A system for detecting retinal diseases and a method thereof

IN604935BActive Publication Date: 2026-10-05SRM UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
IN202441066371
Authority / Receiving Office
IN · IN
Patent Type
Patents
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2026-10-05
Estimated Expiration
2044-09-02

AI Technical Summary

Technical Problem

Traditional methods for detecting retinal diseases, such as manual interpretation of Optical Coherence Tomography (OCT) images and fundus photography, are subjective, labor-intensive, and prone to variability, leading to potential misdiagnoses and disparities in healthcare outcomes, especially in resource-limited settings.

Method used

A system comprising an image acquisition module, data collection module, pre-processing module, feature extraction module, data management module, detection module, diagnosis and report generation module, and user interface module, utilizing convolutional neural networks for automated analysis and classification of retinal images to enhance diagnostic accuracy and efficiency.

Benefits of technology

The system achieves high accuracy in detecting retinal diseases like Central Serous Chorioretinopathy, Choroidal Neovascularization, Diabetic Macular Edema, and Drusen, providing real-time diagnostic results and seamless integration with healthcare systems, reducing reliance on manual expertise and improving accessibility.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

ABSTRACT A SYSTEM FOR DETECTING RETINAL DISEASES AND A METHOD THEREOF The present disclosure discloses a retinal disease detection system (100) and method (200) thereof employing Optical Coherence Tomography (OCT). The system integrates advanced modules to capture, process, and analyze high-resolution retinal images. Key components include an image acquisition module (102) for OCT imaging, a pre-processing module (108) for standardizing data, and a feature extraction module (110) for identifying disease-related features. A data management module (112) utilizes CNN-based transfer learning for model training, validation, and testing. The detection module (114) classifies retinal conditions such as Central Serous Chorioretinopathy (CSC), Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), Drusen, or normal states, supported by diagnostic analysis. Diagnostic reports are generated via a diagnosis and report generation module (116) with real-time results displayed on a user interface (120). A communication module (118) facilitates data sharing and model updates. The system enhances clinical workflows, offering precise, real-time diagnostic capabilities and seamless integration with healthcare platforms.
Need to check novelty before this filing date? Find Prior Art

