picture preparing by CNN technique and executed utilizing IOT.

Current clinical approaches to diabetic foot (DF) treatment mainly rely on clinician vigilance and laboratory test, which have significant limitations, such as the high cost involved in the diagnosis and the high demands for professional skills of clinicians. At present, the research on DF prediction has mainly focused on the regression analysis of clinical data and the recognition based on foot ulcers skin. In view of this situation, we examine the patients’ fundus images to explore an efficient way for DF prediction. In this paper, we have proposed a DF prediction model through fundus images by radiomics features. Twelve kinds of radiomics features are extracted, including a variety of features first applied in the field of medical imaging, describing the information of image texture, direction, phase, and gradient. Subsequently, a two-step feature selection model is put forward for a large number of radiomics features we extracted for the purpose of searching for the best combination. Considering the simplicity and performance of the model, we chose 19 features to train the support vector machine model. The obtained model is evaluated with 5-fold cross validation on abundant clinical data, and the mean prediction performance: area under the curve: 0.9678; sensitivity: 0.9786; specificity: 0.9161; accuracy: 0.9247, showed the excellence of the model prediction. Here, we present a new, noninvasive, and efficient detection means for the automatic prediction of DF, which can help clinicians find potential diabetic foot patients earlier and is expected to be a novel auxiliary diagnostic tool.

In order to study the application of color fundus image (CFI) visual cup segmentation (VCS) algorithm in the design of eye health monitoring system and solve the problem that traditional image processing technology requires artificial design features, which is time-consuming, laborious, and inaccurate, deep learning (DL) algorithm is combined with residual neural network (RNN) to propose a new eye health monitoring system based on CFI VCS algorithm. Firstly, the forward propagation (FP) algorithm and back propagation (BP) algorithm of DL are theoretically derived and improved. Then, the eye health monitoring and image processing system is designed and optimized with RNN. Finally, the effect analysis VCS of the system is carried out on the real data center platform; the accuracy, sensitivity, and specificity of image processing are compared with the traditional manual processing technology and the general residual network technology. The results show that the results of RIM-ONE data set and Glaucoma Repo data set are consistent with the expert standard; in the RIM-ONE data set, the accuracy rate, sensitivity, and specificity of traditional manual processing technology are 69.68%, 71.38%, and 74.71% respectively; the accuracy, sensitivity, and specificity of common residual network technology are 81.89%, 80.56%, and 86.82% respectively; the accuracy of the system is 93.57%, the sensitivity is 87.31%, and the specificity is 90.16%. In the Glaucoma Repo data set, the accuracy, sensitivity, and specificity of conventional manual processing technology are 75.28%, 73.35%, and 82.63% respectively; the accuracy, sensitivity, and specificity of common residual network technology are 82.74%, 81.49%, and 88.82% respectively; the accuracy, sensitivity, and specificity of the system are 94.66%, 91.33%, and 94.22% respectively, which show that the system has a good effect on the segmentation of CFI, can adapt to various features of the image, has a high degree of accuracy and robustness, and provides a new way of thinking for the monitoring and management of eye health in daily life.

Image classification using convolutional neural networks (CNNs) outperforms other state-of-the-art methods. Moreover, attention can be visualized as a heatmap to improve the explainability of results of a CNN. We designed a framework that can generate heatmaps reflecting lesion regions precisely. We generated initial heatmaps by using a gradient-based classification activation map (Grad-CAM). We assume that these Grad-CAM heatmaps correctly reveal the lesion regions; then we apply the attention mining technique to these heatmaps to obtain integrated heatmaps. Moreover, we assume that these Grad-CAM heatmaps incorrectly reveal the lesion regions and design a dissimilarity loss to increase their discrepancy with the Grad-CAM heatmaps. In this study, we found that having professional ophthalmologists select 30% of the heatmaps covering the lesion regions led to better results, because this step integrates (prior) clinical knowledge into the system. Furthermore, we design a knowledge preservation loss that minimizes the discrepancy between heatmaps generated from the updated CNN model and the selected heatmaps. Experiments using fundus images revealed that our method improved classification accuracy and generated attention regions closer to the ground truth lesion regions in comparison with existing methods.

