Diabetic Retinopathy Detection through Multiclass Classification of Fundus Image Using Convolutional Neural Network

Artificial intelligence for diabetic retinopathy screening in Africa

Artificial intelligence using deep learning to screen for referable and vision-threatening diabetic retinopathy in Africa: a clinical validation study

This project proposes machine learning methods for detecting diabetic retinopathy by using CNN. The classification of retinal lesions from non-lesions using CNN classifiers is examined. Retinal retina images can be evaluated using machine learning-based medical image analysis. The image is pre-processed using the Contrast limited adaptive histogram equalization (CLAHE) filter. CLAHE is used to improve the visibility level of blurry images. The image is segmented using the FCM method to generate an appropriate threshold value for data segmentation. The method of feature extraction is implemented by RLC. The Gray Level Co – occurrence Matrix (GLCM) method is used for extracting statistical skin parameters in a given direction and distance. The use of convolutional neural network (CNN) algorithms has facilitated the early identification of diabetic retinopathy. This project provides excellent specificity and sensitivity for classifying images as with or without diabetic retinopathy. This project is implemented using MATLAB software.

The most common causes of blindness in adults worldwide, 2.6% of them are caused by diabetic retinopathy, which is a progressive disease caused by complications of diabetes mellitus. Early detection and prompt treatment help save eyesight. The artificial intelligence technology can provide objective and accurate screening results. Deep learning, especially the Convolutional Neural Network (CNN), which is part of artificial intelligence, has proven successful in solving image problems and is very well used in medical image analysis. CNN works well on large datasets, but it will affect network performance to overfitting in fewer datasets. So to solve the problem of small data, data augmentation techniques can be used. There are various kinds of data augmentation techniques. This study used the CNN method to classify diabetic retinopathy disease, compared several suitable data augmentation techniques for retinal fundus images, and used Contrast Limited Adaptive Histogram Equalization (CLAHE) for image enhancement. This study found that the augmented random zoom technique, together with CLAHE, provided the best accuracy of 98% with 96% sensitivity and 100% specificity.

Highlights from the Diabetes UK Professional Conference 2017

Diabetes is a chronic disease affecting millions of people worldwide, and more than 25% of them have diabetic retinopathy (DR). DR is the leading cause of blindness in adults. DR is usually diagnosed with a biomicroscopic examination of the fundus after pupillary dilation, supplemented by fundus photographs. The challenge lies in the interpretation of these images by a specialist physician. In this work we propose a computer-assisted pipeline for the diagnosis of DR using the convolutional neural network (CNN). This pipeline has three main stages: (i) image preprocessing, (ii) feature extraction and (iii) classification. First, we used the Discrete Wavelet Transform (DWT) method for edge segmentation, and the Contrast Limited Adaptive Histogram Equalization (CLAHE) method to improve image contrast. In the second and third step we used the ResNet50 convolutional neural network for the detection and classification of five levels of disease severity (0 - No DR, 1 - Mild, 2 - Moderate, 3 - Severe, 4 - Proliferative) on 35 122 images from the publicly available Kaggle database.To analyze the performance of the learning model, different metrics were used to detect different diseases, such as sensitivity, specificity, and the confusion matrix which includes the number of true positive, false positive, true negative and false negative.In conclusion, we achieved 85% accuracy in the validation phase and 81% in the testing phase, demonstrating the reliability of CNNs to identify and automate the diagnosis of diabetic retinopathy using digital fundus images.