Description

FIELDThe present disclosure relates to advanced systems for disease detection and diagnosis.BACKGROUNDThe background information herein below relates to the present disclosure but is notnecessarily prior art.Eye infections, prevalent across all age groups, are caused by infectious agents such asviruses, bacteria, and fungi, typically impacting the eyelids, cornea, and conjunctiva.These infections manifest in various conditions, including Conjunctivitis, Dry EyeSyndrome, Cataracts, Glaucoma, Central Serous Chorioretinopathy (CSC), ChoroidalNeovascularization (CNV), Diabetic Macular Edema (DME), and Drusen. CentralSerous Chorioretinopathy (CSC) is characterized by the accumulation of fluid betweenthe retina and the underlying choroid, which may lead to retinal detachment and centralvision impairment. CSC presents in both acute (aCSC) and chronic (cCSC) forms, withacute CSC often resolving spontaneously, while chronic CSC may involve structuralchanges in the retina, such as retinal pigment epithelium alterations, necessitating morecomplex management. Choroidal Neovascularization (CNV), associated with exudativeage-related macular degeneration and other conditions, is marked by distorted visionand color alterations due to abnormal blood vessel growth in the choroid. DiabeticMacular Edema (DME), a complication of diabetes, results from damage to retinalblood vessels, leading to fluid leakage and macular swelling, which can significantlyimpair central vision and potentially cause blindness if untreated. Drusen are tinyyellow or white deposits that form behind the retina, commonly found in adults over theage of 60 as a natural part of the aging process. Although drusen themselves are usuallynot dangerous, larger or more numerous deposits can signal early macular degeneration,increasing the risk of significant vision loss. Given the potential for significant visionloss associated with these conditions, early detection through regular eye examinationsand timely intervention is essential for preventing long-term complications andpreserving visual function.Traditional methods for detecting retinal diseases, such as manual interpretation ofOptical Coherence Tomography (OCT) images and fundus photography, face severallimitations. These approaches heavily depend on the expertise and experience ofclinicians, introducing subjectivity and variability into the diagnostic process. Thesubtle and intricate nature of retinal changes can result in misdiagnosis or delayeddetection, particularly in the early stages of the disease. Moreover, manual analysis islabor-intensive and may be impractical in clinics with high patient volumes. Therequirement for specialized training further restricts the accessibility of these diagnosticservices, especially in resource-limited environments, leading to disparities inhealthcare outcomes.Therefore, there is a need for a system for detecting retinal diseases and a methodthereof that alleviates the aforementioned drawbacks.OBJECTSSome of the objects of the present disclosure, which at least one embodiment hereinsatisfies, are as follows:It is an object of the present disclosure to ameliorate one or more problems of the priorart or to at least provide a useful alternative.An object of the disclosure is to provide a system for detecting retinal diseases.Another object of the disclosure is to enhance the accuracy of retinal disease diagnosis.Still, another object of the disclosure is to enable early detection of retinal diseases,facilitating prompt intervention and potentially preventing vision loss in patients.Yet another object of the disclosure is to minimize dependence on manual examinationby trained ophthalmologists, thereby enhancing efficiency and reducing the risk ofhuman error.Still, another object of the disclosure is to provide a system that streamlines thediagnostic process to enable faster analysis of OCT images, facilitating the managementof larger volumes of cases.Yet another object of the disclosure is to provide a system capable of real-time analysisof retinal images, delivering immediate feedback and facilitating prompt decisionmaking during patient evaluation.Still, another object of the disclosure is to provide a system that standardizes theassessment process, thereby improving diagnostic consistency.Yet another object of the disclosure is to provide a system that can ensure seamlessintegration with existing electronic health record systems and medical imagingplatforms, enabling easy incorporation into current clinical workflows.Still, another object of the disclosure is to provide a method for detecting retinaldiseases.Other objects and advantages of the present disclosure will be more apparent from thefollowing description when read in conjunction with the accompanying figures, whichare not intended to limit the scope of the present disclosure.SUMMARYThe present disclosure provides a system for detecting retinal diseases, the systemcomprising: an image acquisition module, a data collection module, a pre-processingmodule, a feature extraction module, a data management module, a detection module, adiagnosis and report generation module, a communication module, and a user interfacemodule.The image acquisition module is configured to capture high-resolution, cross-sectionalimages of retinal and ocular structures using Optical Coherence Tomography (OCT)imaging techniques.The data collection module is operatively coupled to the image acquisition module andconfigured to collect the captured image data, and store the image data in a repositoryfor subsequent processing;The pre-processing module is configured to receive the collected image data from therepository, standardize and normalize the collected image data using pre-processingtechniques, including noise reduction, image enhancement, and segmentation, andgenerate processed image data suitable for analysis.The feature extraction module is configured to receive the processed image data fromthe pre-processing module to extract structural and texture-based features related toretinal diseases, so as to generate extracted image data for subsequent classification.The data management module comprises:The data splitting module is configured to receive the extracted image data from thefeature extraction module to split the extracted image data to generate training,validation, and testing datasets.The training module is configured to receive the training dataset from the data splittingmodule to train a convolutional neural network (CNN)-based transfer learning model;The validation module is configured to receive the validation dataset from the datasplitting module to monitor and optimize model performance during the training; andThe testing module is configured to receive the test dataset from the data splittingmodule to evaluate the model post-training.The detection module comprises:The classification module employs the trained CNN-based transfer learning model toclassify the extracted image data into classified data of disease categories includingCentral