We propose a new information technology for mapping macular photocoagulation patterns in the laser treatment of diabetic retinopathy and evaluating the treatment strategy efficacy. The technology relies on processing optical coherence tomography (OCT) data, fundus image segmentation, mapping photocoagulation plans, and development of prognostic indicators of the laser photocoagulation treatment outcome. Analysis of the OCT data enables a pathological area to be extracted and then refined based on retinal image segmentation results, allowing a photocoagulation treatment region to be outlined. A photocoagulation plan mapped for the region of interest (ROI) is then analyzed to identify the treatment success probability based on prognostic indicators. The technology developed will offer the retina specialist a preliminary photocoagulation plan, as well as allowing the assessment of the diabetic retinopathy treatment efficacy to be carried out.

Color fundus image quality greatly influence the doctors’ diagnostic accuracy. However, the problems of imbalance data and small sample are the key issues of the color fundus images quality assessment. Hence, this paper purposes a small sample color fundus image quality assessment based on gcforest to solve these problems. Firstly, this paper extracts color and texture features to represent the quality of color fundus image. Next, re-sampling process is used to re-balance training data. Thirdly, the training data after re-balanced is sent to train gcforest which is a forest integration model. Finally, the trained gcforest which is good for small sample problem is used to evaluate color fundus images quality. Experiments demonstrate that the proposed method not only in color fundus image quality assessment but also in glaucoma classification task get good results.

A gainful fuzzy k-means clustering algorithm under Morphological Image Processing (MIP) is performed. Image processing is one of quickly developing examination territory nowadays and now it is particularly coordinated with all identified with science field. Image Processing can be utilized for breaking down various restorative and MRI Image to get the uncommon and anomaly in the image. Image segmentation manages segmentation of vein segmentation algorithm utilizing fundus Image. In this task, this segmentation is done utilizing k-means clustering and c-means clustering algorithm and Morphological operator for better execution. This upgrades the vein variations from the norm progressively and in a moderately brief time when contrasted with numerous other clustering algorithms.

Early detection of eye diseases can forestall visual deficiency and vision loss. There are several types of human eye diseases, for example, diabetic retinopathy, glaucoma, arteriosclerosis, and hypertension. Diabetic retinopathy (DR) which is brought about by diabetes causes the retinal vessels harmed and blood leakage in the retina. Retinal blood vessels have a huge job in the detection and treatment of different retinal diseases. Thus, retinal vasculature extraction is significant to help experts for the finding and treatment of systematic diseases. Accordingly, early detection and consequent treatment are fundamental for influenced patients to protect their vision. The aim of this paper is to detect blood vessels from the digital fundus images. In this research, a novel methodology was introduced to separate retinal blood vessel network. The suggested system in this research involves four stages, after image acquisition, the pre-processes of the image to preparing and improving the image quality is the first stage. Morphological operations are used for the detection of blood vessels. In this research, we will use two morphological operations: erosion and dilation. These two operations have two inputs, a binary image, and a structuring element object. We will use two morphological processes (boundary extraction and top, bottom hat transform). Before these operations, we will use applying a canny edge detector technique to obtain the edges of the retina image. The technique is tried on shading retinal pictures acquired from STARE and DRIVE databases which are accessible on the web as well as the samples of retinal images were obtained from the digital camera from Ibn Al-Haytham specialist Hospital for Eye in Baghdad, Iraq. Good results and effective were obtained for blood vessel detected and extract

To correlate choriocapillaris (CC) flow deficits (FD) in eyes with geographic atrophy (GA) at various distances from the border of the GA lesion with yearly enlargement rate (yER) of GA. In this retrospective study, spectral domain optical coherence tomography (SD-OCT) and SD optical coherence tomography angiography (OCTA) images were collected from patients with GA, who were imaged at Doheny Eye Centers between 2016 and 2018, using the Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA). All enrolled patients had one baseline 6 × 6 mm OCTA scan and two 6 × 6 mm SD-OCT cubes, one at baseline and one at a follow-up visit at least 12 months later. The border of the GA was manually outlined on the en face OCT fundus image and the yER was calculated after square root transformation. A grid composed of 100-μm-wide successive concentric rings was created around the GA lesion on the OCTA CC slab using ImageJ and the FD% was calculated from the binarized image. FD% from each ring was correlated with the yER of GA. Thirty eyes of 22 patients were included in the study. The mean yER was 0.2 ± 0.15 mm. The FD% in the first five rings (from 0 to 500 μm from the border of GA) was significantly correlated with the yER. However, there was no statistically significant correlation between the yER and CC FD% beyond 500 μm from the GA lesion. Only the choriocapillaris FD% in the 500-μm region immediately surrounding GA lesions appears to predict the rate of enlargement of these lesions.

