OculusNet: Detection of retinal diseases using a tailored web-deployed neural network and saliency maps for explainable AI

Retinal diseases become more complex for people at any age. In the early stages, many people suffering from retinal diseases have very mild symptoms. It is observed that retinal diseases mainly damage the blood vessels that cause the leakage of fluid. The accretion of fluid can impact the retina and cause vision changes. Retinal diseases such as diabetic retinopathy (DR), age-related macular degeneration (AMD), diabetic macular edema (DME), drusen, and choroidal neovascularization (CNV) are complex diseases that show a huge impact on human retinal health. If the patients are not detected with retinal diseases in the early stages, this may lead to permanent vision loss. These diseases have a lot of other side effects, such as brain disorders. Prevention of these diseases may stop permanent vision loss in patients. Machine learning (ML) algorithms are most widely used to detect retinal diseases in their early stages. The main disadvantage of ML is that these algorithms will take more time to process the data and select complex methods. In this paper, deep learning (DL) algorithms are discussed, and various optical coherence tomography (OCT) datasets are used for experiments. Experimental results show the performance of various ML and DL approaches applied to OCT and retinal image datasets. The performance and comparison of various algorithms is also discussed in this paper.

To investigate morphological variations in the macular area with optical coherence tomography (OCT) after vitrectomy for diabetic fibrovascular proliferation. We reviewed 108 cases using OCT 7-15 months after vitrectomy. Of these, 32 received OCT within 3 months postoperatively. Morphological variations were categorized and correlated with visual outcome. Only 24 cases (21.4%) had no obvious abnormalities. The most frequent findings were epiretinal membrane (52.8%), macular thickening (37.0%) and macular cysts (28.7%). Multivariate regression showed that diffuse macular thickening, loss of foveal depression and diffuse retinal thinning were significantly associated with poor visual acuity. Sequential OCT (< 3 and >or= 7 months) revealed that epiretinal membrane and oedema outside of fovea changed significantly between two examinations. OCT may identify diverse morphological changes in the macular area after diabetic vitrectomy for fibrovascular proliferation. Macular appearance may change over time, and certain types of morphological changes may be associated with poor visual function.

Classifying retinal diseases is a complex problem because the early problematic areas of retinal disorders are quite small and conservative. In recent years, Transformer architectures have been successfully applied to solve various retinal related health problems. Age-related macular degeneration (AMD) and diabetic macular edema (DME), two prevalent retinal diseases, can cause partial or total blindness. Diseases therefore require an early and accurate detection. In this study, we proposed Vision Transformer (ViT), Tokens-To-Token Vision Transformer (T2T-ViT) and Mobile Vision Transformer (Mobile-ViT) algorithms to detect choroidal neovascularization (CNV), drusen, and diabetic macular edema (DME), and normal using optical coherence tomography (OCT) images. The predictive accuracies of ViT, T2T-ViT and Mobile-ViT achieved on the dataset for the classification of OCT images are 95.14%, 96.07% and 99.17% respectively. Experimental results obtained from ViT approaches showed that Mobile-ViT have superior performance with regard to classification accuracy in comparison with the others. Overall, it has been observed that ViT architectures have the capacity to classify with high accuracy in the diagnosis of retinal diseases.

Artificial intelligence (AI) has revolutionized optical coherence tomography (OCT) imaging for the better detection of retinal diseases like diabetic retinopathy (DR), age-related macular degeneration (AMD), and glaucoma. Here, bibliometric analysis of AI in OCT is presented. We conducted a systematic search in the Web of Science Core Collection (WoSCC) database up to January 22, 2025 with the help of AI and OCT-related terms. The journal articles and reviews on the topic were examined based on publication trends, author productivity, institution affiliation, and keyword co-occurrence. Visualization was performed through VOSviewer software (1.6.20). A total of 765 articles were analyzed with an increasing trend in publications, led by the USA (30.7%), China (23.3%), and the UK (11.1%) in research production. The most cited institution was the Medical University of Vienna (9.0%), and the most productive journal was Scientific Reports (7.3%). The most common keywords were "Optical Coherence Tomography", "Deep Learning", and "Diabetic Retinopathy" with frequency of 469, 297, 221, respectively. Our review of AI-enhanced OCT for retinal diseases was both comprehensive and systematic, identifying key trends and current research hotspots. As research productivity in this field continues to grow, the focus is increasingly shifting toward developing more accurate AI-driven imaging techniques to improve the diagnosis of various retinal diseases. Future work should prioritize algorithm validation and clinical implementation to facilitate widespread adoption