Serous Chorioretinopathy (CSC), Choroidal Neovascularization (CNV),Diabetic Macular Edema (DME), Drusen, or normal condition.The analysis module is configured to receive the classified data from the classifiedmodule to analyse the classified data to identify pathological patterns and generate adisease-specific detection dataset.The diagnosis and report generation module is configured to receive the disease-specificdetection dataset from the analysis module to generate diagnostic outputs includingdetailed reports of detected retinal conditions, provide real-time diagnostic results on auser interface, and facilitate integration with external healthcare systems for streamlinedclinical workflows;The communication module is operatively connected to the detection module anddiagnosis and report generation module, configured to transmit diagnostic data to cloudstorage or telemedicine platforms, and receive updated datasets or algorithms forretraining and model enhancement; andThe user interface module implements visualization tools for displaying diagnosticresults, including heatmaps overlaid on retinal images, and user-configurable optionsfor selecting disease-specific analyses or customizing reports.In an embodiment, the transfer learning model is based on a pre-trained ConvolutionalNeural Network (CNN) architecture, selected from VGG-16, ResNet-50, or similarmodels, and is fine-tuned on retinal OCT datasets.In an embodiment, the pre-processing module employs techniques including:- noise reduction methods including Gaussian filtering and wavelet denoising;- image segmentation and thresholding for retinal layer extraction; and- data augmentation techniques including rotation, scaling, and flipping toincrease the training dataset diversity.In an embodiment, the feature extraction module applies feature extraction techniques,including histogram of oriented gradients (HOG), wavelet transform, texture analysis,and principal component analysis (PCA).In an embodiment, the detection module integrates dropout layers within the CNNarchitecture to reduce overfitting and improve generalization during classification tasks.In an embodiment, the classification module achieves accuracy levels of up to 96.9% byleveraging transfer learning techniques and extensive training on OCT datasets.In an embodiment, the diagnosis and report generation module provides:- real-time or near-real-time diagnostic results during patient evaluation;- seamless integration with electronic health record (EHR) systems forstreamlined clinical workflows; and- patient-friendly reports to improve awareness and engagement.In an embodiment, the data splitting module enables the use of external datasets, such asthe Kaggle OCT dataset and OCT2017 dataset, for training and validation of thetransfer learning model.In an embodiment, the system supports deployment across multiple platforms, includingmobile applications, web-based platforms, desktop applications, and cloud-basedsystems for telemedicine applications.In an embodiment, the pre-processing module integrates retinal layer removal andimage registration techniques to enhance diagnostic accuracy.In an embodiment, the detection module provides automated alerts for criticalconditions requiring urgent clinical intervention.In an embodiment, the classification module uses dropout layers to mitigate overfittingand improve the generalization of the transfer learning model.In an embodiment, the system further comprises a feedback mechanism for iterativeimprovement of the model using real-world clinical data collected post-deployment.In an embodiment, the feature extraction module identifies patterns specific to earlystage retinal diseases, enabling preventive interventions and reducing the risk of visionloss.In an embodiment, the communication module supports deployment on multipleplatforms, including mobile devices, desktop systems, web applications, and cloudbased platforms.The present disclosure provides a method for detecting retinal diseases, comprising thesteps of:- capturing, by an image acquisition module, high-resolution, cross-sectionalimages of the retina and other ocular structures using Optical CoherenceTomography (OCT) imaging techniques;- collecting and storing, by a data collection module, the image data from theinput module in a repository ;- standardizing and normalizing, by a pre-processing module, the collectedimage data using pre-processing techniques including noise reduction, imageenhancement, and segmentation to generate processed image data;- extracting, by a feature extraction module, structural and texture-basedfeatures related to retinal diseases and generating extracted image data;- splitting, by a data splitting module, the extracted image data to generatetraining, validation, and testing datasets;- training, by a training module, a convolutional neural network (CNN)-basedtransfer learning model based on the training dataset;- monitoring and optimizing, by a validation module, model performanceduring training based on the validation dataset;- testing, by a testing module, the model post-training based on the testingdataset;- classifying, by a classification module, the extracted image data intoclassified data of disease categories including Central SerousChorioretinopathy (CSC), Choroidal Neovascularization (CNV), DiabeticMacular Edema (DME), Drusen, or normal condition, by implementing thetrained CNN-based transfer learning model;- analysing, by an analysis module, the classified data to identify pathologicalpatterns and generate a disease-specific detection dataset; and- generating, by a diagnosis and report generation module, diagnostic outputsincluding detailed reports of detected retinal conditions based on analyseddata, so as to provide real-time diagnostic results on a user interface, andfacilitate integration with external healthcare systems for streamlined clinicalworkflows.BRIEF DESCRIPTION OF ACCOMPANYING DRAWINGA system for detecting retinal diseases and a method thereof, of the present disclosurewill now be described with the help of the accompanying drawings, in which:Figure 1 illustrates the block diagram representing a system for detecting retinaldiseases in accordance with one embodiment of the present disclosure;Figures 2A, 2B, and 2C illustrate a method for detecting retinal diseases, in accordancewith the present disclosure; andFigure 3 depicts a graph illustrating the performance of a transfer learning modelemployed by the system for detecting retinal diseases in accordance with anembodiment of the present disclosure.DETAILED DESCRIPTIONEmbodiments, of the present disclosure, will now be described with reference to theaccompanying drawings.Embodiments are provided so as to thoroughly and fully convey the scope of the presentdisclosure to the person skilled in the art. Numerous details, are set forth, relating tospecific components, and methods, to provide a complete understanding ofembodiments of the present disclosure. It will be apparent to the person skilled in the artthat the