Purpose: To compare the quality of retinal images captured with a smartphone-based, handheld fundus camera with that of retinal images captured with a commercial fundus camera and to analyze their agreement in determining the cup-to-disc ratio for a cohort of ophthalmological patients. Methods: A total of 50 patients from a secondary ophthalmic outpatient service center underwent a bilateral fundus examination under mydriasis with a smartphone-based, handheld fundus camera and with a commercial fundus camera (4 images/patient by each). Two experienced ophthalmologists evaluated all the fundus images and graded them on the Likert 1-5 scale for quality. Multivariate regression analyses was then performed to evaluate the factors associated with the image quality. Two masked ophthalmologists determined the vertical cup-to-disc ratio of each fundus image, and both the intraobserver (between devices) and interobserver agreement between them was calculated. Results: Ninety-eight images from 49 patients were processed in this study for their quality analysis. Ten images from five patients (four from commercial fundus camera and one from smartphone-based, handheld fundus camera) were not included in the analyses due to their extremely poor quality. The medians [interquartile interval] of the image quality were not significantly different between those from the smartphone-based, handheld fundus camera and from the commercial fundus camera (4 [4-5] versus 4 [3-4] respectively, p=0.06); however, both the images captured with the commercial fundus camera and the presence of media opacity presented a significant negative correlation with the image quality. Both the intraobserver [intraclass correlation coefficient (ICC)=0.82, p

Diabetic Retinopathy (DR) is the complicatedness of diabetes that happens due to macular degeneration among Type II diabetic patients. The early symptom of this disease is predicted through annual eye checkups. Hence, one can save their vision at an early stage. Later on, it prompts retinal detachment. There is a requirement for awareness among diabetic patients about this disease to prevent their life from vision misfortune. Along these lines, there is a need for a computer-assisted method to analyze the disease. The proposed system used Adaptive Histogram Equalization (AHE) technique for image enhancement, Hop Field Neural Network for blood vessel segmentation, and Adaptive Resonance Theory (ART) for blood vessel classification. The proposed system analyzes the disease and classifies the disease level effectively with high accuracy. Also, the system notifies the users about the stages of the disease. The proposed system is evaluated with the clinical as well as open fundus image data sets like DRIVE, STARE, MESSIDOR, HRF, DRIONS, and REVIEW for diabetic retinopathy prediction. Also, physicians evaluated the system and concluded that the proposed system does not deviate from the quality of disease analysis and grading. The proposed techniques accomplished 99.99% accuracy. The system is evaluated by the ophthalmologists and witnesses that the proposed system has not veered off as far as quality.

Diabetes mellitus, also known as diabetes, is a metabolic disease which may cause high blood sugar, blurred vision and several other complications. Patients already suffering from diabetes have a greater probability to develop Diabetic Retinopathy. Most prominently in the blood vessels of the tissue residing in the rear portion of the eye gets damaged causing vision problems. This disease is especially prevalent in India. Test for Diabetic Retinopathy involves a dilated eye exam, which is time consuming and requires specialized doctors and distinct equipment which may not be easily accessible to the people especially in rural areas. Predicting the possibility of Diabetic Retinopathy utilizing fundus image is considered an effectual analysis subject by many. In this paper a method is explored to diagnose Diabetic Retinopathy using fundus images and deep learning to accurately tell the patient whether he has a Diabetic Retinopathy. Features used from the eye fundus images are soft exudate and hard exudate along with optic disc. A website is also made to deploy the model and promote mass diagnosis at a faster pace.

Subretinal fluid (SRF) can lead to irreversible visual loss in patients with central serous chorioretinopathy (CSC) if not absorbed in time. Early detection and intervention of SRF can help improve visual prognosis and reduce irreversible damage to the retina. As fundus image is the most commonly used and easily obtained examination for patients with CSC, the purpose of our research is to investigate whether and to what extent SRF depicted on fundus images can be assessed using deep learning technology. In this study, we developed a cascaded deep learning system based on fundus image for automated SRF detection and macula-on/off serous retinal detachment discerning. The performance of our system is reliable, and its accuracy of SRF detection is higher than that of experienced retinal specialists. In addition, the system can automatically indicate whether the SRF progression involves the macula to provide guidance of urgency for patients. The implementation of our deep learning system could effectively reduce the extent of vision impairment resulting from SRF in patients with CSC by providing timely identification and referral.

