A Privacy-Preserving Collaborative Federated Learning Framework for Detecting Retinal Diseases

Identification and classification of microaneurysms for early detection of diabetic retinopathy

This paper explores the application of deep learning architectures in retinal image analysis, with a particular focus on the early detection of diabetic retinopathy (DR). Leveraging large-scale retinal imaging datasets, the study examines the diagnostic capacity of advanced neural models in identifying subtle pathological markers such as microaneurysms, hemorrhages, and exudates at the earliest clinical stages. Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and hybrid deep learning frameworks are evaluated for their performance in image classification and lesion localization. Results demonstrate that CNN-based architectures achieve an accuracy of 94.2%, with sensitivity and specificity values exceeding 93%, underscoring their robustness in medical image interpretation. Hybrid models integrating CNN and attention mechanisms further enhance feature discrimination, achieving 95.1% accuracy and improved area under the curve (AUC) metrics. RNN-driven architectures exhibit competitive performance, particularly in sequential image feature learning, with overall accuracy surpassing 92%. The findings highlight the transformative role of deep learning in automating ophthalmic diagnostics, enabling scalable, precise, and timely detection of diabetic retinopathy to mitigate preventable vision loss.

Diabetic retinopathy (DR) is the most common complication of diabetes mellitus and one of the leading causes of preventable blindness in the adult working population. Visual loss can be prevented from the early stages of DR, when the treatments are effective. Therefore, early diagnosis is paramount. However, DR may be clinically asymptomatic until the advanced stage, when vision is already affected and treatment may become difficult. For this reason, diabetic patients should undergo regular eye examinations through screening programs. Traditionally, DR screening programs are run by trained specialists through visual inspection of the retinal images. However, this manual analysis is time consuming and expensive. With the increasing incidence of diabetes and the limited number of clinicians and sanitary resources, the early detection of DR becomes non-viable. For this reason, computed-aided diagnosis (CAD) systems are required to assist specialists for a fast, reliable diagnosis, allowing to reduce the workload and the associated costs. We hypothesize that the application of novel, automatic algorithms for fundus image analysis could contribute to the early diagnosis of DR. Consequently, the main objective of the present Doctoral Thesis is to study, design and develop novel methods based on the automatic analysis of fundus images to aid in the screening, diagnosis, and treatment of DR. In order to achieve the main goal, we built a private database and used five retinal public databases: DRIMDB, DIARETDB1, DRIVE, Messidor and Kaggle. The stages of fundus image processing covered in this Thesis are: retinal image quality assessment (RIQA), the location of the optic disc (OD) and the fovea, the segmentation of RLs and EXs, and the DR severity grading. RIQA was studied with two different approaches. The first approach was based on the combination of novel, global features. Results achieved 91.46% accuracy, 92.04% sensitivity, and 87.92% specificity using the private database. We developed a second approach aimed at RIQA based on deep learning. We achieved 95.29% accuracy with the private database and 99.48% accuracy with the DRIMDB database. The location of the OD and the fovea was performed using a combination of saliency maps. The proposed methods were evaluated over the private database and the public databases DRIVE, DIARETDB1 and Messidor. For the OD, we achieved 100% accuracy for all databases except Messidor (99.50%). As for the fovea location, we also reached 100% accuracy for all databases except Messidor (99.67%). The joint segmentation of RLs and EXs was accomplished by decomposing the fundus image into layers. Results were computed per pixel and per image. Using the private database, 88.34% per-image accuracy (ACCi) was reached for the RL detection and 95.41% ACCi for EX detection. An additional method was proposed for the segmentation of RLs based on superpixels. Evaluating this method with the private database, we obtained 84.45% ACCi. Results were validated using the DIARETDB1 database. Finally, we proposed a deep learning framework for the automatic DR severity grading. The method was based on a novel attention mechanism which performs a separate attention of the dark and the bright structures of the retina. The Kaggle DR detection dataset was used for development and validation. The International Clinical DR Scale was considered, which is made up of 5 DR severity levels. Classification results for all classes achieved 83.70% accuracy and a Quadratic Weighted Kappa of 0.78. The methods proposed in this Doctoral Thesis form a complete, automatic DR screening system, contributing to aid in the early detection of DR. In this way, diabetic patients could receive better attention for their ocular health avoiding vision loss. In addition, the workload of specialists could be relieved while healthcare costs are reduced.

