Fundus cameras are widely used by ophthalmologists for monitoring and diagnosing retinal pathologies. Unfortunately, no optical system is perfect, and the visibility of retinal images can be greatly degraded due to the presence of problematic illumination, intraocular scattering, or blurriness caused by sudden movements. To improve image quality, different retinal image restoration/enhancement techniques have been developed, which play an important role in improving the performance of various clinical and computer-assisted applications. This paper gives a comprehensive review of these restoration/enhancement techniques, discusses their underlying mathematical models, and shows how they may be effectively applied in real-life practice to increase the visual quality of retinal images for potential clinical applications including diagnosis and retinal structure recognition. All three main topics of retinal image restoration/enhancement techniques, i.e., illumination correction, dehazing, and deblurring, are addressed. Finally, some considerations about challenges and the future scope of retinal image restoration/enhancement techniques will be discussed.

UNASSIGNED: We developed an Infant Retinal Intelligent Diagnosis System (IRIDS), an automated system to aid early diagnosis and monitoring of infantile fundus diseases and health conditions to satisfy urgent needs of ophthalmologists.
UNASSIGNED: We developed IRIDS by combining convolutional neural networks and transformer structures, using a dataset of 7697 retinal images (1089 infants) from four hospitals. It identifies nine fundus diseases and conditions, namely, retinopathy of prematurity (ROP) (mild ROP, moderate ROP, and severe ROP), retinoblastoma (RB), retinitis pigmentosa (RP), Coats disease, coloboma of the choroid, congenital retinal fold (CRF), and normal. IRIDS also includes depth attention modules, ResNet-18 (Res-18), and Multi-Axis Vision Transformer (MaxViT). Performance was compared to that of ophthalmologists using 450 retinal images. The IRIDS employed a five-fold cross-validation approach to generate the classification results.
UNASSIGNED: Several baseline models achieved the following metrics: accuracy, precision, recall, F1-score (F1), kappa, and area under the receiver operating characteristic curve (AUC) with best values of 94.62% (95% CI, 94.34%-94.90%), 94.07% (95% CI, 93.32%-94.82%), 90.56% (95% CI, 88.64%-92.48%), 92.34% (95% CI, 91.87%-92.81%), 91.15% (95% CI, 90.37%-91.93%), and 99.08% (95% CI, 99.07%-99.09%), respectively. In comparison, IRIDS showed promising results compared to ophthalmologists, demonstrating an average accuracy, precision, recall, F1, kappa, and AUC of 96.45% (95% CI, 96.37%-96.53%), 95.86% (95% CI, 94.56%-97.16%), 94.37% (95% CI, 93.95%-94.79%), 95.03% (95% CI, 94.45%-95.61%), 94.43% (95% CI, 93.96%-94.90%), and 99.51% (95% CI, 99.51%-99.51%), respectively, in multi-label classification on the test dataset, utilizing the Res-18 and MaxViT models. These results suggest that, particularly in terms of AUC, IRIDS achieved performance that warrants further investigation for the detection of retinal abnormalities.
UNASSIGNED: IRIDS identifies nine infantile fundus diseases and conditions accurately. It may aid non-ophthalmologist personnel in underserved areas in infantile fundus disease screening. Thus, preventing severe complications. The IRIDS serves as an example of artificial intelligence integration into ophthalmology to achieve better outcomes in predictive, preventive, and personalized medicine (PPPM / 3PM) in the treatment of infantile fundus diseases.
UNASSIGNED: The online version contains supplementary material available at 10.1007/s13167-024-00350-y.

UNASSIGNED: The precise identification of retinal disorders is of utmost importance in the prevention of both temporary and permanent visual impairment. Prior research has yielded encouraging results in the classification of retinal images pertaining to a specific retinal condition. In clinical practice, it is not uncommon for a single patient to present with multiple retinal disorders concurrently. Hence, the task of classifying retinal images into multiple labels remains a significant obstacle for existing methodologies, but its successful accomplishment would yield valuable insights into a diverse array of situations simultaneously.
UNASSIGNED: This study presents a novel vision transformer architecture called retinal ViT, which incorporates the self-attention mechanism into the field of medical image analysis. To note that this study supposed to prove that the transformer-based models can achieve competitive performance comparing with the CNN-based models, hence the convolutional modules have been eliminated from the proposed model. The suggested model concludes with a multi-label classifier that utilizes a feed-forward network architecture. This classifier consists of two layers and employs a sigmoid activation function.
UNASSIGNED: The experimental findings provide evidence of the improved performance exhibited by the suggested model when compared to state-of-the-art approaches such as ResNet, VGG, DenseNet, and MobileNet, on the publicly available dataset ODIR-2019, and the proposed approach has outperformed the state-of-the-art algorithms in terms of Kappa, F1 score, AUC, and AVG.