Retinal disease classification is a significant problem in computer-aided diagnosis (CAD) for medical applications. This paper is focused on a 4-class classification problem to automatically detect choroidal neovascularization (CNV), diabetic macular edema (DME), DRUSEN, and NORMAL in optical coherence tomography (OCT) images. The proposed classification algorithm adopted an ensemble of four classification model instances to identify retinal OCT images, each of which was based on an improved residual neural network (ResNet50). The experiment followed a patient-level 10-fold cross-validation process, on development retinal OCT image dataset. The proposed approach achieved 0.973 (95% confidence interval [CI], 0.971-0.975) classification accuracy, 0.963 (95% CI, 0.960-0.966) sensitivity, and 0.985 (95% CI, 0.983-0.987) specificity at the B-scan level, achieving a matching or exceeding performance to that of ophthalmologists with significant clinical experience. Other performance measures used in the study were the area under receiver operating characteristic curve (AUC) and kappa value. The observations of the study implied that multi-ResNet50 ensembling was a useful technique when the availability of medical images was limited. In addition, we performed qualitative evaluation of model predictions, and occlusion testing to understand the decision-making process of our model. The paper provided an analytical discussion on misclassification and pathology regions identified by the occlusion testing also. Finally, we explored the effect of the integration of retinal OCT images and medical history data from patients on model performance.

RETRACTED: Deep learning network (DL-Net) based classification and segmentation of multi-class retinal diseases using OCT scans

To evaluate the optical coherence tomographic (OCT) features of the retina of rats, we compared the OCT images with the histological appearance of normal retinas and retinas with laser-induced choroidal neovascularization. Twelve eyes of 12 adult pigmented rats (Brown-Norway) were used. Color fundus photography, fluorescein angiography (FAG), and OCT images of normal retinas and retinas with laser photocoagulation-induced choroidal neovascularization were studied. OCT showed a double-layered structure in the normal sensory retina with a highly reflective layer located in the inner retina and a low reflective layer located in the outer retina. The retinal pigment epithelium (RPE) and choriocapillaris were imaged as a layer with the highest reflection. On the first day after photocoagulation, OCT showed a disruption of the highly reflective layer corresponding to the RPE, and an enhanced reflectivity in the choroid under the lesion. Choroidal neovascularizations (CNVs) which appeared 2 weeks after photocoagulation was seen as a multi-layered, highly reflective area protruding from the RPE into the subretinal space A CNV beneath a subretinal hematoma was difficult to detect because of the low transmission of the scanning light through the hematoma. The histopathological appearance was well correlated with the OCT images. The two reflective bands in the OCT images were identified as coming from the inner layers of the retina and from the photoreceptors. The highest reflective band arose from the RPE and choriocapillaris. In the future, OCT combined with FAG or indocyanine-green angiography will be a useful tool for the evaluation of animal studies of choroidal neovascularization and other retinal diseases.

ScLNet: A cornea with scleral lens OCT layers segmentation dataset and new multi-task model

High-Definition and 3-dimensional Imaging of Macular Pathologies with High-speed Ultrahigh-Resolution Optical Coherence Tomography

Microaneurysms (MAs) are pathognomonic signs that help clinicians to detect diabetic retinopathy (DR) in the early stages. Automatic detection of MA in retinal images is an active area of research due to its application in screening processes for DR which is one of the main reasons of blindness amongst the working-age population. The focus of these works is on the automatic detection of MAs in en face retinal images like fundus color and Fluorescein Angiography (FA). On the other hand, detection of MAs from Optical Coherence Tomography (OCT) images has 2 main advantages: first, OCT is a non-invasive imaging technique that does not require injection, therefore is safer. Secondly, because of the proven application of OCT in detection of Age-Related Macular Degeneration, Diabetic Macular Edema, and normal cases, thanks to detecting MAs in OCT, extensive information is obtained by using this imaging technique. In this research, the concentration is on the diagnosis of MAs using deep learning in the OCT images which represent in-depth structure of retinal layers. To this end, OCT B-scans should be divided into strips and MA patterns should be searched in the resulted strips. Since we need a dataset comprising OCT image strips with suitable labels and such large labelled datasets are not yet available, we have created it. For this purpose, an exact registration method is utilized to align OCT images with FA photographs. Then, with the help of corresponding FA images, OCT image strips are created from OCT B-scans in four labels, namely MA, normal, abnormal, and vessel. Once the dataset of image strips is prepared, a stacked generalization (stacking) ensemble of four fine-tuned, pre-trained convolutional neural networks is trained to classify the strips of OCT images into the mentioned classes. FA images are used once to create OCT strips for training process and they are no longer needed for subsequent steps. Once the stacking ensemble model is obtained, it will be used to classify the OCT strips in the test process. The results demonstrate that the proposed framework classifies overall OCT image strips and OCT strips containing MAs with accuracy scores of 0.982 and 0.987, respectively.