details provided in the embodiments should not be construed to limit the scopeof the present disclosure. In some embodiments, well-known processes, well-knownapparatus structures, and well-known techniques are not described in detail.The terminology used, in the present disclosure, is only for the purpose of explaining aparticular embodiment and such terminology shall not be considered to limit the scopeof the present disclosure. As used in the present disclosure, the forms "a," "an," and"the" may be intended to include the plural forms as well, unless the context clearlysuggests otherwise. The terms "comprises," "comprising," "including," and "having,"are open ended transitional phrases and therefore specify the presence of stated features,elements, modules, units, and / or components, but do not forbid the presence or additionof one or more other features, elements, components, and / or groups thereof.Eye infections affect people of all ages and are caused by viruses, bacteria, or fungi,impacting areas like the eyelids, cornea, and conjunctiva. Conditions such asConjunctivitis, Dry Eye Syndrome, Cataracts, Glaucoma, Central SerousChorioretinopathy (CSC), Choroidal Neovascularization (CNV), Diabetic MacularEdema (DME), and Drusen can result from these infections. CSC, for instance, involvesfluid buildup between the retina and choroid and may lead to retinal detachment andvision problems. CNV is characterized by abnormal blood vessel growth affectingvision, while DME causes macular swelling due to diabetes. Drusen, deposits behindthe retina, can indicate early macular degeneration. Early detection through regular eyeexams is crucial to prevent vision loss. However, traditional diagnostic methods likemanual Optical Coherence Tomography (OCT) interpretation and fundus photographycan be subjective, labor-intensive, and prone to variability, particularly in high-volumeor resource-limited settings, leading to potential misdiagnoses and disparities inhealthcare outcomes. Therefore, there is a need for a system for detecting retinaldiseases and a method thereof that alleviates the aforementioned drawbacks.In accordance with one embodiment of the present disclosure, there is disclosed asystem for detecting retinal diseases and a method thereof (hereinafter referred to as'system 100').Referring to Figure 1, The present disclosure provides a system (100) for detectingretinal diseases, the system (100) comprising: an image acquisition module (102), a datacollection module (106), a pre-processing module (108), a feature extraction module(110), a data management module (112), a detection module (114), a diagnosis andreport generation module (116), a communication module (118), and a user interfacemodule (120).The image acquisition module (102) is configured to capture high-resolution, crosssectional images of retinal and ocular structures using Optical Coherence Tomography(OCT) imaging techniques.The data collection module (106) is operatively coupled to the image acquisitionmodule (102) and configured to collect the captured image data, and store the imagedata in a repository (106a) for subsequent processing.The pre-processing module (108) is configured to receive the collected image data fromthe repository (106a), standardize and normalize the collected image data using preprocessing techniques, including noise reduction, image enhancement, andsegmentation, and generate processed image data suitable for analysis.The feature extraction module (110) is configured to receive the processed image datafrom the pre-processing module (108) to extract structural and texture-based featuresrelated to retinal diseases, so as to generate extracted image data for subsequentclassification.The data management module (112) comprises:The data splitting module (112A) is configured to receive the extracted image data fromthe feature extraction module (110) to split the extracted image data to generate training,validation, and testing datasets.The training module (112B) is configured to receive the training dataset from the datasplitting module (112A) to train a convolutional neural network (CNN)-based transferlearning model.The validation module (112C) is configured to receive the validation dataset from thedata splitting module (112) to monitor and optimize model performance during thetraining.The testing module (112D) is configured to receive the test dataset from the datasplitting module (112) to evaluate the model post-training.The detection module (114) comprises:The classification module (114A) employs the trained CNN-based transfer learningmodel to classify the extracted image data into classified data of disease categoriesincluding Central Serous Chorioretinopathy (CSC), Choroidal Neovascularization(CNV), Diabetic Macular Edema (DME), Drusen, or normal condition.The analysis module (114b) is configured to receive the classified data from theclassified module (114a) to analyse the classified data to identify pathological patternsand generate a disease-specific detection dataset.The diagnosis and report generation module (116) is configured to receive the diseasespecific detection dataset from the analysis module (114b) to generate diagnosticoutputs including detailed reports of detected retinal conditions, provide real-timediagnostic results on a user interface, and facilitate integration with external healthcaresystems for streamlined clinical workflows.The communication module (118) is operatively connected to the detection module(114) and diagnosis and report generation module (116), configured to transmitdiagnostic data to cloud storage or telemedicine platforms, and receive updated datasetsor algorithms for retraining and model enhancement.The user interface module (120) implements visualization tools for displayingdiagnostic results, including heatmaps overlaid on retinal images, and user-configurableoptions for selecting disease-specific analyses or customizing reports.In an embodiment, the transfer learning model is based on a pre-trained ConvolutionalNeural Network (CNN) architecture, selected from VGG-16, ResNet-50, or similarmodels, and is fine-tuned on retinal OCT datasets.In an embodiment, the pre-processing module (108) employs techniques including:- noise reduction methods including Gaussian filtering and wavelet denoising;- image segmentation and thresholding for retinal layer extraction; and- data augmentation techniques including rotation, scaling, and flipping toincrease the training dataset diversity.In an embodiment, the feature extraction module (110) applies feature extractiontechniques, including histogram of oriented gradients (HOG), wavelet transform,texture analysis, and principal component analysis (PCA).In an embodiment, the detection module (114) integrates dropout layers within the CNNarchitecture to reduce overfitting and improve generalization during classification tasks.In an embodiment, the classification module (114A) achieves accuracy levels of up to96.9% by leveraging transfer learning techniques and extensive training on