Claims:STATEMENT OF CLAIMS
We claim:
1. A method (100) for predicting retinal diseases using machine learning, the method comprising:
obtaining (101), by a neural network (303), a low resolution medical image of retina of a subject; and
predicting (105), by the neural network (303), presence of at least one retinal disease/condition from the low resolution medical image using deep learning, wherein the neural network (303) utilizes at least one classifier to identify the presence of the at least one retinal disease/condition.
2. The method, as claimed in claim 1, wherein the prediction of the presence of the at least one retinal disease/condition is based on computed at least one value of probability of presence of a retinal disease/condition, wherein the probability of presence of a retinal disease/condition quantifies a likelihood of occurrence of the at least one retinal disease/condition in the retina of the subject.
3. The method, as claimed in claim 1, wherein the neural network (303) is trained to predict the at least one retinal disease or condition in the low resolution medical image based on a few-shot training methodology.
4. The method, as claimed in claim 1, wherein the method further comprises generating a heat map superimposed on the low resolution medical image for localizing the predicted at least one retinal disease or condition in the low resolution medical image, wherein a region encompassed by the heat map indicates a presence of high energy.
5. The method, as claimed in claim 1, wherein the method further comprises displaying at least one of the low resolution medical image of the retina, the predicted at least one retinal disease or condition, the at least one value of probability of presence of the predicted at least one retinal disease/condition, and the heat map on a display (305).
6. The method, as claimed in claim 1, wherein the at least one retinal disease/condition comprises Diabetic Retinopathy, Age-Related Macular Degeneration, Branch Retinal Vein Occlusion, Media Haze, Drusen, Tessellation, Myopia, Epiretinal Membrane, Laser Scar, Macular Scar, Central Serous Retinopathy, Optic Disc Cupping, Central Retinal Vein Occlusion, Tortuous Vessels, Asteroid Hyalosis, Optic Disc Pallor, Optic Disc Edema, Shunt, Anterior Ischemic Optic Neuropathy, Retinitis, Chorioretinitis, Exudation, Retinal Pigment Epithelium Changes, Macular Hole, Retinitis Pigmentosa, Parafoveal Telangiectasia, and Retinal Traction.
7. A system (300) for predicting retinal diseases using machine learning, the system (300) comprising:
an input receiving unit (301), wherein the input receiving unit (301) is configured to receive an input selection of at least one retinal disease/condition to be diagnosed;
an imaging device (302), wherein the imaging device (302) is configured to
capture a high resolution medical image of retina of a subject; and
generate a low resolution medical image of the retina of the subject by down-sampling the high resolution medical image;
and
a neural network (303), wherein the neural network (303) is configured to
obtain the low resolution medical image of the retina of the subject from the imaging device (302); and
predict presence of the at least one retinal disease/condition from the low resolution medical image using deep learning, wherein the neural network (303) utilizes at least one classifier to identify the presence of the at least one retinal disease/condition.
8. The system (300), as claimed in claim 7, wherein the prediction of the presence of the at least one retinal disease/condition is based on computed at least one value of probability of presence of a retinal disease/condition, wherein the probability of presence of a retinal disease/condition quantifies a likelihood of occurrence of the at least one retinal disease/condition in the retina of the subject.
9. The system (300), as claimed in claim 7, wherein the neural network (303) is trained to predict the at least one retinal disease or condition in the low resolution medical image based on a few-shot training methodology.
10. The system (300), as claimed in claim 7, wherein the system (300) is further configured to generate a heat map superimposed on the low resolution medical image for localizing the predicted at least one retinal disease or condition in the low resolution medical image, wherein a region encompassed by the heat map indicates a presence of high energy.
11. The system (300), as claimed in claim 7, wherein the system (300) is further configured to display at least one of the low resolution medical image of the retina, the predicted at least one retinal disease or condition, the at least one value of probability of presence of the predicted at least one retinal disease/condition, and the heat map on a display (305).