UNASSIGNED: The purpose of this study is to identify predictive activation biomarkers in retinal microvascular characteristics of non-exudative macular neovascularization (MNV) and avoid delayed treatment or overtreatment of subclinical MNV. The main objective is to contribute to the international debate on a new understanding of the role of retinal vessel features in the pathogenesis and progression of non-exudative MNV and age-related macular degeneration (AMD). A discussion on revising-related clinical protocols is presented.
UNASSIGNED: In this retrospective study, the authors included eyes with non-exudative MNV, eyes with exudative AMD, and normal eyes of age-matched healthy subjects. The parameters were obtained by optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA).
UNASSIGNED: In total, 21 eyes with exudative AMD, 21 eyes with non-exudative MNV, and 20 eyes of 20 age-matched healthy subjects without retinal pathology were included. Vessel density (VD) of the deep vascular complex (DVC) in eyes with non-exudative MNV was significantly greater than that in eyes with exudative AMD (p = 0.002), while for superficial vascular plexus (SVP) metrics, no VD differences among sectors were observed between eyes with non-exudative MNV and eyes with exudative AMD.
UNASSIGNED: The reduction in retinal vessel density, especially in the DVC, seems to be involved in or be accompanied by non-exudative MNV activation and should be closely monitored during follow-up visits in order to ensure prompt anti-angiogenic therapy. A discussion on applicable clinical protocols is presented aiming to contribute to new insights into ophthalmology service development which is directed to this specific type of patient and diagnosis.

Retinal blood vessel segmentation is a valuable tool for clinicians to diagnose conditions such as atherosclerosis, glaucoma, and age-related macular degeneration. This paper presents a new framework for segmenting blood vessels in retinal images. The framework has two stages: a multi-layer preprocessing stage and a subsequent segmentation stage employing a U-Net with a multi-residual attention block. The multi-layer preprocessing stage has three steps. The first step is noise reduction, employing a U-shaped convolutional neural network with matrix factorization (CNN with MF) and detailed U-shaped U-Net (D_U-Net) to minimize image noise, culminating in the selection of the most suitable image based on the PSNR and SSIM values. The second step is dynamic data imputation, utilizing multiple models for the purpose of filling in missing data. The third step is data augmentation through the utilization of a latent diffusion model (LDM) to expand the training dataset size. The second stage of the framework is segmentation, where the U-Nets with a multi-residual attention block are used to segment the retinal images after they have been preprocessed and noise has been removed. The experiments show that the framework is effective at segmenting retinal blood vessels. It achieved Dice scores of 95.32, accuracy of 93.56, precision of 95.68, and recall of 95.45. It also achieved efficient results in removing noise using CNN with matrix factorization (MF) and D-U-NET according to values of PSNR and SSIM for (0.1, 0.25, 0.5, and 0.75) levels of noise. The LDM achieved an inception score of 13.6 and an FID of 46.2 in the augmentation step.

UNASSIGNED: Due to limited imaging conditions, the quality of fundus images is often unsatisfactory, especially for images photographed by handheld fundus cameras. Here, we have developed an automated method based on combining two mirror-symmetric generative adversarial networks (GANs) for image enhancement.
UNASSIGNED: A total of 1047 retinal images were included. The raw images were enhanced by a GAN-based deep enhancer and another methods based on luminosity and contrast adjustment. All raw images and enhanced images were anonymously assessed and classified into 6 levels of quality classification by three experienced ophthalmologists. The quality classification and quality change of images were compared. In addition, image-detailed reading results for the number of dubiously pathological fundi were also compared.
UNASSIGNED: After GAN enhancement, 42.9% of images increased their quality, 37.5% remained stable, and 19.6% decreased. After excluding the images at the highest level (level 0) before enhancement, a large number (75.6%) of images showed an increase in quality classification, and only a minority (9.3%) showed a decrease. The GAN-enhanced method was superior for quality improvement over a luminosity and contrast adjustment method (P＜0.001). In terms of image reading results, the consistency rate fluctuated from 86.6% to 95.6%, and for the specific disease subtypes, both discrepancy number and discrepancy rate were less than 15 and 15%, for two ophthalmologists.
UNASSIGNED: Learning the style of high-quality retinal images based on the proposed deep enhancer may be an effective way to improve the quality of retinal images photographed by handheld fundus cameras.

Background and Objectives: Glaucoma is a major cause of irreversible visual impairment and blindness, so its timely detection is crucial. Retinal images from diabetic retinopathy screening programmes (DRSP) provide an opportunity to detect undiagnosed glaucoma. Our aim was to find out which retinal image indicators are most suitable for referring DRSP patients for glaucoma assessment and to determine the glaucoma detection potential of Slovenian DRSP. Materials and Methods: We reviewed retinal images of patients from the DRSP at the University Medical Centre Ljubljana (November 2019-January 2020, May-August 2020). Patients with at least one indicator and some randomly selected patients without indicators were invited for an eye examination. Suspect glaucoma and glaucoma patients were considered accurately referred. Logistic regression (LOGIT) with patients as statistical units and generalised estimating equation with logistic regression (GEE) with eyes as statistical units were used to determine the referral accuracy of indicators. Results: Of the 2230 patients reviewed, 209 patients (10.1%) had at least one indicator on a retinal image of either one eye or both eyes. A total of 149 (129 with at least one indicator and 20 without) attended the eye exam. Seventy-nine (53.0%) were glaucoma negative, 54 (36.2%) suspect glaucoma, and 16 (10.7%) glaucoma positive. Seven glaucoma patients were newly detected. Neuroretinal rim notch predicted glaucoma in all cases. The cup-to-disc ratio was the most important indicator for accurate referral (odds ratio 7.59 (95% CI 3.98-14.47; p < 0.001) and remained statistically significant multivariably. Family history of glaucoma also showed an impact (odds ratio 3.06 (95% CI 1.02-9.19; p = 0.046) but remained statistically significant only in the LOGIT multivariable model. Other indicators and confounders were not statistically significant in the multivariable models. Conclusions: Our results suggest that the neuroretinal rim notch and cup-to-disc ratio are the most important for accurate glaucoma referral from retinal images in DRSP. Approximately half of the glaucoma cases in DRSPs may be undiagnosed.