Background: To design a novel anomaly detection and localization approach using artificial intelligence methods using optical coherence tomography (OCT) scans for retinal diseases. Methods: High-resolution OCT scans from the publicly available Kaggle dataset and a local dataset were used by four state-of-the-art self-supervised frameworks. The backbone model of all the frameworks was a pre-trained convolutional neural network (CNN), which enabled the extraction of meaningful features from OCT images. Anomalous images included choroidal neovascularization (CNV), diabetic macular edema (DME), and the presence of drusen. Anomaly detectors were evaluated by commonly accepted performance metrics, including area under the receiver operating characteristic curve, F1 score, and accuracy. Results: A total of 25,315 high-resolution retinal OCT slabs were used for training. Test and validation sets consisted of 968 and 4000 slabs, respectively. The best performing across all anomaly detectors had an area under the receiver operating characteristic of 0.99. All frameworks were shown to achieve high performance and generalize well for the different retinal diseases. Heat maps were generated to visualize the quality of the frameworks' ability to localize anomalous areas of the image. Conclusions: This study shows that with the use of pre-trained feature extractors, the frameworks tested can generalize to the domain of retinal OCT scans and achieve high image-level ROC-AUC scores. The localization results of these frameworks are promising and successfully capture areas that indicate the presence of retinal pathology. Moreover, such frameworks have the potential to uncover new biomarkers that are difficult for the human eye to detect. Frameworks for anomaly detection and localization can potentially be integrated into clinical decision support and automatic screening systems that will aid ophthalmologists in patient diagnosis, follow-up, and treatment design. This work establishes a solid basis for further development of automated anomaly detection frameworks for clinical use.

Choroid neovascularization (CNV) is caused by new blood vessels growing in the choroid and penetrating the bruch membrane. It is the major cause of vision disability in many retinal diseases. Though anti-vascular endothelial growth factor injection has proved to be effective for treating CNV, treatment planning is essential to ensure the efficacy while reducing the risk. For this purpose, we propose a CNV growth model based on longitudinal optical coherence tomography (OCT) images. The reaction-diffusion model is applied to simulate the growth and shrinkage of CNV volumes, and is solved by using the finite-element method. A fitted curve of the CNV growth/shrinkage rate is obtained by optimizing the growth parameters. Then, the trained parameters are applied to the predicted image to get the simulated image, which is compared with the validated image to evaluate the accuracy of prediction. The proposed method was tested on a dataset with seven patients in which each patient has 12 longitudinal OCT images. The resulted mean dice coefficient is 76.40%±8.20%. The experimental results show a promising step towards the image-guided patient-specific treatment.

Multi-scale deep feature fusion for automated classification of macular pathologies from OCT images

Optical Coherence Tomography (OCT) is a widely used non-invasive imaging modality that enables detailed visual-ization of retinal microstructures and supports early diagnosis of retinal diseases. However, manual interpretation of OCT scans is time-consuming, subjective, and highly dependent on clinical ex-pertise, which limits scalability in real-world healthcare settings. Although deep learning-based approaches have demonstratedstrong performance in automated retinal disease classification, most existing systems lack interpretability, clinical workflow integration, and automated diagnostic reporting. This paper presents a comprehensive Smart OCT-based Reti-nal Disease Diagnosis System that integrates ensemble deep learn-ing with AI-driven intelligent agents to enhance clinical usability.An ensemble model combining EfficientNetB0 and VGG-16 is employed to perform eight-class retinal disease classification,including Age-related Macular Degeneration (AMD), Choroidal Neovascularization (CNV), Central Serous Retinopathy (CSR),Diabetic Macular Edema (DME), Diabetic Retinopathy (DR),Drusen, Macular Hole (MH), and Normal retina. Beyond classi-fication, the system incorporates LangGraph-based AI agents to automatically generate diagnostic reports, provide visual explana-tions using Grad-CAM, and produce patient-friendly summaries.The proposed framework achieves 95.6% classification accu-racy while addressing key limitations of existing OCT-based di-agnostic systems through improved interpretability, automation,and scalability. Extensive experimental validationdemonstratesthe system’s capability to assist ophthalmologists in making fasterand more accurate diagnostic decisions. Index Terms—Optical Coherence Tomography, Retinal DiseaseClassification, Deep Learning, EfficientNet, VGG-16, Ensem-ble Learning, Explainable AI, LangGraph, Medical Imaging,Computer-Aided Diagnosis

Objective To investigate on the image futures of OCT of CNV in pathologic myopia, and to analyze that if the ocular axial length and dilopter had any correlation of development of CNV. Methods Thirty-four patients (sixty eyes) were diagnosed as myopia from November 2004 to June 2006 at FUSHUN OPHTHALMIC HOSPITAL.All patients were examined under A scan、B scan、diopter、FFA and OCT.CNV was detected by OCT and FFA .Record the pathologic change on the macula. The condition of 50 eyes with oc- ular axial length ≥ 26mm diopter ≥-6.00D (test group) was compared with other 10 eyes with axial 0.1).Conclusion The longer of AL and higher of D , the more dangerous of development of CNV. The AL and D have correlation of development of CNV. OCT and FFA both have higher sensitivity on detecting CNV. OCT have characteristic image of CNV in patho- logic myopia and it can find pathological change of CNV in pathologic myopia early, safely and correctly. The results suggest that OCT is a valuable detection for CNV in pathologic myopia. Key words: Pathologic myopia; CNV(choroidal neovascularization); OCT(optical coherence tomog-raphy); FFA(fundus fluorescein angiography)