OCTdatasets.In an embodiment, the diagnosis and report generation module (116) provides:- real-time or near-real-time diagnostic results during patient evaluation;- seamless integration with electronic health record (EHR) systems for streamlinedclinical workflows; and- patient-friendly reports to improve awareness and engagement.In an embodiment, the data splitting module (112A) enables the use of external datasets,such as the Kaggle OCT dataset and OCT2017 dataset, for training and validation of thetransfer learning model.In an embodiment, the system (100) supports deployment across multiple platforms,including mobile applications, web-based platforms, desktop applications, and cloudbased systems for telemedicine applications.In an embodiment, the pre-processing module (108) integrates retinal layer removal andimage registration techniques to enhance diagnostic accuracy.In an embodiment, the detection module (114) provides automated alerts for criticalconditions requiring urgent clinical intervention.In an embodiment, the classification module (114A) uses dropout layers to mitigateoverfitting and improve the generalization of the transfer learning model.The system (100) as claimed in claim 1, further comprises a feedback mechanism foriterative improvement of the model using real-world clinical data collected postdeployment.In an embodiment, the feature extraction module (110) identifies patterns specific toearly-stage retinal diseases, enabling preventive interventions and reducing the risk ofvision loss.In an embodiment, the communication module (118) supports deployment on multipleplatforms, including mobile devices, desktop systems, web applications, and cloudbased platforms.Figures 2A, 2B, and 2C provide a method (200) for detecting retinal diseases,comprising the steps of:- capturing, by an image acquisition module (102), high-resolution, crosssectional images of the retina and other ocular structures using OpticalCoherence Tomography (OCT) imaging techniques;- collecting and storing, by a data collection module (106), the image datafrom the input module (102) in a repository (106a);- standardizing and normalizing, by a pre-processing module (108), thecollected image data using pre-processing techniques including noisereduction, image enhancement, and segmentation to generate processedimage data;- extracting, by a feature extraction module (110), structural and texture-basedfeatures related to retinal diseases and generating extracted image data;- splitting, by a data splitting module (112A), the extracted image data togenerate training, validation, and testing datasets;- training, by a training module (112B), a convolutional neural network(CNN)-based transfer learning model based on the training dataset;- monitoring and optimizing, by a validation module (112C), modelperformance during training based on the validation dataset;- testing, by a testing module (112D), the model post-training based on thetesting dataset;- classifying, by a classification module (114A), the extracted image data intoclassified data of disease categories including Central SerousChorioretinopathy (CSC), Choroidal Neovascularization (CNV), DiabeticMacular Edema (DME), Drusen, or normal condition, by implementing thetrained CNN-based transfer learning model;- analysing, by an analysis module (114B), the classified data to identifypathological patterns and generate a disease-specific detection dataset; and- generating, by a diagnosis and report generation module (116), diagnosticoutputs including detailed reports of detected retinal conditions based onanalysed data, so as to provide real-time diagnostic results on a userinterface, and facilitate integration with external healthcare systems forstreamlined clinical workflows.In a preferred embodiment of the present disclosure, the data splitting module (112)may utilize third-party datasets to train the transfer learning model in order to detectvarious retinal diseases. The Kaggle OCT dataset may be utilized to train the transferlearning model deployed by the detection module (114) in order to detect CSC atvarious stages including acute CSC and chronic CSC, as shown in Table 1. TheOCT2017 dataset may be utilized to train the transfer learning model for detectingretinal diseases including CNV, DME, and Drusen as shown in Table 2.Table 1 depicts a dataset division for training the transfer learning model for CSCdetection in accordance with one embodiment of the disclosure.Table 1Table 2 depicts a dataset division for training the transfer learning model for detectingvarious retinal diseases in accordance with one embodiment of the disclosure.Table 2In one embodiment of the present disclosure, the detection module (114) can be atransfer learning (TL) based model with pre-trained convolutional neural networks(CNNs) to classify retinal OCT images based on various retinal diseases and detect thepresence of retinal diseases. The detection module (114) can receive featured data fromthe feature extraction module (110). Furthermore, the classification module (114a) canemploy the VGG-16 model trained on the ImageNEt dataset and fine-tuned on theOCT2017 dataset for classification. The VGG-16 leverages the architecture comprisinga total of 16 layers, including 13 convolutional layers and 3 fully connected layers. Theconvolutional layers can be arranged in sequential blocks, with each block comprisingone or more 3x3 convolutional layers followed by a ReLU activation function and a 2x2max-pooling layer. Dropout layers can be included in the final two blocks to mitigateoverfitting. The network accepts 224x224x3 RGB images as input and theconvolutional base extracts features that three fully connected layers can process. Thefirst two fully connected layers comprise 4096 neurons, followed by a ReLU activationfunction and a 0.5 dropout layer. The third fully connected layer can have 1000 neuronscorresponding to the ImageNet classes and can use a softmax activation function toproduce a probability distribution over these classes. The model may focus on a subsetof these classes for specific applications and can generate classified data.In another embodiment of the present disclosure, the analysis module (114b) cananalyze the classified data to determine the presence of retinal diseases by comparing itwith the classified data and producing analyzed data.In one embodiment of the present disclosure, the diagnosis and report generationmodule (116) can generate a comprehensive report utilizing the analyzed data receivedfrom the analysis module (114b).In a preferred embodiment of the present disclosure, Figure 3 depicts a graphillustrating the performance of a transfer learning model employed by the system fordetecting retinal diseases (100). The transfer learning model can utilize the trainingmodule (112a), the validation module (112b), and the testing module (112c) todetermine the model performance. The model can be initiated with training up to 2epochs with a validation batch size of 4, utilizing the OCT2017 dataset comprising 