To evaluate the ability of trained nonphysician retinal imagers to perform diabetic retinopathy (DR) evaluation at the time of ultrawide field retinal (UWF) imaging in a teleophthalmology program. Clinic patients with diabetes received Joslin Vision Network protocol retinal imaging as part of their standard medical care. Retinal imagers evaluated UWF images for referable DR at the time of image capture. Training of the imagers included 4 h of standardized didactic lectures and 12 h of guided image review. Real-time evaluations were compared with standard masked gradings performed at a centralized reading center. A total of 3,978 eyes of 1,989 consecutive patients were imaged and evaluated. By reading center evaluation, 3,769 eyes (94.7%) were gradable for DR, 1,376 (36.5%) had DR, and 580 (15.3%) had referable DR. Compared with the reading center, real-time image evaluation had a sensitivity and specificity for identifying more than minimal DR of 0.95 (95% CI 0.94-0.97) and 0.84 (0.82-0.85), respectively, and 0.99 (0.97-1.00) and 0.76 (0.75-0.78), respectively, for detecting referable DR. Only three patients with referable DR were not identified by imager evaluation. Point-of-care evaluation of UWF images by nonphysician imagers following standardized acquisition and evaluation protocols within an established teleophthalmology program had good sensitivity and specificity for detection of DR and for identification of referable retinal disease. With immediate image evaluation, <0.1% of patients with referable DR would be missed, reading center image grading burden would be reduced by 60%, and patient feedback would be expedited.

Incidence of diabetic eye disease among migrants: A cohort study of 100,000 adults with diabetes in Denmark

Purpose:To assess the diagnostic accuracy measures such as sensitivity and specificity of smartphone-based artificial intelligence (AI) approaches in the detection of diabetic retinopathy (DR).Methods:A literature search of the EMBASE and MEDLINE databases (up to March 2020) was conducted. Only studies using both smartphone-based cameras and AI software for image analysis were included. The main outcome measures were pooled sensitivity and specificity, diagnostic odds ratios and relative risk of smartphone-based AI approaches in detecting DR (of all types), and referable DR (RDR) (moderate nonproliferative retinopathy or worse and/or the presence of diabetic macular edema).Results:Smartphone-based AI has a pooled sensitivity of 89.5% (95% confidence interval [CI]: 82.3%–94.0%) and pooled specificity of 92.4% (95% CI: 86.4%–95.9%) in detecting DR. For referable disease, sensitivity is 97.9% (95% CI: 92.6%-99.4%), and the pooled specificity is 85.9% (95% CI: 76.5%–91.9%). The technology is better at correctly identifying referable retinopathy.Conclusions:The smartphone-based AI programs demonstrate high diagnostic accuracy for the detection of DR and RDR and are potentially viable substitutes for conventional diabetic screening approaches. Further, high-quality randomized controlled trials are required to establish the effectiveness of this approach in different populations.

Diabetic retinopathy detection through convolutional neural networks with synaptic metaplasticity

The prospective correlation between subclinical atherosclerosis and diabetic retinopathy (DR) incidence in Chinese patients with type 2 diabetes mellitus (T2DM) remains elusive. Prospective data were obtained from 2781 patients with diabetes, among whom 1,964 and 2,180 T2DM patients without any and referable DR at baseline, respectively, were included in the analysis. Multivariate analyses were performed using the Cox proportional hazards model. Over a median follow-up of 22.2months (interquartile range 12.7-27.7), 282 (14.36%) and 125 (5.73%) patients developed any and referable DR, respectively. After adjustment for confounders, each standard deviation (SD) increase in brachial-ankle pulse wave velocity (ba-PWV) was associated with 31% (95% confidence interval 1.15-1.50) and 38% (1.14-1.66) higher risks of incident any and referable DR, respectively. Compared with the lowest ba-PWV quartile, the highest ba-PWV quartile had 135% (1.48-3.72) and 293% (1.83-8.44) higher risks of developing any and referable DR, respectively. Per SD increase of pulse pressure (PP) was associated with 22% (1.09-1.38) and 22% (1.02-1.46) higher risks of incident any and referable DR, respectively. The restricted cubic spline models further indicated a significant linear association of baseline subclinical atherosclerosis with referable DR, and a nonlinear association with any DR. In addition, adding the ba-PWV to the prognostic model for DR incidence improved the C-statistic value, the integrated discrimination improvement value, and the net reclassification improvement value (all P < 0.05). Baseline subclinical atherosclerosis was significantly associated with an increased risk of DR incidence, and elevated ba-PWV independently predicted incident DR in T2DM patients.