Diabetes is one of the most dangerous diseases that is facing by many people in the world. Diabetic Retinopathy (DR) is a retinal disease that is caused by the Diabetes Mellitus (DM). DM causes the lesions on the retinal layer thereby affecting the vision. If it is not detected at the initial stages, it might lead to complete blindness and also DR is an irreversible disease. Hence, early detection of DR is inevitable to avoid the vision loss of the patients. Expeditious detection can decrease the complications of the DR thereby vision of the patient is preserved. It is a laborious task for the ophthalmologists to diagnosis DR manually, since it takes a lot more time and cost-effective. Mis-diagnosis might happen if the ophthalmologists are not skillful and experienced in detecting DR from the fundus images. Over the past two decades, deep learning has shown a significant raise in the bio-medical image processing and its niche areas obtaining the best performance. The state-of-the-art Convolutional Neural Networks (CNN) models achieved notable performance in classifying DR, but the severity levels of DR are not analyzed. To address these challenges, we propose a CNN based model that is used to analyze the fundus oculi retinal images to locate the eyeball structure and observe the presence of DR. The proposed model's hyperparameters are regulated by the transfer learning techniques for mapping the label of the images. The dataset used for training and testing the model is taken from the Kaggle that contains the retinal images and its corresponding severity in a scale. Severity level of the images is classified into five different categories from a normal to DR presented eyeball. The accuracy of the proposed model is 94.92%, proving the confident detection and prediction of the DR from the retinal images.

ABSTRACTThe early detection of diabetic retinopathy (DR) illnesses improves diagnosis and lowers the risk of permanent blindness. As a result, screening for DR in fundus images is an important method used to diagnose diabetes and other eye diseases. However, detecting diseases manually requires a significant amount of time and work. Deep learning (DL) techniques have produced encouraging results in categorizing fundus images. Still, the multi‐class DR disease remains a difficult task. Thus, the proposed framework adopted a novel Criss‐Cross Attention‐Based Squeeze‐and‐Excitation Assisted Vision Transformer (CCA‐SE‐ViT) model to classify DR from fundus images. Initially, the defective region of the retina is segmented using a novel dilated depth‐wise separable convolutional U‐Net model (dDSC‐UNet). Then, using the segmented regions, the fundus images are classified into multiple classes as age‐related macular degeneration (AMD), DR, glaucoma, cataracts, myopia, hypertension, normal, other abnormalities, and DR cases are classified into no DR, mild, moderate, severe, and proliferative DR, respectively. The retinal fundus images are obtained from publicly available datasets like OIA‐ODIR and APTOS 2019. The proposed methodology for multi‐class categorization of retinal illnesses in the OIA‐ODIR dataset yielded 97.2% accuracy, 96.7% precision, 96.1% recall, 95.9% F1‐score, and 96.4% specificity. The APTOS dataset was used for multi‐class classification of DR illnesses, and the results were 99.68% accuracy, 99.08% precision, 99.31% recall, 99.19% F1‐score, and 99.26% specificity. The results demonstrated that the proposed method accurately identifies DR using retinal fundus images.

Diabetic retinopathy is a diabetes complication that has been a leading cause of blindness worldwide. Segmentation of blood vessels and detection of exudates in fundus images through an automated system will help the ophthalmologist to provide a proper treatment that may cure or decrease the severity of the retinal diseases. Realizing its significance, a diabetic retinopathy screening system was proposed in this paper which classifies normal from the abnormal fundus image. Segmentation of blood vessel is based on match filter followed by fuzzy c-means clustering. The small and thin vessels were obtained from directional filter bank by incorporating the use of line-like directional features. In this study a combination of both statistical and geometric features were extracted from image regions. A firefly algorithm for discriminative feature selection for the early detection of diabetic retinopathy was proposed. Publically available dataset such as DRIVE, HRF, DIAREDB1, MESSIDOR and one local dataset are used to validate the suggested system. Moreover, the proposed algorithm has been compared with few state-of-the-art techniques such as particle swarm optimization, genetic algorithm and ant colony optimization. The results demonstrated that the proposed method improves the classification accuracy with a minimized feature set for the early detection of diabetic retinopathy.