12. The system (300), as claimed in claim 7, wherein the at least one retinal disease/condition comprises Diabetic Retinopathy, Age-Related Macular Degeneration, Branch Retinal Vein Occlusion, Media Haze, Drusen, Tessellation, Myopia, Epiretinal Membrane, Laser Scar, Macular Scar, Central Serous Retinopathy, Optic Disc Cupping, Central Retinal Vein Occlusion, Tortuous Vessels, Asteroid Hyalosis, Optic Disc Pallor, Optic Disc Edema, Shunt, Anterior Ischemic Optic Neuropathy, Retinitis, Chorioretinitis, Exudation, Retinal Pigment Epithelium Changes, Macular Hole, Retinitis Pigmentosa, Parafoveal Telangiectasia, and Retinal Traction. , Description:TECHNICAL FIELD
[001] Embodiments herein relate to diagnosis of diseases related to retina, and more particularly to methods and systems for diagnosing at least one retinal disease through prediction using machine learning based classifiers.
BACKGROUND
[002] Currently, medical imaging is being used for diagnosing or detecting numerous types of ophthalmic diseases and disorders. For example, color fundus camera/photography can be utilized for extracting images, which can be used for diagnosing the ophthalmic diseases and disorders. The diagnostic processes involving medical imaging generally involve reliance upon human experts for individual analysis of images. With advancement in field of image processing or other related fields, the number of medical imaging procedures has increased significantly. This has contributed to an increase in demand for efficient and accurate medical image analysis, which in turn necessitates immense availability of expert manpower that is capable of performing such efficient and accurate medical image analysis. Therefore, the reliance upon human experts for individual analysis of images has increased.
[003] In order to minimize increasing involvement of human experts in diagnosis of the ophthalmic diseases, machine learning was introduced. However, a machine learning tool or method is likely to function as a black box. Therefore, acceptance of such diagnoses generated through the machine learning tool and/or method may be hindered. This is particularly due to lack of transparency on specific or objective criteria used by classifiers in the machine learning tool and/or method for evaluating a medical image, which had led to the generation of a prediction (diagnosis) of existence of a disease.
[004] In case, the number of medical images used for training machine learning classifiers is not sufficient (deficient training data due to non-availability of adequate medical images in a training set), the accuracy of prediction of diseases may be impacted. This issue (lack of confidence in the prediction) may be exacerbated if the machine learning classifiers have been trained to predict diseases or conditions that are relatively rare. Further, the machine learning classifiers are likely to include a single endpoint and, therefore, may not be able to predict multiple diseases or conditions from the medical images. The early detection and treatment of the ophthalmic diseases can prevent progression of the ophthalmic diseases to the advanced stages of illness.
OBJECTS
[005] The principal object of the embodiments herein is to predict one or more ophthalmic (retinal) disease/disorder/condition from retinal medical images using machine learning.
[006] Another object of the embodiments herein is to predict the at least one ophthalmic disease by computing a likelihood of at least one condition present in the retina of a subject using a retinal medical image obtained from the subject using one or more medical imaging equipment.
[007] Another object of the embodiments herein is to provide an interpretive result relating to at least one diagnosed (predicted) retinal disease/condition.
[008] Another object of the embodiments herein is to enable a machine learning system to accurately predict the at least one retinal disease, even if the machine learning system is trained using limited training data (medical images).
[009] Another object of the embodiments herein is to enable machine learning classifiers to diagnose a plurality of ophthalmic diseases, disorders, or conditions in one or more subjects.

BRIEF DESCRIPTION OF FIGURES
[0010] Embodiments herein are illustrated in the accompanying drawings, throughout which like reference letters indicate corresponding parts in the various figures. The embodiments herein will be better understood from the following description with reference to the drawings, in which:
[0011] FIG. 1 is a flow diagram depicting a method for predicting one or more retinal diseases/conditions using machine learning and providing an interpretive result relating to diagnosed one or more retinal diseases/conditions, according to embodiments as disclosed herein;
[0012] FIG. 2 is an example illustration of prediction of three retinal diseases using machine learning, according to embodiments as disclosed herein;
[0013] FIG. 3 depicts various units of an example system configured to predict one or more retinal diseases or conditions using machine learning and provide an interpretive result relating to diagnosed one or more retinal diseases or conditions, according to embodiments as disclosed herein;
[0014] FIG. 4 depicts a few-shot training methodology used for training a Convolutional Neural Network (CNN) to simultaneously predict the presence of one or more retinal diseases, according to embodiments as disclosed herein;
[0015] FIG. 5 is an example depiction of prediction of plurality of retinal diseases based on likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein;
[0016] FIG. 6 is another example depiction of prediction of plurality of retinal diseases based on the likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein;
[0017] FIG. 7 is yet another example depiction of prediction of plurality of retinal diseases based on the likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein;
[0018] FIG. 8 is an example depicting the generation of maps that allow human experts to interpret the results of prediction of one or more retinal diseases using machine learning, according to embodiments as disclosed herein; and
[0019] FIG. 9 is another example depicting the generation of maps that allow human experts to interpret the results of prediction of one or more retinal diseases using machine learning, according to embodiments as disclosed herein.

