Server for predicting disease through ai-based iris analysis and method for operating same

Abstract

A server for predicting a disease through AI-based iris analysis and a method for operating same are disclosed. The server for predicting a disease through AI-based iris analysis comprises: an image preprocessing unit that performs preprocessing on an iris image received from a user terminal; a prediction model training unit that trains an iris-based disease prediction model in a supervised manner using pre-collected learning data; a prediction model tuning unit that performs additional learning on the supervised iris-based disease prediction model; and a disease prediction unit that determines a predicted disease corresponding to the iris image using candidate diseases obtained from the additionally trained iris-based disease prediction model and transmits disease prediction information including the determined predicted disease to the user terminal.

Description

Server and its operation method for predicting diseases through AI-based iris analysis

[0001] The present invention relates to AI technology, and more particularly, to a server for predicting diseases through AI-based iris analysis and its operating method.

[0002]

[0003] Predicting diseases in advance can save lives and prevent them from developing into serious illnesses, making predicting them a crucial part of the healthcare industry.

[0004] The iris is a donut-shaped membrane surrounding the pupil. Its function is to control the size of the pupil by contracting and relaxing, thereby regulating the amount of light entering the eye. The iris's color varies due to the pigment layer within it, and this variation is known to vary significantly across races and individuals.

[0005] Research results have reported that changes in the color of the iris or the occurrence of inflammation have a strong causal relationship with the onset of certain diseases, and the iris is attracting attention as one of the main factors for predicting diseases.

[0006] Meanwhile, as artificial intelligence (AI) has recently gained attention, efforts are being made to apply it to disease prediction, and in particular, deep learning-based AI is being utilized as a major research subject as its performance is recognized.

[0007]

[0008] The purpose of the present invention to solve the above problems is to provide a server and its operation method for predicting diseases through AI-based iris analysis.

[0009]

[0010] One aspect of the present invention to achieve the above object provides a server for predicting diseases through AI-based iris analysis.

[0011] A server for predicting a disease through the AI-based iris analysis includes an image preprocessing unit that performs preprocessing on an iris image received from a user terminal; a prediction model learning unit that supervises an iris-based disease prediction model using pre-collected learning data; a prediction model tuning unit that performs additional learning on the supervised-learned iris-based disease prediction model; and a disease prediction unit that determines a predicted disease corresponding to the iris image using candidate diseases obtained from the additionally-learned iris-based disease prediction model and transmits disease prediction information including the determined predicted disease to the user terminal.

[0012] The image preprocessing unit filters the iris image using a predefined image filter, searches for two points corresponding to the outer corners of the eyes belonging to the left and right edge areas of the eyes in the filtered iris image, generates a reference line which is a straight line connecting the two searched points, rotates the iris image by a predetermined angle so that the generated reference line is parallel, detects an outline in the rotated iris image, and extracts the pupil area from the outermost circular portion of the detected outline, thereby obtaining an iris image to be analyzed.

[0013] The above prediction model learning unit includes a learning data management unit that corrects the collected learning data using an iris coordinate system and performs supervised learning for the iris-based disease prediction model using the corrected learning data; and the iris-based disease prediction model that predicts a candidate disease based on the iris image.

[0014] The above iris coordinate system includes an inner circle region (ICC) located at the most central part of the pupil region, a central circle region (CC) sharing the same center point as the inner circle region (ICC) and spaced apart from the central circle region (CC) by a predetermined first distance in the radial direction, an outer circle region (ECC) sharing the same center point as the central circle region (CC) and spaced apart from the central circle region (CC) by a predetermined second distance in the radial direction, and an outer circle region (OCC) sharing the same center point as the outer circle region (ECC) and spaced apart from the central circle region (CC) by a predetermined third distance in the radial direction to form the outermost outline line of the pupil region.

[0015] The first region between the inner circumferential region (ICC) and the central circumferential region (CC) corresponds to the viscera, including the transverse colon, small intestine, cecum, duodenum, and ascending colon.

[0016] The second region between the outer circumferential region (ECC) and the outer circumferential region (OCC) is the region corresponding to the skin, lymphatic circulation, and cerebral blood vessels.

[0017]

[0018] When using the server and its operation method for predicting diseases through AI-based iris analysis according to the present invention as described above, it is possible to predict various types of diseases with a fairly high degree of accuracy using iris images, thereby enabling early detection of the onset of the disease.

[0019] Additionally, there is an advantage in that prediction accuracy can be further improved by using two artificial neural networks simultaneously.

[0020]

[0021] FIG. 1 is a conceptual diagram illustrating a server and its operation method for predicting a disease through AI-based iris analysis according to one embodiment.

[0022] Figure 2 is a block diagram showing the functional configuration of the server according to Figure 1.

[0023] Figure 3 is a conceptual diagram for explaining the operation of the image preprocessing unit according to Figure 2.

[0024] Figures 4a and 4b are coordinate systems showing areas according to the disease to be predicted.

[0025] FIG. 5 is a drawing showing the iris coordinate system corresponding to the 'right eye' in the iris coordinate system according to FIGS. 4a and 4b.