It has rarely been studied whether the presence and severity of diabetic retinopathy (DR) could influence the renal disease progression among all chronic kidney disease (CKD) diabetic patients. This study investigates the characteristics of diabetic patients, with different stages of chronic kidney disease (CKD), according to the occurrence of diabetic retinopathy and determines the influence of retinopathy in the deterioration of renal function. We conduct a multicenter, longitudinal cohort study based on the Epidemiology and Risk Factors Surveillance of the CKD project (2008–2013) and the National Health Insurance Research Database (NHIRD) (2001–2013). A total of 4050 diabetic patients with CKD, 20–85 years of age, from 14 hospitals and the community are included in this study. As compared to CKD patients without DR, CKD patients with DR have a lower baseline estimated glomerular filtration rate (eGFR) (39.17 ± 30.36 mL/min per 1.73 m2 vs. 54.38 ± 33.67 mL/min per 1.73 m2 ); poorer glycemic control (higher glycated hemoglobin (HbA1c) 7.85 ± 4.97 vs. 7.29 ± 4.02, p < 0.01); higher proteinuria (urine protein-to-creatinine ratio (UPCR )1.94 ± 2.96 g/dL vs. 0.91 ± 2.11 g/dL, p < 0.01); more anemia (Hb 11.22 ± 2.43 g/dL vs. 12.39 ± 3.85 g/dL, p < 0.01), and more hypoalbuminemia (3.88 ± 0.95 g/dL vs. 4.16 ± 1.74 g/dL, p < 0.01). Later stage (stage 3b–5) CKD patients with DR had significantly higher CKD progression compared with patients without DR (OR (odds ratio) 1.66 (1.36–2.02)). Patients with proliferative DR had significantly higher CKD progression events compared to patients with non-proliferative DR (OR 2.18 (1.71–2.78)). The presence and severity of DR is a risk factor for CKD progression among our Taiwanese CKD patients with diabetes. Prevention and early detection of DR are important and DR should be routinely screened as early as possible among diabetic CKD patients.

Background: Diabetic retinopathy (DR) is a leading cause of blindness among working-age individuals globally. In Kazakhstan, the prevalence of type 2 diabetes mellitus (T2DM) is rising, making DR a significant public health concern. Early detection of DR is challenging due to its asymptomatic onset, necessitating the identification of reliable biomarkers for timely diagnosis. Objective: This study aimed to evaluate the potential of plasma microRNAs (miRNAs) as biomarkers for DR in the Kazakhstani population. Materials and Methods: We conducted a case-control study involving 100 T2DM patients with DR, 98 T2DM patients without DR, and 30 healthy controls. Plasma samples were collected from four medical institutions in Almaty between 2020 and 2021. Quantitative real-time PCR was employed to assess the expression levels of ten candidate miRNAs. Statistical analyses included Mann– Whitney U tests, Pearson’s χ² tests, Spearman’s correlation coefficients, and multiple linear regression. Receiver operating characteristic (ROC) analysis was performed to evaluate the diagnostic performance of miRNAs. Results: The expression level of miR-423-3p was significantly lower in DR patients compared to those without DR (pFDR = 5.4 × 10−3). Additionally, miR-423-3p levels were reduced in DR patients compared to healthy controls (pFDR = 5.4 × 10−3). However, the diagnostic accuracy of miR- 423-3p was limited, with area under the curve (AUC) values ranging from 0.6 to 0.7, indicating poor diagnostic potential. Conclusion: While miR-423-3p demonstrates differential expression in DR patients, its low diagnostic accuracy precludes its use as a standalone biomarker for DR in the Kazakhstani population. Further research is needed to identify more reliable biomarkers for early detection of DR. Acknowledgements: We thank the Science Committee of the Ministry of Education and Science of the Republic of Kazakhstan for funding this research (grant AP08857430). Key words: miR-423-3p, Diabetic Retinopathy, Type 2 Diabetes, Biomarker, Plasma microRNA, Kazakh Population