Feature-based retinal fundus image registration (RIR) technique aligns fundus images according to geometrical transformations estimated between feature point correspondences. To ensure accurate registration, the feature points extracted must be from the retinal vessels and throughout the image. However, noises in the fundus image may resemble retinal vessels in local patches. Therefore, this paper introduces a feature extraction method based on a local feature of retinal vessels (CURVE) that incorporates retinal vessels and noises characteristics to accurately extract feature points on retinal vessels and throughout the fundus image. The CURVE performance is tested on CHASE, DRIVE, HRF and STARE datasets and compared with six feature extraction methods used in the existing feature-based RIR techniques. From the experiment, the feature extraction accuracy of CURVE (86.021%) significantly outperformed the existing feature extraction methods (p ≤ 0.001*). Then, CURVE is paired with a scale-invariant feature transform (SIFT) descriptor to test its registration capability on the fundus image registration (FIRE) dataset. Overall, CURVE-SIFT successfully registered 44.030% of the image pairs while the existing feature-based RIR techniques (GDB-ICP, Harris-PIIFD, Ghassabi’s-SIFT, H-M 16, H-M 17 and D-Saddle-HOG) only registered less than 27.612% of the image pairs. The one-way ANOVA analysis showed that CURVE-SIFT significantly outperformed GDB-ICP (p = 0.007*), Harris-PIIFD, Ghassabi’s-SIFT, H-M 16, H-M 17 and D-Saddle-HOG (p ≤ 0.001*).

Background: Pathologic myopia (PM) associated with myopic maculopathy (MM) and “Plus” lesions is a major cause of irreversible visual impairment worldwide. Therefore, we aimed to develop a series of deep learning algorithms and artificial intelligence (AI)–models for automatic PM identification, MM classification, and “Plus” lesion detection based on retinal fundus images.Materials and Methods: Consecutive 37,659 retinal fundus images from 32,419 patients were collected. After excluding 5,649 ungradable images, a total dataset of 32,010 color retinal fundus images was manually graded for training and cross-validation according to the META-PM classification. We also retrospectively recruited 1,000 images from 732 patients from the three other hospitals in Zhejiang Province, serving as the external validation dataset. The area under the receiver operating characteristic curve (AUC), sensitivity, specificity, accuracy, and quadratic-weighted kappa score were calculated to evaluate the classification algorithms. The precision, recall, and F1-score were calculated to evaluate the object detection algorithms. The performance of all the algorithms was compared with the experts’ performance. To better understand the algorithms and clarify the direction of optimization, misclassification and visualization heatmap analyses were performed.Results: In five-fold cross-validation, algorithm I achieved robust performance, with accuracy = 97.36% (95% CI: 0.9697, 0.9775), AUC = 0.995 (95% CI: 0.9933, 0.9967), sensitivity = 93.92% (95% CI: 0.9333, 0.9451), and specificity = 98.19% (95% CI: 0.9787, 0.9852). The macro-AUC, accuracy, and quadratic-weighted kappa were 0.979, 96.74% (95% CI: 0.963, 0.9718), and 0.988 (95% CI: 0.986, 0.990) for algorithm II. Algorithm III achieved an accuracy of 0.9703 to 0.9941 for classifying the “Plus” lesions and an F1-score of 0.6855 to 0.8890 for detecting and localizing lesions. The performance metrics in external validation dataset were comparable to those of the experts and were slightly inferior to those of cross-validation.Conclusion: Our algorithms and AI-models were confirmed to achieve robust performance in real-world conditions. The application of our algorithms and AI-models has promise for facilitating clinical diagnosis and healthcare screening for PM on a large scale.