4032training images, 32 validation images, and 500 test images, which can achieve anaccuracy of 91.5% and a loss of 3.5%. Subsequently, the training up to 10 epochs,maintaining the validation batch size at 4, with the same 4032 training images and 32validation images, but expanding the test set to 1000 images, can result in an improvedaccuracy of 95.5% and a reduced loss of 1.46%. Further, the increase of training up to20 epochs, with the training set expanded to 8016 images, while keeping the validationset at 32 images and the test set at 1000 images, an accuracy of 95.8% can be achievedwith a loss of 1.38%. Finally, when extensive training up to 200 epochs, again with avalidation batch size of 4, using the same 8016 training images, 32 validation images,and 1000 test images, the model performance can lead to a peak accuracy of 96.9% anda minimal loss of 1.2%.In one embodiment of the present disclosure, the system (100) can be deployed ondiverse platforms including mobile platforms, web-based applications, desktopapplications, and cloud platforms for detecting retinal diseases.In an operative configuration, the retinal disease detection system (100) captures highresolution OCT images through the image acquisition module (102), with data stored inthe data collection module (106) for subsequent processing. The pre-processing module(108) ensures image quality via noise reduction and enhancement, while the featureextraction module (110) identifies structural and texture features indicative of retinaldiseases. The data management module (112) trains, validates, and tests a CNN-basedmodel to classify retinal conditions. The detection module (114) analyzes and classifiesthese conditions, providing disease-specific datasets. Diagnostic outputs are generatedby the diagnosis and report generation module (116) and presented via the user interfacemodule (120). The communication module (118) enables cloud integration,telemedicine support, and system updates.Advantageously, the retinal disease detection system offers significant benefits inenhancing diagnostic accuracy and efficiency. By utilizing high-resolution OCTimaging and advanced CNN-based transfer learning, the present subject matter ensuresprecise detection of retinal diseases such as CSC, CNV, DME, and Drusen. The systemoperates in real-time, providing instant diagnostic results and detailed reports,streamlining clinical workflows. Its integration capabilities with telemedicine platformsand external healthcare systems facilitate seamless data sharing and system scalability.Additionally, the user interface allows customizable visualization, including heatmapsand tailored analysis, ensuring user-friendly interaction. The system's modular designalso supports continual updates and retraining for model enhancement, making it arobust and future-ready diagnostic solution.The foregoing description of the embodiments has been provided for purposes ofillustration and is not intended to limit the scope of the present disclosure. Individualcomponents of a particular embodiment are generally not limited to that particularembodiment but are interchangeable. Such variations are not to be regarded as adeparture from the present disclosure, and all such modifications are considered to bewithin the scope of the present disclosure.TECHNICAL ADVANCES AND ECONOMIC SIGNIFICANCEThe present disclosure described herein above has several technical advantagesincluding, but not limited to, the realization of a system for detecting retinal diseasesand a method thereof, that:- can provide a system for detecting retinal diseases;- can enhance retinal disease detection accuracy by utilizing advanced models trainedusing extensive datasets;- can reduce the time required for image interpretation by automating OCT imageanalysis and enable quicker identification of retinal diseases;- can enhance the capability for real-time or near-real-time processing of OCTimages to provide immediate feedback during patient evaluations, enabling in-timeclinical intervention;- can facilitate seamless integration with existing healthcare infrastructure withoutrequiring significant hardware and software modification, ensuring scalability;- can improve average accuracy of retinal disease diagnosis by up to 96.9 percent;- can deliver an efficient system by minimizing reliance on manual diagnosis,improving efficiency, and ensuring objective and comprehensive analysis of retinalimages;- can support remote diagnosis and telemedicine applications, expanding access tohigh-quality retinal care for patients in remote or underserved areas; and- can allow patients to view and understand their diagnostic results, enhancingengagement and awareness, and thereby contributing to improved outcomes andpatient satisfaction.The foregoing description of the specific embodiments so fully reveals the generalnature of the embodiments herein that others can, by applying current knowledge,readily modify and / or adapt for various applications such specific embodimentswithout departing from the generic concept, and, therefore, such adaptations andmodifications should and are intended to be comprehended within the meaning andrange of equivalents of the disclosed embodiments. It is to be understood that thephraseology or terminology employed herein is for the purpose of description and notof limitation. Therefore, while the embodiments herein have been described in terms ofpreferred embodiments, those skilled in the art will recognize that the embodimentsherein can be practiced with modification within the spirit and scope of theembodiments as described herein.Throughout this specification the word "comprise", or variations such as "comprises" or"comprising", will be understood to imply the inclusion of a stated element, or group ofelements, but not the exclusion of any other element, or group of elements.While considerable emphasis has been placed herein on the components and componentparts of the preferred embodiments, it will be appreciated that many embodiments canbe made and that many changes can be made in the preferred embodiments withoutdeparting from the principles of the disclosure. These and other changes in the preferredembodiment as well as other embodiments of the disclosure will be apparent to thoseskilled in the art from the disclosure herein, whereby it is to be distinctly understoodthat the foregoing descriptive matter is to be interpreted merely as illustrative of thedisclosure and not as a limitation.LIST OF REFERENCE NUMERALS USED IN DETAILED DESCRIPTIONAND DRAWING100 - Retinal Disease Detection System102 - Image Acquisition Module106 - Data Collection Module106a - Repository for Image Data108 - Pre-Processing Module110 - Feature Extraction Module112 - Data Management Module112A - Data Splitting Module112B - Training Module112C - Validation Module112D - Testing Module114 - Detection Module114A - Classification Module114B - Analysis Module116 - Diagnosis and Report Generation Module118 - Communication Module120 - User Interface Module200 - Method