Background: Diabetic Retinopathy (DR) is a common, potentially blinding and visually disabling complication of diabetes. Early detection of retinopathy and its progression to severity with routine referral for screening by Ophthalmologist can save vision by timely management. Methods: This was a hospital based cross-sectional study done to study the association of glycated hemoglobin (HbA1c), fasting and postprandial blood sugar, serum lipid profile, serum creatinine and urine albumin in diabetic patient with and without DR. A total of 50 patients with Diabetic retinopathy and 50 patients without diabetic retinopathy were enrolled in this study. All patients included were Type 2 DM aged 35 years and above. Results: The mean HbA1c was 8.62±1.5 and 5.54±1.2, total cholesterol was 228.9±63 and 184.9±39.8 mg/dl, serum triglyceride was 226.6±80.7 and 160.8±45.1 mg/dl, LDL-C was 152.3±49 and 127.2±37 mg/dl and serum creatinine was 1.15±0.45 and 0.66±0.27 mg/dl in diabetic retinopathy group and no diabetic retinopathy group respectively. There was significant association of elevated HbA1c, serum triglyceride, LDL- C and total cholesterol with diabetic retinopathy in patients with type 2 DM. The mean values of serum lipoproteins were higher in the diabetic retinopathy group. Conclusion: Elevated fasting and postprandial blood sugar, glycated hemoglobin, total cholesterol, serum triglyceride, LDL-C, serum creatinine and urine albumin were significantly associated with DR in our study. So, all patients with diabetes mellitus should be screened routinely for serum lipoproteins along with blood sugar profile, as it would help in early detection of diabetic retinopathy.

Diabetic retinopathy (DR) is a common and potentially sight-threatening complication of diabetes. Early detection of DR through screening can prevent visual loss. Handheld fundus cameras combined with artificial intelligence (AI) technology may improve DR screening. We evaluated the Aireen AI algorithm's performance in grading DR in fundus images captured by the handheld Optomed Aurora. Two retina specialists and Aireen graded 624 fundus images for DR. Sensitivity, specificity, and predictive values were measured against the ophthalmologists' grading. Overall, 97% of images were sufficient for DR classification. Aireen demonstrated 94.8% sensitivity, 91.4% specificity, and 92.7% diagnostic accuracy for DR. Aireen showed high diagnostic accuracy in detecting DR in Optomed Aurora images, suggesting its potential for effective screening. The validated use of AI with a handheld fundus camera may streamline the screening process, reduce the burden on health care professionals, and improve access to screening and patient outcomes through enhanced diagnostic accuracy.

Diabetic retinopathy (DR) is a microvascular complication of diabetes with a prevalence of ~35%, and is one of the leading causes of visual impairment in people of working age in most developed countries. The earliest stage of DR, non-proliferative DR (NPDR), may progress to sight-threatening DR (STDR). Thus, early detection of DR and active regular screening of patients with diabetes are necessary for earlier intervention to prevent sight loss. While some countries offer systematic DR screening, most nations are reliant on opportunistic screening or do not offer any screening owing to limited healthcare resources and infrastructure. Currently, retinal imaging approaches for DR screening include those with and without mydriasis, imaging in single or multiple fields, and the use of conventional or ultra-wide-field imaging. Advances in telescreening and automated detection facilitate screening in previously hard-to-reach communities. Despite the heterogeneity in approaches to fit local needs, an evidence base must be created for each model to inform practice. In this review, we appraise different aspects of DR screening, including technological advances, identify evidence gaps, and propose several studies to improve DR screening globally, with a view to identifying patients with moderate-to-severe NPDR who would benefit if a convenient treatment option to delay progression to STDR became available.