In recent years, an increasing number of people have myopia in China, especially the younger generation. Common myopia may develop into high myopia. High myopia causes visual impairment and blindness. Parapapillary atrophy (PPA) is a typical retinal pathology related to high myopia, which is also a basic clue for diagnosing high myopia. Therefore, accurate segmentation of the PPA is essential for high myopia diagnosis and treatment. In this study, we propose an optimized Unet (OT-Unet) to solve this important task. OT-Unet uses one of the pre-trained models: Visual Geometry Group (VGG), ResNet, and Res2Net, as a backbone and is combined with edge attention, parallel partial decoder, and reverse attention modules to improve the segmentation accuracy. In general, using the pre-trained models can improve the accuracy with fewer samples. The edge attention module extracts contour information, the parallel partial decoder module combines the multi-scale features, and the reverse attention module integrates high- and low-level features. We also propose an augmented loss function to increase the weight of complex pixels to enable the network to segment more complex lesion areas. Based on a dataset containing 360 images (Including 26 pictures provided by PALM), the proposed OT-Unet achieves a high AUC (Area Under Curve) of 0.9235, indicating a significant improvement over the original Unet (0.7917).

Dust particle artifacts are present in all imaging modalities but have more adverse consequences in medical images like retinal images. They could be mistaken as small lesions, such as microaneurysms. We propose a method for detecting and accurately segmenting dust artifacts in retinal images based on multi-scale template-matching on several input images and an iterative segmentation via an inpainting approach. The inpainting is done through dictionary learning and sparse-based representation. The artifact segmentation is refined by comparing the original image to the initial restoration. On average, 90% of the dust artifacts were detected in the test images, with state-of-theart restoration results. All detected artifacts were accurately segmented and removed. Even the most challenging artifacts located on top of blood vessels were removed. Thus, ensuring the continuity of the retinal structures. The proposed method successfully detects and removes dust artifacts in retinal images, which could be used to avoid false-positive lesion detections or as an image quality criterion. An implementation of the proposed algorithm can be accessed and executed through a Code Ocean compute capsule. The authors acknowledge the financial support from the Centre de Cooperació i Desenvolupament (CCD) at the Universitat Politècnica de Catalunya under project ref. CCD 2019-B004, and from the Universidad Tecnológica de Bolívar. Authors are grateful to Juan Luís Fuentes from the Miguel Servet University Hospital (Zaragoza, Spain) for providing the real images from clinical practice. E. Barrios thanks Minciencias and Sistema General de Regalías (Programa de Becas de Excelencia) for a PhD scholarship. E. Sierra thanks the Universidad Tecnológica de Bolívar for a post-graduate scholarship. Parts of this work were presented at the Pattern Recognition and Tracking XXX - SPIE DCS, 2019 [39]. L. Romero, A. Marrugo, and M.S. Millán thank the funds provided by the Spanish Ministerio de Ciencia e Innovación under the project reference PID2020-114582RB-I00.

Unprecedented modern research in medical science reveals a new horizon in detection, diagnosis, and treatment of the various ophthalmology-related disorders by analyzing vessel configurations in retinal fundus images. The main motivation is to come up with a new scheme to separate out vessels from the others in fundus image without manual intervention. In the suggested method, first, we generate three layers from the fundus with different features namely ‘prominent vessel network’, ‘minute vessel details’, and ‘illumination improved fundus’. In the second step, a new algorithm has been defined to clean unwanted ‘dust-like’ noise and restructure the vascular network with all the subtle branches by means of a binary image. The new scheme is executed and validated on images available in open-sourced DRIVE repository. Outcomes of the proposed scheme are tallied with other schemes for establishing its superiority.

Purpose Fundus images are typically used as the sole training input for automated diabetic retinopathy (DR) classification. In this study, we considered several well-known DR risk factors and attempted to improve the accuracy of DR screening. Metphods Fusing nonimage data (e.g., age, gender, smoking status, International Classification of Disease code, and laboratory tests) with data from fundus images can enable an end-to-end deep learning architecture for DR screening. We propose a neural network that simultaneously trains heterogeneous data and increases the performance of DR classification in terms of sensitivity and specificity. In the current retrospective study, 13,410 fundus images and their corresponding nonimage data were collected from the Chung Shan Medical University Hospital in Taiwan. The images were classified as either nonreferable or referable for DR by a panel of ophthalmologists. Cross-validation was used for the training models and to evaluate the classification performance. Results The proposed fusion model achieved 97.96% area under the curve with 96.84% sensitivity and 89.44% specificity for determining referable DR from multimodal data, and significantly outperformed the models that used image or nonimage information separately. Conclusions The fusion model with heterogeneous data has the potential to improve referable DR screening performance for earlier referral decisions. Translational relevance Artificial intelligence fused with heterogeneous data from electronic health records could provide earlier referral decisions from DR screening.