Claims

1. A system (100) for detecting retinal diseases, said system (100) comprising: - an image acquisition module (102) configured to capture highresolution, cross-sectional images of retinal and ocular structures using Optical Coherence Tomography (OCT) imaging techniques; - a data collection module (106) operatively coupled to the image acquisition module (102) and configured to collect the captured image data, and store said image data in a repository (106a) for subsequent processing; - a pre-processing module (108) configured to receive said collected image data from said repository (106a), standardize and normalize said collected image data using pre-processing techniques, including noise reduction, image enhancement, and segmentation, and generate processed image data suitable for analysis; - a feature extraction module (110) configured to receive said processed image data from said pre-processing module (108) to extract structural and texture-based features related to retinal diseases, so as to generate extracted image data for subsequent classification; - a data management module (112) comprising: - a data splitting module (112A) configured to receive said extracted image data from said feature extraction module (110) to split said extracted image data to generate training, validation, and testing datasets; - a training module (112B) configured to receive said training dataset from said data splitting module (112A) to train a convolutional neural network (CNN)-based transfer learning model; - a validation module (112C) configured to receive said validation dataset from said data splitting module (112) to monitor and optimize model performance during the training; and- a testing module (112D) configured to receive said test dataset from said data splitting module (112) to evaluate the model post-training; - a detection module (114) comprising: - a classification module (114A) employing the trained CNNbased transfer learning model to classify said extracted image data into classified data of disease categories including Central Serous Chorioretinopathy (CSC), Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), Drusen, or normal condition; and - an analysis module (114b) configured to receive said classified data from said classified module (114a) to analyse said classified data to identify pathological patterns and generate a disease-specific detection dataset; - a diagnosis and report generation module (116) configured to receive said disease-specific detection dataset from analysis module (114b) to generate diagnostic outputs including detailed reports of detected retinal conditions, provide real-time diagnostic results on a user interface, and facilitate integration with external healthcare systems for streamlined clinical workflows; - a communication module (118) operatively connected to the detection module (114) and diagnosis and report generation module (116), configured to transmit diagnostic data to cloud storage or telemedicine platforms, and receive updated datasets or algorithms for retraining and model enhancement; and - a user interface module (120) implementing visualization tools for displaying diagnostic results, including heatmaps overlaid on retinal images, and user-configurable options for selecting disease-specific analyses or customizing reports.