DETAILED DESCRIPTION
[0020] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
[0021] Embodiments herein disclose methods and systems for predicting at least one retinal disease using machine learning. In an embodiment, medical images can be obtained using a medical imaging system. In an example, the medical images can be color images obtained using a fundus camera. The medical images can be analyzed using machine learning for detecting, classifying, and displaying candidate diseases. The embodiments include predicting the likelihood of the at least one disease present in the retina based on the analysis. The embodiments include displaying the position and location of the retinal conditions. In an embodiment, heat maps can be generated, which can localize at least one region of interest (where retinal disease or condition has been detected or diagnosed).
[0022] In an embodiment, retinal diseases that can be diagnosed include, but not limited to, Diabetic Retinopathy (DR), Age-Related Macular Degeneration (ARMD), Media Haze (MH), Drusen (DN), Tessellation (TSLN), Branch Retinal Vein Occlusion (BRVO), Myopia (MYA), Epiretinal Membrane (ERM), Laser Scar (LS), Macular Scar (MS), Central Serous Retinopathy (CSR), Optic Disc Cupping (ODC), Central Retinal Vein Occlusion (CRVO), Tortuous Vessels (TV), Asteroid Hyalosis (AH), Optic Disc Pallor (ODP), Optic Disc Edema (ODE), Shunt (ST), Anterior Ischemic Optic Neuropathy (AION), Retinitis (RS), Chorioretinitis (CRS), Exudation (EDN), Retinal Pigment Epithelium Changes (RPEC), Macular Hole (MHL), Retinitis Pigmentosa (RP), Parafoveal Telangiectasia (PT), Retinal Traction (RT), and so on.
[0023] Referring now to the drawings and more particularly to methods and systems for FIGS. 1 through 9, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments.
[0024] FIG. 1 is a flow diagram 100 depicting a method for predicting one or more retinal diseases or conditions using machine learning and providing an interpretive result relating to diagnosed one or more retinal diseases/conditions, according to embodiments as disclosed herein. The embodiments include predicting one or more retinal diseases or conditions using deep learning (DL). The embodiments include training a Convolutional Neural Network (CNN) model for predicting the one or more retinal diseases or conditions. Once the training is completed, the embodiments include testing a deep learning model (CNN), which involves predicting one or more retinal diseases or conditions.
[0025] As depicted in FIG. 1, the flow diagram comprises two parts, viz., a training part and a testing part. At step 101, the method includes obtaining high-resolution medical images using a medical imaging equipment. In an embodiment, a fundus camera can be the medical imaging equipment used for obtaining the high-resolution medical images. The high-resolution medical images are acquired from three different retinal fundus cameras. At step 102, the method includes cropping the high-resolution medical images and down-sampling the high-resolution medical images for obtaining low-resolution medical images. The low-resolution medical images can be part of a training dataset. In an embodiment, the training dataset includes 2560 low-resolution medical images, which can be used for training, validating, and testing.
[0026] Once the training is completed, the CNN model can be used in the testing phase. In an embodiment, the trained CNN model can be tested using 640 retinal medical images. The embodiments include predicting one or more retinal diseases and conditions in each of the 640 retinal medical images. It is determined that the average Area Under the Curve (AUC) after cross validation of the model is 0.8754.
[0027] At step 103, the method includes obtaining a high-resolution medical images of a retina of a subject (patient), using medical imaging equipment, which needs to be analyzed for detecting one or more retinal diseases or retinal conditions. In an embodiment, the medical imaging equipment used for obtaining the high-resolution medical images of the retina is a fundus camera. At step 104, the method includes cropping the high-resolution medical image of the retina of the subject and down-sampling the high-resolution medical image. The cropping and down-sampling allows obtaining a low-resolution retinal medical image.
[0028] At step 105, the method includes performing deep learning based classification on the low-resolution retinal medical image to predict the presence of one or more retinal diseases and/or conditions. In an embodiment, the CNN model can provide values of probabilities of presence of the one or more retinal diseases and/or conditions. In an embodiment, a user can configure the CNN model to predict a predefined number of retinal diseases and/or conditions. In an example, consider that the user had configured the CNN model to predict 28 retinal diseases and/or conditions. In this instance, the embodiments include computing the probability of presence of retinal disease and/or condition corresponding to each of the 28 retinal diseases and/or conditions.
