Explainable AI For Diabetic Retinopathy Detection

To study the effectiveness of two digital 50 degree photographic fields per eye, stored compressed or integrally, in the grading of diabetic retinopathy, in comparison to 35-mm colour slides. Two-field digital non-stereoscopic retinal photographs and two-field 35-mm retinal photographs were made at the same time from patients visiting a diabetic retinopathy outpatient clinic. The digital images were stored integrally (TIFF-file) and in a compressed way (JPEG-file). Two ophthalmologists assessed the photographs in a masked fashion. The results were compared. The sensitivity and specificity for the detection of vision-threatening diabetic retinopathy were calculated, using only the grading of the most affected eye. The differences between the retinopathy gradings of the two kinds of photographs were analysed. The agreement for the grading of DR compared to slides was good, both for the compressed and for the integrally stored images (kappa 0.63-0.68). The sensitivity for the detection of vision-threatening diabetic retinopathy using the JPEG-stored images was 0.72-0.74, specificity 0.93-0.98. The sensitivity for vision-threatening retinopathy detection using the integrally stored images was 0.86-0.92, specificity 0.93. Two-field digital retinal photography is effective in diabetic retinopathy grading and it can replace 35-mm retinal photography. Vision-threatening retinopathy can be detected on the images with reasonable to good sensitivity and specificity. An experienced grader should assess the images. The compression of the digital images seems to have some adverse effect on the detection of diabetic retinopathy.

Aim: This study was conducted to evaluate the knowledge, attitude, and practice of the general health practitioners toward diabetic retinopathy (DR), Kingdom of Saudi Arabia (KSA). Materials and Methods: This was a cross-sectional descriptive study. Responses were obtained using a structured multi-point questionnaire. It included closed-ended questions that were arranged in a Likert's scale. Informed consent and ethical approval were obtained. Data were analyzed using SPSS version 16 statistical software. Results: Of the 180 general practitioners (GPs) who participated in the study, 56.7% were males and 43.3% were females. Of those, 51.1% were practicing in rural areas, and 41.1% had postgraduate training. Participants had good knowledge about the prevalence of diabetes mellitus in KSA (97.2%). The majority (92.8%) agreed that early detection and treatment of DR plays a critical role in the prevention of permanent visual loss. However, participants had good attitude toward screening for DR. However, the practice of fundus examination was poorly mastered by 43.9% of the GPs. Conclusion: This study displays the need for hands-on training of GPs about detection of DR by direct use of ophthalmoscopes. Barriers for ophthalmoscope examination, as perceived, need to be further addressed and evaluated. Furthermore, It is of great importance to improve the screening facilities at the primary health care setting.

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.

Artificial intelligence for diabetic retinopathy screening in Africa

Rapid, bloody, and blinding diabetic retinopathy

IntroductionIn April 2018, the US Food and Drug Administration (FDA) approved the world’s first artificial intelligence (AI) medical device for detecting diabetic retinopathy (DR), the IDx-DR. However, there is a lack of evaluation systems for DR intelligent diagnostic technology.MethodsFive hundred color fundus photographs of diabetic patients were selected. DR severity varied from grade 0 to 4, with 100 photographs for each grade. Following that, these were diagnosed by both ophthalmologists and the intelligent technology, the results of which were compared by applying the evaluation system. The system includes primary, intermediate, and advanced evaluations, of which the intermediate evaluation incorporated two methods. Main evaluation indicators were sensitivity, specificity, and kappa value.ResultsThe AI technology diagnosed 93 photographs with no DR, 107 with mild non-proliferative DR (NPDR), 107 with moderate NPDR, 108 with severe NPDR, and 85 with proliferative DR (PDR). The sensitivity, specificity, and kappa value of the AI diagnoses in the primary evaluation were 98.8%, 88.0%, and 0.89, respectively. According to method 1 of the intermediate evaluation, the sensitivity of AI diagnosis was 98.0%, specificity 97.0%, and the kappa value 0.95. In method 2 of the intermediate evaluation, the sensitivity of AI diagnosis was 95.5%, the specificity 99.3%, and kappa value 0.95. In the advanced evaluation, the kappa value of the intelligent diagnosis was 0.86.ConclusionsThis article proposes an evaluation system for color fundus photograph-based intelligent diagnostic technology of DR and demonstrates an application of this system in a clinical setting. The results from this evaluation system serve as the basis for the selection of scenarios in which DR intelligent diagnostic technology can be applied.