2. The system (100) as claimed in claim 1, wherein the transfer learning model is based on a pre-trained Convolutional Neural Network (CNN) architecture,selected from VGG-16, ResNet-50, or similar models, and is fine-tuned on retinal OCT datasets.

3. The system (100) as claimed in claim 1, wherein the pre-processing module (108) employs techniques including: - noise reduction methods including Gaussian filtering and wavelet denoising; - image segmentation and thresholding for retinal layer extraction; and - data augmentation techniques including rotation, scaling, and flipping to increase the training dataset diversity.

4. The system (100) as claimed in claim 1, wherein the feature extraction module (110) applies feature extraction techniques, including histogram of oriented gradients (HOG), wavelet transform, texture analysis, and principal component analysis (PCA).

5. The system (100) as claimed in claim 1, wherein the detection module (114) integrates dropout layers within the CNN architecture to reduce overfitting and improve generalization during classification tasks.

6. The system (100) as claimed in claim 1, wherein the classification module (114A) achieves accuracy levels up to 96.9% by leveraging transfer learning techniques and extensive training on OCT datasets.

7. The system (100) as claimed in claim 1, wherein the diagnosis and report generation module (116) provides: - real-time or near-real-time diagnostic results during patient evaluation; - seamless integration with electronic health record (EHR) systems for streamlined clinical workflows; and - patient-friendly reports to improve awareness and engagement.

8. The system (100) as claimed in claim 1, wherein the data splitting module (112A) enables the use of external datasets, such as the Kaggle OCT dataset and OCT2017 dataset, for training and validation of the transfer learning model.

9. The system (100) as claimed in claim 1, wherein the system (100) supports deployment across multiple platforms, including mobile applications, webbased platforms, desktop applications, and cloud-based systems for telemedicine applications.

10. The system (100) as claimed in claim 1, wherein the pre-processing module (108) integrates retinal layer removal and image registration techniques to enhance diagnostic accuracy.

11. The system (100) as claimed in claim 1, wherein the detection module (114) provides automated alerts for critical conditions requiring urgent clinical intervention.

12. The system (100) as claimed in claim 1, wherein the classification module (114A) uses dropout layers to mitigate overfitting and improve the generalization of the transfer learning model.

13. The system (100) as claimed in claim 1, further comprises a feedback mechanism for iterative improvement of the model using real-world clinical data collected post-deployment.

14. The system (100) as claimed in claim 1, wherein the feature extraction module (110) identifies patterns specific to early-stage retinal diseases, enabling preventive interventions and reducing the risk of vision loss.

15. The system (100) as claimed in claim 1, wherein the communication module (118) supports deployment on multiple platforms, including mobile devices, desktop systems, web applications, and cloud-based platforms.

16. A method (200) for detecting retinal diseases, comprising the steps of: - capturing, by an image acquisition module (102), high-resolution, cross-sectional images of the retina and other ocular structures using Optical Coherence Tomography (OCT) imaging techniques; - collecting and storing, by a data collection module (106), said image data from said input module (102) in a repository (106a); - standardizing and normalizing, by a pre-processing module (108), said collected image data using pre-processing techniques including noise reduction, image enhancement, and segmentation to generate processed image data; - extracting, by a feature extraction module (110), structural and texturebased features related to retinal diseases and generating extracted image data; - splitting, by a data splitting module (112A), said extracted image data to generate training, validation, and testing datasets; - training, by a training module (112B), a convolutional neural network (CNN)-based transfer learning model based on said training dataset; - monitoring and optimizing, by a validation module (112C), model performance during training based on said validation dataset; - testing, by a testing module (112D), the model post-training based on said testing dataset; - classifying, by a classification module (114A), said extracted image data into classified data of disease categories including Central Serous Chorioretinopathy (CSC), Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), Drusen, or normal condition, by implementing the trained CNN-based transfer learning model; - analysing, by an analysis module (114B), said classified data to identify pathological patterns and generate a disease-specific detection dataset; and - generating, by a diagnosis and report generation module (116), diagnostic outputs including detailed reports of detected retinal conditions based on analysed data, so as to provide real-time diagnostic results on a user interface, and facilitate integration with external healthcare systems for streamlined clinical workflows.