Diagnosing and detection of ophthalmology related disorders, extraction of vessels from retinal fundus images is very important task in Medical Science. Locating retinal vessels manually is not only tough and time taking task for ophthalmologist but also it is prone to have human error because of complex nature of blood vessels in Retina. Across the globe different techniques are being applied to overcome this challenge by deploying different strategies and techniques. This paper proposes a new method for extraction of vessels from retinal images. Proposed method first extracts thick vessels from the fundus image. Finally, flood filling algorithm in-conjugation with BFS is applied to discover the tiny blood vessels and reconstruct the final result by combining it with the thick vessels extracted in the first place. Experiment is performed on publicly available DRIVE data set and it is found that proposed method produced quite outstanding results.

Purpose This study aimed to quantitative assess the fundus tessellated density (FTD) and associated factors on the basis of fundus photographs using artificial intelligence. Methods A detailed examination of 3468 individuals was performed. The proposed method for FTD measurements consists of image preprocessing, sample labeling, deep learning segmentation model, and FTD calculation. Fundus tessellation was extracted as region of interest and then the FTD could be obtained by calculating the average exposed choroid area of per unit area of fundus. Besides, univariate and multivariate linear regression analysis have been conducted for the statistical analysis. Results The mean FTD was 0.14 ± 0.08 (median, 0.13; range, 0-0.39). In multivariate analysis, FTD was significantly (P < 0.001) associated with thinner subfoveal choroidal thickness, longer axial length, larger parapapillary atrophy, older age, male sex and lower body mass index. Correlation analysis suggested that the FTD increased by 33.1% (r = 0.33, P < .001) for each decade of life. Besides, correlation analysis indicated the negative correlation between FTD and spherical equivalent (SE) in the myopia participants (r = -0.25, P < 0.001), and no correlations between FTD and SE in the hypermetropia and emmetropic participants. Conclusions It is feasible and efficient to extract FTD information from fundus images by artificial intelligence-based imaging processing. FTD can be widely used in population screening as a new quantitative biomarker for the thickness of the subfoveal choroid. The association between FTD with pathological myopia and lower visual acuity warrants further investigation. Translational Relevance Artificial intelligence can extract valuable clinical biomarkers from fundus images and assist in population screening.

Diabetic retinopathy has become one of the major reasons for blindness in the world. Early and precise diagnosis of the disease may save one’s eyesight from irreversible damage. Manual detection of lesions is time consuming and may not be as accurate as desirable. Many automated systems have been developed recently to help ophthalmologists in their endeavors. Exudates are one of the early signs of manifestation of diabetic retinopathy. In this paper, the methodologies detecting exudates in retinal fundus images were reviewed. These methods were categorized into deep learning, machine learning and methods primarily focusing on image processing techniques. The comprehensive view of the performances of the methods was given. Several datasets were described briefly. Most of the researchers preferred combination of multiple publically available databases. Also, the potential areas of research were discussed. It was found that sensitivity which identifies the abnormal images correctly, is the most widely used performance measure. The study will be helpful to the researchers wanting to explore more in this field.

This paper presents an unsupervised and novel method for extraction of vasculature map from colored retinal fundus images. Proposed technique makes use of a fusion of bimodal masking and global thresholding technique for extraction of vessels. For this, adaptive histogram equalization method is used for enhancement of the retinal input fundus images while, on the other hand, an average filter is used on the masked images to remove the noise from the image. At this stage, bimodal masking is used to generate the masked image to exclude the pixels that belong to the background. The use of this technique reduces the analysis time and computational effort as operations will be focused only on the object pixels. After that, a global thresholding technique is utilized on the masked foreground image, which produces a vasculature map with border. Since the border has no concern in our system, hence, it is removed using the mask generated through bimodal masking. Fundus images of DRIVE and STARE database is used to perform extensive computations. The results are encouraging as the proposed system shows better outcomes on 3-quality measures: average sensitivity, specificity, and accuracy which comes out to be (84.18%, 96.68% and 95.68%) and (81.79%, 90.74% and 90.08%) for DRIVE and STARE database respectively. The results prove that the proposed methodology is capable of extracting the vasculature map accurately, which can be further helpful for diagnosis, screening, and treatment of various disorders of fundus images.