The study aimed to apply to Machine Learning (ML) researchers working in image processing and biomedical analysis who play an extensive role in comprehending and performing on complex medical data, eventually improving patient care. Developing a novel ML algorithm specific to Diabetic Retinopathy (DR) is a challenge and need of the hour. Biomedical images include several challenges, including relevant feature selection, class variations, and robust classification. Although the current research in DR has yielded favourable results, several research issues need to be explored. There is a requirement to look at novel pre-processing methods to discard irrelevant features, balance the obtained relevant features, and obtain a robust classification. This is performed using the Steerable Kernalized Partial Derivative and Platt Scale Classifier (SKPD-PSC) method. The novelty of this method relies on the appropriate non-linear classification of exclusive image processing models in harmony with the Platt Scale Classifier (PSC) to improve the accuracy of DR detection. First, a Steerable Filter Kernel Pre-processing (SFKP) model is applied to the Retinal Images (RI) to remove irrelevant and redundant features and extract more meaningful pathological features through Directional Derivatives of Gaussians (DDG). Next, the Partial Derivative Image Localization (PDIL) model is applied to the extracted features to localize candidate features and suppress the background noise. Finally, a Platt Scale Classifier (PSC) is applied to the localized features for robust classification. For the experiments, we used the publicly available DR detection database provided by Standard Diabetic Retinopathy (SDR), called DIARETDB0. A database of 130 image samples has been collected to train and test the ML-based classifiers. Experimental results show that the proposed method that combines the image processing and ML models can attain good detection performance with a high DR detection accuracy rate with minimum time and complexity compared to the state-of-the-art methods. The accuracy and speed of DR detection for numerous types of images will be tested through experimental evaluation. Compared to state-of-the-art methods, the method increases DR detection accuracy by 24% and DR detection time by 37.

INTRODUCTION: Diabetic retinopathy (DR) is one of the consequences of diabetes which if untreated may lead to loss of vision. Generally, for DR detection, retinal images are obtained using a traditional fundus camera. A recent trend in the acquisition of eye fundus images is the usage of smartphones to acquire images.
 OBJECTIVES: This paper focuses on the study of existing works which incorporated smartphones for obtaining fundus images and various devices available in the market. Also, the common datasets used for carrying out DR detection using smartphone-based fundus images as well as the classification models used for the diagnosis of DR are explored.
 METHODS: A search of information was carried out on articles based on DR detection from fundus images published in the state-of-the-art literatures.
 RESULTS: Majority of the works uses SBFI devices like 20D lens, EyeExaminer etc. to obtain fundus image. The common databases used for the study are EyePACS, Messidor, etc. and the classification models mostly rely on deep learning frameworks.
 CONCLUSION: The use of smartphones for capturing fundus images for DR detection are explored. Smartphone devices, datasets used for the study and currently available classification models for SBFI based DR detection are discussed in detail. This paper portrays various approaches currently being employed in SBFI based DR detection.