[0029] At step 106, the method includes generating a heat map for indicating location, in the retina, in which one or more retinal diseases and/or conditions have been predicted. The heat maps can indicate the presence of energy in the low-resolution retinal medical image, which can be interpreted for diagnosis of presence of one or more retinal diseases and/or conditions. Thereafter, experts can access the medical image, along with the prediction of the one or more retinal diseases and/or conditions and the heat maps, for determining the best course of treatment for enabling the retina to recover from the one or more retinal diseases and/or conditions diagnosed in the retina.
[0030] The various actions in the flowchart 100 may be performed in the order presented, in a different order, or simultaneously. Further, in some embodiments, some actions listed in FIG. 1 may be omitted.
[0031] FIG. 2 is an example illustration of prediction of three retinal diseases using machine learning, according to embodiments as disclosed herein. As depicted in FIG. 2, a fundus camera is used for obtaining a high resolution medical image of the retina of a subject. The high resolution medical image can be subsequently cropped and down-sampled to obtain a low resolution image. The embodiments utilize a deep learning to diagnose one or more diseases from the low resolution image of the retina. In an embodiment, the diagnosis involves predicting the presence of one or more diseases in the retina based on probabilities of presence of the one or more diseases in the retina using machine learning classifiers.
[0032] Consider that three retinal diseases have been diagnosed based on the probabilities of presence of the three diseases. The three diagnosed diseases are DR, ODC, and LS. The probabilities of presence of each the three diseases indicate the likelihood of the occurrence of each of the diseases in the retina. The likelihood of the occurrence of the disease DR is 0.634. The likelihood of the occurrence of the disease ODC is 0.728. The likelihood of the occurrence of the disease LS is 0.317. Once the likelihood of the occurrence of the three diseases is determined, the embodiments include generating a heat map in the low resolution medical image of the retina for localizing the diagnosed (predicted) diseases. The heat map depicts the position or location of the retina in which the diseases are present (predicted).
[0033] FIG. 3 depicts various units of an example system 300 configured to predict one or more retinal diseases or conditions using machine learning and provide an interpretive result relating to diagnosed one or more retinal diseases or conditions, according to embodiments as disclosed herein. As depicted in FIG. 3, the example system 300 comprises an input receiving unit 301, an imaging device 302, a neural network 303, a memory 304, and a display 305. The input receiving unit 301 allows a user to configure the system 300 to select retinal diseases or conditions that the user intends to diagnose. In an embodiment, the user can select all retinal diseases or conditions, which allows the user to diagnose all retinal diseases or conditions that the system 300 is capable of diagnosing. The user can refer to a human expert specialized in analyzing retinal medical images and is acquainted with the possible retinal diseases or conditions that can occur in the retina.
[0034] The imaging device 302 can be considered as medical imaging equipment, which can be used for obtaining high-resolution medical images of retina of a subject. In an embodiment, the imaging device 302 comprises of one or more fundus cameras. The one or more fundus cameras can be utilized for obtaining high-resolution medical images of the retina of the subject. The imaging device 302 can store the high-resolution medical images, obtained from multiple subjects, in the memory 304. The high-resolution medical images can be cropped and down-sampled for obtaining low-resolution medical images. The imaging device 302 can store the low-resolution medical images in the memory 304.
[0035] The neural network 303 is configured to predict one or more retinal diseases or conditions using deep learning. In an embodiment, the neural network 303 can be a CNN. The neural network 303 can be trained using a few-shot training methodology for predicting the one or more retinal diseases or conditions. This enables the neural network 303 to accurately predict rare retinal diseases during the testing phase, even if the neural network 303 has been trained using a deficient training dataset. The training dataset comprises of low-resolution retinal medical images.
[0036] Once the training is completed, the neural network 303 the testing phase can be initiated. The neural network 303 can be tested using medical retinal images of multiple subjects, wherein the accuracy of the trained neural network 303 in predicting one or more retinal diseases or conditions is determined. In an embodiment, the neural network 303 can be tested using a training set comprising of retinal medical images. The embodiments include predicting one or more retinal diseases and conditions in each of the retinal medical images. The neural network 303 can be cross-validated during the testing phase based on average AUC values.