Identifikasi potensi glaukoma dan retinopati diabetes dapat dilakukan melalui citra fundus. Jaringan syaraf tiruan dapat digunakan untuk mengidentifikasi potensi glaukoma dan diabetes retinopati. Dataset yang digunakan berjumlah 60 citra fundus yang terdiri dari 20 citra fundus terjangkit glaukoma, 20 citra fundus terjangkit diabetes retinopati dan 20 citra fundus mata normal. Hasil penelitian menghasilkan rata-rata untuk recall sebesar 86,6%, precission sebesar 86,6%, dan untuk accuracy sebesar 91,06%.

The tissue layer located at the back of the eye is known as retina which converts the incoming light into nerve signals and those signals are sent to the brain for understanding. The damage onto the retina is termed as retinopathy and that may lead to vision weakening or vision loss. The hard exudates are small white or yellowish white deposits with their edges being clear and sharp. In the proposed methods we take color image of retina then extract the green channel of that image then apply top hat transformation and bottom hat transformation on that image. The DIARETDB1 and High-Resolution Fundus (HRF) databases are used for performance evaluation of the proposed method. The proposed technique achieves accuracy 97%, sensitivity 95%, and specificity 96% and it takes average 5.6135 second for detection of hard exudates in an image.

Glaucoma as an irreversible blinding opioid neuropathy disease, its blindness rate is the second only after cataract in the world. The optic cup-to-disc ratio (CDR) is generally considered to be an important clinical indicator for judging the severity of glaucoma by ophthalmologists from retinal fundus image. In this letter, we propose an automatic CDR measurement method that consists of a novel optic disc localization method and a simultaneous optic disc and cup segmentation network based on the improved U shape deep convolutional neural network. Experimental results demonstrate that the proposed method can achieve superior performance when compared with other existing methods. Thus, our method can be used as a powerful tool for glaucoma-assisted diagnosis.

In this paper, we propose a non-invasive method to detect glaucoma eye disease at the earlier stage from aqueous humor (AH) fluid’s biomagnetic signal in anterior chamber of eye. The AH fluid has biomagnetic emission because of magnetic material, such as sodium, potassium, chloride, bicarbonate, and protein. At present, the diagnosis of glaucoma was done at the later stage because of undetectable microvariation in the fundus image cup to disc size ratio measurement and because of similar symptoms in glaucoma eye and vision problem. In the proposed method, glaucoma can be detected through the acquisition of AH fluid biomagnetic signal from eye through giant magnetoresistance magnetometers (GMRs) sensor. The biomagnetic signal of AH fluid’s dynamic and static variation in eye is analyzed with the rational dilation wavelet transform (RADWT) algorithm. The variations of AH fluid is obtained through beta-blocker eye drops. From experiment, The AH fluid’s biomagnetic signal shows variations for the rate of dissipation of fluid in eye, such as normal and glaucoma eye, with and without eye drops. The AH biomagnetic signal correlates glaucoma disease with signal variance and RADWT subband energy for 50 glaucoma and 10 normal eye. The proposed method detect glaucoma eye in the early stage with 90% accuracy. The experimental result is validated with glaucoma measuring tonometer instrument. The earlier detection of glaucoma eye is measured through regression modeling.

Objectives: To make the automatic detection of Glaucoma at early stage of the disease. Methods/Statistical Analysis: Fundoscopy is one of the medical specialty techniques to understand the inner structure of the membrane. In this study, we tend to project the combination of structural and textural characteristics of a fundus image to solidify the eye disease diagnosis. Findings: The planned technique introduces an extraction of structural and textural features by the Naive Bayes algorithm and it is classified using the decision tree algorithm to detect the glaucoma cases. Naive Bayes can be implemented fast and ease in the less dataset. The planned technique introduces a category of perpetrators in the machinecontrolled diagnosis in the event of any conflict in the determination of structural features. Application/Improvements: The estimate of the proposed algorithm is carried out using the local database of 100 patients with fundus images. It calculates the instance value to identify the glaucoma level. The proposed system gives exceptional results with better accuracy. Keywords: Decision Tree, Fundus Images, Naive Bayes, PCA

Diabetic Retinopathy (DR) is an abnormality related with eye. It is caused due to alteration of the blood vessels present in the retina. If this disease is not treated at the early time, it...