Diabetic Retinopathy caused by Diabetes Mellitus is a major vision threatening condition across the global population. Early detection and grading of diabetic retinopathy are pivotal to avoiding the associated vision impairment, and retinal image analysis serves as an effective method for the screening process. Analysing retinal images manually to detect diabetic retinopathy and grade them is a time-consuming pro cess that necessitates the involvement of experts to perform the classification. Computer-aided diagnosis of diabetic retinopathy through retinal image analysis is an effective tool to reduce the time involved in the screening process. This study proposed a convolutional neural network based Selective Feature Map Fusion architecture that leverages the fused feature maps of ResNet50 and EfficientNetV2L networks for the detection and grading of diabetic retinopathy. Feature set of ResNet50 and EfficientNetV2L captures different aspects of the underlying lesion distribution, resulting in a more comprehensive representation of the features. This approach helps the model generalize better to unseen data by incorporating diverse data aspects, thereby preventing overfitting to a specific feature set and developing a model that is more robust to noise. Selective feature fusion helps to reduce the computational overhead of the the feature fusion computation. Comparative analysis against the ResNet50 and EfficientNetV2L models individually revealed the superior performance of the fused model in both detection and grading of diabetic retinopathy. The proposed model attained an excellent level of accuracy of 95.2% in diabetic retinopathy detection with a sensitivity of 95.2% and a specificity of 95.2%, when tested on the IDRiD dataset. The model achieved a classification accuracy of 92% for diabetic retinopathy grading with a sensitivity of 88%, a specificity of 98% and an F1-score of 0.89. Moreover, the fused framework attained superior performance when compared against existing methodologies on the IDRiD dataset. The developed model exhibited robust performance also in the DeepDRiD dataset, and it can be employed successfully in the screening processes for the diagnosis and grading of diabetic retinopathy.

Purpose To investigate prevalence of diabetes (DM), diabetic retinopathy (DR), and areas with highest rates of undetected DR. To quantify and map locations of disparities as they relate to poverty and minority populations. Methods Retrospective cohort study from large regional health data repository (HealthLNK). Geographic Information System (GIS) analysis mapped rates of DM and DR in Chicago area ZIP Codes. Results Of 1,086,921 adults who met the inclusion criteria, 143,790 with DM were identified. ZIP Codes with higher poverty rates were correlated with higher prevalence of DM and DR (Pearson’s correlation coefficient 0.614, p < .05, 0.333, p < .05). Poverty was negatively correlated with likelihood of DR diagnosis (−0.638, p < .05). Relative risks of DM and DR were calculated in each ZIP Code and compared to actual rates. 36 high-risk ZIP Codes had both high-risk of DM and low DR detection. In high-risk ZIP Codes 85.4% of households self-identified as ethnic minority and 33.0% were below the Federal Poverty Level (FPL). Both percentages were significantly higher than the Chicago average of 50.5% minority and 19.9% below FPL (p < .05). 67 ideal ZIP Codes had both low risk of DM and high DR detection. In ideal ZIP Codes 32.6% of households self-identified as minority, and 10.2% were below the FPL (p < .05). Conclusions A health care disparity exists with regards to DM and DR. High-risk ZIP Codes are associated with higher poverty and higher minority population, and they are highly concentrated in just 17% of the ZIP codes in the Chicago area.

There are countless number of people with diabetes around the world. diabetic retinopathy (DR), a major complication of diabetes, is a retinal disease that results in blindness if not treated in the earlier stages. Early detection of diabetic retinopathy in patients can help avoid further loss of eyesight. For early detection, it is important for diabetic patients to undergo medical tests after a frequent amount of times. Diagnosing DR using fundus images is a difficult and time-consuming task which requires professional expertise to find and identify if the DR traits are present in the eye. In our proposed system, we make use of artificial intelligence, convolutional neural networks precisely to make the process of detection and classification of diabetic retinopathy a lot simpler and fast. The proposed model makes use of AlexNet which is a convolutional neural network architecture which is trained based on fundus image database to accurately diagnose DR with minimum efforts. The database contains multiple cases of retinal hemorrhage, microaneurism, cotton wool spots, etc. The retinal image of the patient is fed to the system where initially the image is trimmed to match the input layer of AlexNet and then the neural network extracts features and the model gets trained in order to further classify the diabetic retinopathy patients. The last layer of the convolutional neural network is developed in such a way each segment of image is assigned a specific weight factor where higher the weight, more is the chances of getting severe DR.