[0037] The neural network 303 can obtain a low-resolution retinal medical image from the memory 304, which was stored by the imaging device 302. The neural network 303 can predict the presence of one or more retinal diseases and/or conditions from the low-resolution retinal medical images. In an embodiment, the neural network 303 can compute probabilities of presence of the one or more retinal diseases and/or conditions from low-resolution retinal medical image. The neural network 303 can predict the predefined number of retinal diseases and/or conditions configured by the user or predict all the retinal diseases and/or conditions that the system 300 is capable of diagnosing.
[0038] In an embodiment, the neural network 303 can be trained using a sigmoid layer for predicting at least twenty-eight retinal disease classes. Therefore, the neural network 303 can compute the probabilities produced by the sigmoid layer, which are independent from each other (each representing a retinal disease), and allow for more than one correct retinal disease. Each output provided by the neural network 303 is an estimated classification probability of the low-resolution medical image belonging to a particular disease. The estimated classification probability can be, thereafter, considered as an individual disease with a score ranging from 0.0 to 1.0.
[0039] The neural network 303 can generate heat maps for indicating locations, in the low-resolution retinal medical image, in which one or more retinal diseases and/or conditions have been predicted. The heat maps indicate the regions in the low-resolution retinal medical image, wherein there is a significant presence of energy. In an embodiment, the heat maps can be utilized for interpreting the probabilities of presence of the one or more retinal diseases and/or conditions as diagnosis of presence of one or more retinal diseases and/or conditions based on the values of the probabilities. The prediction and the heat maps can be utilized for determining the treatment for recovering from the one or more retinal diseases and/or conditions diagnosed in the retina.
[0040] The neural network 303 can display the low resolution retinal medical image, the predicted one or more retinal diseases and/or conditions, and the values of the probabilities of presence of the one or more retinal diseases and/or conditions, on the display 305. The neural network 303 can display the generated heat map on the display 305. The user can view the display 305 to access the low resolution retinal medical image, the values of the probabilities of presence of the one or more retinal diseases and/or conditions, and the generated heat map.
[0041] FIG. 3 shows exemplary units of the system 300, but it is to be understood that other embodiments are not limited thereon. In other embodiments, the system 300 may include less or more number of units. Further, the labels or names of the units of the system 300 are used only for illustrative purpose and does not limit the scope of the invention. One or more units can be combined together to perform same or substantially similar function in the system 300.
[0042] FIG. 4 depicts a few-shot training methodology used for training a CNN to simultaneously predict the presence of one or more retinal diseases, according to embodiments as disclosed herein. The few-shot training methodology may be effective even if the training dataset (comprising of training medical images) is deficient. The few-shot training methodology enables the CNN to effectively predict the presence of one or more retinal diseases with a high degree of accuracy. The CNN can be referred to as a few-shot model.
[0043] The few-shot model is trained to recognize a class (disease class) by providing a predefined number class samples. Specifically, the few-shot model is trained on pseudo few-shot task using well labeled ‘base’ classes, for which an abundance of data is available. Specifically, for each training iteration, the few-shot model can be provided with a set of few support images containing target classes (target disease classes), based on which the few-shot model is trained to identify these classes in a set of query images. This allows the few-shot model to learn the few-shot task and recognize unknown classes (rare disease classes) based on a few support samples.
[0044] As shown in the figure FIG. 4, the few-shot model can extract a prototype vector for each target class from a few support images. Similarly, feature maps can be produced for query image(s), and a feature vector corresponding to each pixel location on the feature maps can be compared with the class prototypes obtained from the support set images using a distance metric.
[0045] The few-shot training methodology can be addressed as a multi-label problem wherein the model is few-shot capable of identifying multiple base classes along with few-shot classes.
[0046] FIG. 5 is an example depiction of prediction of plurality of retinal diseases based on the likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein. In an embodiment, the likelihood of occurrence of a retinal disease can be determined based on the probability of presence of the particular retinal disease. In an embodiment, the CNN model can be configured to predict a predefined number of retinal diseases. Consider that the CNN model is configured to determine the likelihoods of occurrence of 27 specific retinal diseases and another non-specific retinal disease. The CNN model can predict the presence of the one or more retinal diseases from a low resolution medical image of the retina of a subject.