Diabetic Retinopathy (DR) is a human eye disease which may cause damage to retina of eye, it may lead to complete blindness. Currently diabetic retinopathy is detected using different types of lesions such as Microaneursyms (MAs), Exudates (EXs) and Hemorrhages (HAs). This paper focuses on the identification and classification of lesions for early detection of Diabetic retinopathy through hybrid morphological approach. In the proposed approach the detection of DR is performed through hybrid pre-post processing methods to enhance the lesion detection execution. In addition by using the area of boundaries scheme, the affected regions are identified as initial, middle and severe stage. Performance analysis of the proposed method is demonstrated using Matlab software. Simulation studies on available database prove that the hybrid approach has improved performance over classical approach.

Diabetic retinopathy is one of the leading disabling chronic diseases, and one of the leading causes of preventable blindness in the world. Early diagnosis of diabetic retinopathy enables timely treatment and in order to achieve it a major effort will have to be invested into screening programs and especially into automated screening programs. For automated screening programs to work robustly a representative fundus image database is required. In this paper we give an overview of currently available databases and present a new diabetic retinopathy database. Our database is to our knowledge the first and only database which has diabetic retinopathy pathologies and major fundus structures annotated for every image from the database which makes it perfect for design and evaluation of currently available and new image processing algorithms for early detection of diabetic retinopathy using color fundus images.

Purpose: Diabetic retinopathy is one of the most common complications of diabetes mellitus and remains a leading cause of visual impairment worldwide. Because early-stage diabetic retinopathy is often asymptomatic, diagnosis is frequently delayed, resulting in progression to vision-threatening stages. This education column reviews the importance of screening, summarizes current diagnostic approaches, and outlines contemporary treatment strategies for diabetic retinopathy in routine clinical practice.Current concepts: Regular screening and timely ophthalmic referral are essential for the early detection of diabetic retinopathy. Fundus examination remains the cornerstone of diagnosis, while advances in retinal imaging—including ultra-widefield fundus photography, optical coherence tomography (OCT), fluorescein angiography, and OCT angiography—enable more precise assessment of disease severity, retinal ischemia, and macular involvement. Management strategies are determined according to disease stage and include systemic risk factor control, intravitreal injections, laser photocoagulation, and surgical intervention. Intravitreal anti–vascular endothelial growth factor therapy is the first-line treatment for diabetic macular edema, whereas panretinal photocoagulation remains the standard treatment for proliferative diabetic retinopathy.Discussion and conclusion: Diabetic retinopathy is a progressive disease; however, vision-threatening complications can be substantially reduced through early detection and appropriate management. Regular screening, accurate staging, and timely treatment are key elements in preventing severe visual loss. Close collaboration among primary care physicians, endocrinologists, and ophthalmologists, together with continuous patient education, plays a critical role in improving screening adherence and clinical outcomes. A comprehensive, multidisciplinary approach is essential to reduce the burden of visual impairment caused by diabetic retinopathy.

Purpose: Early detection of diabetic retinopathy can prevent progressive visual loss. Evaluate the role of Optical Coherence Tomography Angiography (OCTA) in early detection of diabetic retinopathy (DR) in type 1 diabetes (T1D) adolescents with normal fundus photography. Methods: Cross-sectional study included 150 eyes of 75 T1D adolescents from the Diabetes Endocrine and Metabolism Pediatric Unit, Children hospital, Cairo University. Their ages ranged from 10-18 years and diabetes duration 2 years or more. Fundus photography, OCT and OCTA were done in the diagnostic and laser unit, Ophthalmology department. Results: Twenty-eight patients (37.3%) had abnormalities in the OCTA despite normal fundus examination, photography and OCT. Diabetes duration, DKA at presentation and frequency, elevated blood pressure, limited joint mobility, elevated HbA1c and dyslipidemia had a negative correlation with OCTA measurements (foveal, parafoveal and perifoveal densities). Patients with abnormal OCTA had significantly higher mean age, higher DKA frequency, and lower height SDS (p=0.034, 0.011 and 0.001 respectively). Univariate and multivariate analysis revealed age and DKA at diabetes presentation are significant predictors for abnormal OCTA. Conclusion: OCTA can be used in early detection of diabetic retinopathy in T1D adolescents especially with risk factors (Elevated HbA1c, frequent DKA, decreased height SDS and limited joint mobility).