This article demonstrates a revolutionary deep- learning approach for automated diabetic retinopathy (DR) detection using convolutional neural networks (CNNs). With the help of datasets from the Diabetic Retinopathy Detection 2015 and APTOS 2019 Blindness Detection competitions, our model attempts to accurately classify retinal images into two categories: No DR and Proliferative DR, based on the severity of macular degeneration. We enhance the quality of the dataset and the robustness of the model by carefully preprocessing and augmenting it using various approaches like rotation, flipping, shearing, rescaling, translation, brightness correction, and normalization. We train two models: a regression model and a deep learning model. With an 88% validation accuracy, we demonstrated that our deep learning technique performs better than the regression model. In order to properly determine and classify DR severity levels, this study presents a robust and noise-tolerant system that evolves the field of automated DR detection. Our approach enables early detection and intervention by combining cutting-edge CNN architectures with sophisticated preprocessing techniques, which may lessen the strain of DR- related blindness on healthcare systems. We also demonstrate out drawbacks and suggest areas for additional investigation, such as analyzing ensemble learning and using a variety of datasets to improve the generalizability of the model. As we wrap up, our study emphasizes the potential of deep learning techniques to transform DR diagnosis and treatment, with the goal of enhancing the quality of care for diabetics globally.

Diabetic retinopathy (DR) is one of the most threatening complications in diabetic patients, leading to permanent blindness without timely treatment. However, DR screening is not only a time-consuming task that requires experienced ophthalmologists but also easy to produce misdiagnosis. In recent years, deep learning techniques based on convolutional neural networks have attracted increasing research attention in medical image analysis, especially for DR diagnosis. However, dataset labeling is expensive work and it is necessary for existing deep-learning-based DR detection models. For this study, a novel domain adaptation method (multi-model domain adaptation) is developed for unsupervised DR classification in unlabeled retinal images. At the same time, it only exploits discriminative information from multiple source models without access to any data. In detail, we integrate a weight mechanism into the multi-model-based domain adaptation by measuring the importance of each source domain in a novel way, and a weighted pseudo-labeling strategy is attached to the source feature extractors for training the target DR classification model. Extensive experiments are performed on four source datasets (DDR, IDRiD, Messidor, and Messidor-2) to a target domain APTOS 2019, showing that MMDA produces competitive performance for present state-of-the-art methods for DR classification. As a novel DR detection approach, this article presents a new domain adaptation solution for medical image analysis when the source data is unavailable.

Automatic Detection and Classification of Diabetic Retinopathy from Retinal Fundus Images by Abdullah Biran, Master of Applied Science, lectrical and computer engineering Department, Ryerson University, 2017. Diabetic Retinopathy (DR) is an eye disease that leads to blindness when it progresses to proliferative level. The earliest signs of DR are the appearance of red and yellow lesions on the retina called hemorrhages and exudates. Early diagnosis of DR prevents from blindness. In this thesis, an automatic algorithm for detecting diabetic retinopathy is presented. The algorithm is based on combination of several image processing techniques including Circular Hough Transform (CHT), Contrast Limited Adaptive Histogram Equalization (CLAHE), Gabor filter and thresholding. In addition, Support Vector Machine (SVM) classifier is used to classify retinal images into normal or abnormal cases of DR including non-proliferative (NPDR) or proliferative diabetic retinopathy (PDR). The proposed method has been tested on fundus images from Standard Diabetic Retinopathy Database (DIARETDB). The implementation of the presented methodology was done in MATLAB. The methodology is tested for sensitivity and accuracy.

This paper presents a comparative analysis of fundus image enhancement techniques to detect diabetic retinopathy (DR). Medical images frequently suffers from non-uniform illumination, poor contrast and noise, thus these images have to go through pre-processing stage. For image enhancement many techniques are proposed based on spatial domain (histogram). However, these methods usually decline to produce suitable consequences for non-uniform illumination and wide-ranging of low-contract. The comparative analysis and performance evaluation of various enhancement techniques will help in choosing most suitable technique which may significantly improve the detection of diabetic retinopathy. In this paper histogram equalization (HE), adaptive histogram equalization (ADHE), contrast limited adaptive histogram equalization (CLAHE) and exposure based sub-image histogram equalization (ESIHE) techniques are compared for pre-processing of fundus image for detecting DR. For fair analysis of these techniques, histogram, SNR, entropy, absolute mean brightness error and peak signal-to-noise ratio (PSNR) of fundus images are analyzed by using MATLAB.