[0047] As depicted in FIG. 5, the embodiments include diagnosing three retinal diseases, viz., DR, ARMD, and TSLN. The embodiments predict the occurrence of these three retinal diseases based on the probabilities of presence of each of the three retinal diseases, viz., DR, ARMD, and TSLN. The values of the probabilities of presence of each of the three retinal diseases are highest amongst the other 25 values. The probability of presence of DR is 0.6043. The probability of presence of ARMD is 0.2615. The probability of presence of TSLN is 0.2154.
[0048] FIG. 6 is another example depiction of prediction of plurality of retinal diseases based on the likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein. As depicted in FIG. 6, the embodiments include diagnosing four retinal diseases, viz., MH, TSLN, ODC, and ODP. The embodiments predict the occurrence of these four retinal diseases based on the probabilities of presence of each of the four retinal diseases, viz., MH, TSLN, ODC, and ODP. The values of the probabilities of presence of each of the four retinal diseases are highest amongst the other 24 values. The probability of presence of MH is 0.6295. The probability of presence of TSLN is 0.6242. The probability of presence of ODC is 0.4165. The probability of presence of ODP is 0.4018.
[0049] FIG. 7 is yet another example depiction of prediction of plurality of retinal diseases based on the likelihood of occurrence of the retinal diseases, according to embodiments as disclosed herein. As depicted in FIG. 7, the embodiments include diagnosing two retinal diseases, viz., ARMD and MYA. The embodiments predict the occurrence of these two retinal diseases based on the probabilities of presence of each of the two retinal diseases, viz., ARMD and MYA. The values of the probabilities of presence of each of the two retinal diseases are highest amongst the other 26 values. The probability of presence of ARMD is 0.9730. The probability of presence of MYA is 0.7229.
[0050] FIG. 8 is an example depicting the generation of maps that allow human experts to interpret the results of prediction of one or more retinal diseases using machine learning, according to embodiments as disclosed herein. The embodiments include obtaining a low resolution medical image of the retina of a subject. The embodiments include predicting one or more retinal diseases using the low resolution medical image. The embodiments include generating heat maps, which can depict the location in the retina, in which the one or more retinal diseases that are likely (predicted) to have occurred. The heat maps the high energy regions of the low resolution retinal images of the subject. The heat maps allows the human experts to interpret the results of prediction of the one or more retinal diseases, i.e., the values of the probabilities of presence (or likelihood of occurrence) of the one or more retinal diseases. The heat maps allow building confidence in the results of prediction of the one or more retinal diseases. As depicted in FIG. 8, the embodiments include diagnosing two retinal diseases from the low resolution medical image of the retina of a subject. The two diagnosed retinal diseases are TSLN and ODC. The heat map shows the location of the two predicted retinal diseases.
[0051] FIG. 9 is another example depicting the generation of maps that allow human experts to interpret the results of prediction of one or more retinal diseases using machine learning, according to embodiments as disclosed herein. As depicted in FIG. 9, the embodiments include diagnosing three retinal diseases from a low resolution medical image of the retina of a subject. The diagnosed retinal diseases are DR, ODC and LS. The heat map shows the location of the three predicted retinal diseases.
[0052] The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the network elements. The elements shown in FIG. 3 include blocks which can be at least one of a hardware device, or a combination of hardware device and software module.
[0053] The embodiments disclosed herein describe methods and systems for predicting one or more retinal diseases or conditions using machine learning and providing an interpretive result relating to diagnosed one or more retinal diseases/conditions. Therefore, it is understood that the scope of the protection is extended to such a program and in addition to a computer readable means having a message therein, such computer readable storage means contain program code means for implementation of one or more steps of the method, when the program runs on a server or mobile device or any suitable programmable device. The method is implemented in a preferred embodiment through or together with a software program written in example Very high speed integrated circuit Hardware Description Language (VHDL), or any other programming language, or implemented by one or more VHDL or several software modules being executed on at least one hardware device. The hardware device can be any kind of portable device that can be programmed. The device may also include means, which could be, for example, a hardware means, for example, an Application-specific Integrated Circuit (ASIC), or a combination of hardware and software means, for example, an ASIC and a Field Programmable Gate Array (FPGA), or at least one microprocessor and at least one memory with software modules located therein. The method embodiments described herein could be implemented partly in hardware and partly in software. Alternatively, the invention may be implemented on different hardware devices, e.g. using a plurality of Central Processing Units (CPUs).
[0054] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the scope of the embodiments as described herein.