[0026] Figure 6 is a drawing for explaining the operation of the learning data management unit according to Figure 2.

[0027] Figure 7 is a conceptual diagram illustrating a process in which a learning data management unit according to one embodiment generates disease-related indicators for generating learning input data.

[0028] FIG. 8a, FIG. 8b, FIG. 9a, and FIG. 9b are additional conceptual diagrams of a learning data management unit according to one embodiment for producing disease-related indicators.

[0029] FIG. 10 is a conceptual diagram illustrating a process of predicting a disease by simultaneously using two types of iris-based disease prediction models according to one embodiment.

[0030] FIG. 11 is a diagram illustrating an exemplary hardware configuration of a server for predicting a disease through AI-based iris analysis according to one embodiment.

[0031]

[0032] The present invention is susceptible to various modifications and embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the present invention. Throughout the description of each drawing, similar reference numerals have been used to designate similar components.

[0033] Terms such as first, second, A, and B may be used to describe various components, but these components should not be limited by these terms. These terms are used solely to distinguish one component from another.

[0034] For example, without departing from the scope of the present invention, the first component could be referred to as the second component, and similarly, the second component could also be referred to as the first component.

[0035] The term and / or includes any combination of a plurality of related described items or any one of a plurality of related described items.

[0036] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0037] The terminology used in this application is only used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0038] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0039]

[0040] Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the attached drawings.

[0041] FIG. 1 is a conceptual diagram illustrating a server and its operation method for predicting a disease through AI-based iris analysis according to one embodiment.

[0042] Referring to FIG. 1, a server (100, hereinafter abbreviated as “server”) for predicting a disease through AI-based iris analysis can provide a disease prediction service to a user of a user terminal (200) by communicating with the user terminal (200) using a wired or wireless network.

[0043] For example, the user terminal (200) can capture and obtain an iris image of the user using a camera built into the user terminal (200), or can obtain an iris image of the user by communicating with another external computing device.

[0044] Here, the iris image may be a two-dimensional (2D) image capturing the 'eye' area including the user's left eye and / or right eye.

[0045] The user terminal (200) can transmit the acquired iris image to the server (100), and the server (100) can analyze the received iris image using a pre-supervised learning iris-based disease prediction model (102b), determine a predicted disease indicating a disease with a high probability of onset or a disease that has already occurred in the iris image, and generate disease prediction information including the determined predicted disease.

[0046] For example, disease prediction information may include the probability of occurrence of the predicted disease, the time of occurrence, treatment and management methods for the predicted disease, and information on medical institutions for treating the predicted disease.

[0047] When the server (100) transmits the generated disease prediction information to the user terminal (200), the user terminal (200) can display the received disease prediction information to the user of the user terminal (200).

[0048] Examples of the user terminal (200) may include a desktop computer, a laptop computer, a notebook, a smart phone, a tablet PC, a mobile phone, a smart watch, smart glasses, an e-book reader, a portable multimedia player (PMP), a portable game console, a navigation device, a digital camera, a digital multimedia broadcasting (DMB) player, a digital audio recorder, a digital audio player, a digital video recorder, a digital video player, a PDA (Personal Digital Assistant), etc.

[0049]

[0050] Fig. 2 is a block diagram showing the functional configuration of the server according to Fig. 1. Fig. 3 is a conceptual diagram for explaining the operation of the image preprocessing unit according to Fig. 2.

[0051] Referring to FIG. 2, the server (100) may include an image preprocessing unit (101) that performs preprocessing on an iris image, a prediction model learning unit (102) that performs supervised learning on an iris-based disease prediction model (102b) using pre-collected learning data, a prediction model tuning unit (103) that performs additional learning on the supervised-learned iris-based disease prediction model (102b), and a disease prediction unit (104) that determines a predicted disease corresponding to an iris image of a user terminal (200) using candidate diseases obtained from the additionally-learned iris-based disease prediction model (102b) and transmits disease prediction information including the determined predicted disease to the user terminal (200).

[0052] The image preprocessing unit (101) can filter the iris image using a predefined image filter. For example, the image filter may be a Gaussian filter or a median filter that removes noise such as inaccurate pixel values ​​or defective pixels in the photograph from the iris image, a sharpening filter that emphasizes outlines, a correction filter that brightens or darkens brightness or saturation, etc. Since various filters for such image processing are known, a person skilled in the art can use one of the known image filters.

[0053] The image preprocessing unit (101) can search for two points corresponding to the outer corners of the eyes, each belonging to the left and right edge areas of the eye, in the iris image, and generate a reference line, which is a straight line connecting the two searched points. The image preprocessing unit (101) can rotate the iris image by a predetermined angle so that the generated reference lines are parallel.

[0054] The image preprocessing unit (101) can obtain an iris image to be analyzed by detecting an outline from a rotated iris image and extracting the pupil area from the outermost circular portion of the detected outline.

[0055] The prediction model learning unit (102) may include a learning data management unit (102a) that corrects collected learning data using an iris coordinate system and performs supervised learning on an iris-based disease prediction model (102b) using the corrected learning data, and an iris-based disease prediction model (102b) that predicts a candidate disease based on an iris image.