Diabetic retinopathy (DR) will cause blindness if the detection and treatment are not carried out in the early stages. To create an effective treatment strategy, the severity of the disease must first be divided into referral-warranted diabetic retinopathy (RWDR) and non-referral diabetic retinopathy (NRDR). However, there are usually no sufficient fundus examinations due to lack of professional service in the communities, particularly in the developing countries. In this study, we introduce UGAN_Resnet_CBAM (URNet; UGAN is a generative adversarial network that uses Unet for feature extraction), a two-stage end-to-end deep learning technique for the automatic detection of diabetic retinopathy. The characteristics of DDR fundus data set were used to design an adaptive image preprocessing module in the first stage. Gradient-weighted Class Activation Mapping (Grad-CAM) and t-distribution and stochastic neighbor embedding (t-SNE) were used as the evaluation indices to analyze the preprocessing results. In the second stage, we enhanced the performance of the Resnet50 network by integrating the convolutional block attention module (CBAM). The outcomes demonstrate that our proposed solution outperformed other current structures, achieving 94.5% and 94.4% precisions, and 96.2% and 91.9% recall for NRDR and RWDR, respectively.

Eye diseases that are common and many diseases that result in visual ailments, such as diabetes and vascular disease, can be diagnosed through retinal imaging. The enhancement of retinal images often helps in diagnosing diseases related to retinal organ failure. However, today\'s image enhancement methods may lead to artificial boundaries, sudden color gradation, and the loss of image details. Therefore, to prevent these side effects, a new method of retinal image enhancement is proposed. In this work, we propose a new method for enhancing the overall contrast of colored retinal images. That is, we propose low-light image enhancement using a new retinex method based on a powerful semidecoupled retinex method. In particular, illumination layer I gradually approximates the S input image according to the file. This leads to a complete Gaussian transformation model, while the R-layer reflectance is estimated jointly by S and intermediary by I to suppress image noise simultaneously during R estimation on the publicly available Messidor database. From our assessment measurements (PSNR and SSIM), we show that this proposed method is effective in comparison with the relevant and recently proposed retinal imaging methods; moreover, the color, which is determined by the data, does not change the image structure. Finally, a technique is presented to improve the pronounced color of a retinal image, which is useful for ophthalmologists to screen for retinal disease more effectively. Moreover, this technique can be used in the development of robotics for imaging tests to search for clinical markers.

UNASSIGNED: The curation of images using human resources is time intensive but an essential step for developing artificial intelligence (AI) algorithms. Our goal was to develop and implement an AI algorithm for image curation in a high-volume setting. We also explored AI tools that will assist in deploying a tiered approach, in which the AI model labels images and flags potential mislabels for human review.
UNASSIGNED: Implementation of an AI algorithm.
UNASSIGNED: Seven-field stereoscopic images from multiple clinical trials.
UNASSIGNED: The 7-field stereoscopic image protocol includes 7 pairs of images from various parts of the central retina along with images of the anterior part of the eye. All images were labeled for field number by reading center graders. The model output included classification of the retinal images into 8 field numbers. Probability scores (0-1) were generated to identify misclassified images, with 1 indicating a high probability of a correct label.
UNASSIGNED: Agreement of AI prediction with grader classification of field number and the use of probability scores to identify mislabeled images.
UNASSIGNED: The AI model was trained and validated on 17 529 images and tested on 3004 images. The pooled agreement of field numbers between grader classification and the AI model was 88.3% (kappa, 0.87). The pooled mean probability score was 0.97 (standard deviation [SD], 0.08) for images for which the graders agreed with the AI-generated labels and 0.77 (SD, 0.19) for images for which the graders disagreed with the AI-generated labels (P < 0.0001). Using receiver operating characteristic curves, a probability score of 0.99 was identified as a cutoff for distinguishing mislabeled images. A tiered workflow using a probability score of < 0.99 as a cutoff would include 27.6% of the 3004 images for human review and reduce the error rate from 11.7% to 1.5%.
UNASSIGNED: The implementation of AI algorithms requires measures in addition to model validation. Tools to flag potential errors in the labels generated by AI models will reduce inaccuracies, increase trust in the system, and provide data for continuous model development.