PurposeWe studied the validity, usefulness, and relative cost to detect diabetic retinopathy (DR) and sight-threatening DR (STDR) by using a hand-held electrophysiologic tool compared to digital fundus photography.MethodPatients with diabetes attending the screening unit of King Khaled Eye Specialist Hospital, Riyadh, Saudi Arabia were evaluated by “RETeval”, Amsler grid, and digital dilated fundus photography. Fundus images were evaluated by a retina specialist to determine grade of DR. The sensitivity and specificity of STDR and DR screening compared to photography were calculated, as well as “RETeval” combined with Amsler grid testing. The area under the curve (AUC) of “RETeval” screening outcome was calculated.ResultWe analyzed data of 400 diabetic patients. The prevalence of DR of any grade was 48.8% (95% confidence interval [CI], 43.9–53.7) while the prevalence of STDR was 27% (95% CI, 22.6–31.4). The outcome of RETeval test was “fail” (based on 20 μV or more amplitude of electrophysiologic spikes) in 351 (87.8%; 95% CI, 84.5–91.0) eyes. The sensitivity of the device was 95.4% and the specificity was 17.5%. Thus, the sensitivity of sequential testing with RETeval and Amsler grid test was 30.1% and the specificity was 80.1%. The AUCs for STDR and DR in general were 76.6% and 50.6%, respectively.Conclusions“RETeval” is a rapid screening device with excellent sensitivity for detecting STDR. It has potential as a first level screening tool to detect patients who require further evaluation.Translational RelevanceRetinal function, such as electrophysiology, can be used as a new concept for screening for DR.

Highlights from the Diabetes UK Professional Conference 2017

The primary source of vision loss in patients is mainly due to Diabetic retinopathy (DR), caused due to diabetes mellitus. It has become a significant reason for visual impairment among people within 25-74 years of age. If timely medical attention is provided to DR patients, over 90% of people can be saved from vision loss. It's crucial for the early diagnosis of the disease and provide the necessary treatment. The symptoms are more prevalent in type 2 diabetics than associated with type 1 diabetics. Unlike computer-aided diagnosis systems, the traditional procedures of DR detection using fundus photography are both time and cost-consuming. Among the numerous methods for screening and detecting DR, Convolutional Neural Networks are considered extensively in Deep Learning (DL) methods. This review article illustrates the different datasets, pre-processing steps, and DL techniques used in the fundus images for efficient DR detection at an early stage. The main motive of this review article is to provide the research community with an insight into the various pre-processing steps, Public datasets, DL models in DR detection, and some future research directions in this field.

Diabetic retinopathy (DR) is a leading cause of blindness worldwide, affecting millions of dia- betic patients. Early detection and accurate classification of DR stages are crucial for preventing vision loss and improving patient outcomes. This survey paper presents a comprehensive anal- ysis of deep learning approaches for automatic detection and classification of diabetic retinopa- thy from fundus images. The proposed system utilizes Faster Region-based Convolutional Neural Networks (Faster R-CNN) architecture to accurately identify and categorize the sever- ity of DR in retinal images. The methodology encompasses image acquisition using fundus cameras, preprocessing for quality enhancement, deep feature learning for data-driven char- acterizations, and classification into various DR stages including No DR, Mild DR, Moderate DR, Proliferative DR, and Severe DR. This automated approach offers significant advantages over traditional manual screening methods, providing faster, more consistent, and cost-effective diagnosis capabilities. The technical implementation focuses on developing robust algorithms for fundus image analysis, including preprocessing techniques for image enhancement, feature extraction using convolutional neural networks, and classification models optimized for medical imaging appli- cations. The system addresses challenges such as image quality variations, lighting conditions, and anatomical differences across patients. Performance evaluation encompasses accuracy met- rics, sensitivity, specificity, and area under the ROC curve (AUC) to ensure clinical reliability. The proposed Faster R-CNN architecture demonstrates superior performance compared to tra- ditional methods, achieving high accuracy in DR detection while maintaining computational efficiency suitable for clinical deployment. This automated screening system has the potential to significantly improve early detection rates, reduce healthcare costs, and enhance patient care quality in diabetic retinopathy management.