[0056] Here, the learning data may include learning iris images in which the eye area of ​​each of various users is photographed, and preprocessing is performed by the image preprocessing unit (101) on the iris images including the photographed eye area, and correct labels that are the correct answers (labels) for each of the learning iris images, which are diseases of the user who photographed the iris images.

[0057] The learning data management unit (102a) can generate learning input data by comparing each of the learning iris images included in the learning data with a predefined iris coordinate system.

[0058] In addition, the learning data management unit (102a) may generate input data by comparing the iris image received from the user terminal (200) with a predefined iris coordinate system, and since it is obvious that the learning input data and the input data are generated in the same manner, the process of generating the learning input data will be specifically described below to avoid redundant explanation.

[0059] The prediction model tuning unit (103) operates to optimize the parameters of the iris-based disease prediction model (102b). For example, the prediction model tuning unit (103) may receive feedback data from the user terminal (200) indicating whether the predicted disease provided to the user terminal (200) is a disease that has actually occurred (or a disease that has occurred in the past). In this case, the feedback data includes data indicating the type of disease that has actually occurred and is different from the predicted disease.

[0060] If the candidate disease is confirmed through feedback data to be not an actual disease, the prediction model tuning unit (103) sets the actual disease that is different from the predicted disease as the correct answer, uses the iris image provided from the user terminal (200) as a learning iris image to generate learning input data, and uses the generated learning input data to further train the iris-based disease prediction model (102b).

[0061] In this way, by using feedback data obtained during the service operation process to further tune the prediction model, the prediction model operates so that its accuracy can continuously improve.

[0062]

[0063] Figures 4a and 4b are coordinate systems representing areas according to diseases to be predicted. Figure 5 is a drawing showing the iris coordinate system corresponding to the 'right eye' in the iris coordinate system according to Figures 4a and 4b alone.

[0064] Referring to FIGS. 4A and 4B , an iris coordinate system according to one embodiment may include an inner circle region (ICC) located at the most central portion of the pupil region, a central circle region (CC) sharing the same center point as the inner circle region (ICC) and spaced apart from the central circle region (CC) by a predetermined first distance in the radial direction, an outer circle region (ECC) sharing the same center point as the central circle region (CC) and spaced apart from the central circle region (CC) by a predetermined second distance in the radial direction, and an outer circle region (OCC) sharing the same center point as the outer circle region (ECC) and spaced apart from the central circle region (ECC) by a predetermined third distance in the radial direction to form the outermost outline line of the pupil region.

[0065] At this time, referring to FIG. 5, the first region between the inner circle region (ICC) and the central circle region (CC) may be a region corresponding to internal organs including the transverse colon, small intestine, cecum, duodenum, and ascending colon. For example, as shown in FIG. 5, the region between 10 o'clock and 1 o'clock in the first region may be a region corresponding to the transverse colon, the region between 1 o'clock and 5 o'clock may be a region corresponding to the small intestine, the region between 5 o'clock and 7 o'clock may be a region corresponding to the cecum, the region between 7 o'clock and 8 o'clock may be a region corresponding to the duodenum, and the region between 8 o'clock and 10 o'clock may be a region corresponding to the ascending colon.

[0066] Additionally, the second region between the outer circle region (ECC) and the outer circle region (OCC) may correspond to the skin, the lymphatic circulation, and the cerebral blood vessels. For example, the region between 11 o'clock and 12 o'clock and some regions between 12 o'clock and 1 o'clock in the second region correspond to the cerebral blood vessels, the region between 1 o'clock and 6 o'clock and the region between 6 o'clock and 11 o'clock in the second region and a region spaced from the outermost contour line to the inner center by a predetermined interval corresponds to the lymphatic circulation, and the region that surrounds the region corresponding to the lymphatic circulation from the outside and continues to the outermost contour line may be defined as the region corresponding to the skin.

[0067] However, in the present invention, the area between the precise time angles can be finely adjusted and is not clearly limited to a specific angle. It should be understood that the area can be changed by a predetermined time angle error range centered on the area indicated in the drawing. Below, a person skilled in the art can interpret the area corresponding to each part by referring to the corresponding body part for each area described in FIG. 5, so further description is omitted.

[0068] In addition, in Fig. 5, the iris coordinate system corresponding to the 'right eye' is illustrated, but the iris coordinate system corresponding to the 'left eye' can be configured in a form that is the left-right inversion of the iris coordinate system corresponding to the 'right eye', and this will be collectively referred to as the iris coordinate system.

[0069]

[0070] FIG. 6 is a diagram illustrating the operation of the learning data management unit according to FIG. 2. FIG. 7 is a conceptual diagram illustrating the process of the learning data management unit according to one embodiment generating disease-related indicators for generating learning input data. FIG. 8a, FIG. 8b, and FIG. 9a, and FIG. 9b are additional conceptual diagrams illustrating the process of the learning data management unit according to one embodiment generating disease-related indicators.

[0071] Referring to FIG. 6, the learning data management unit (102a) can correct the size of the iris coordinate system to correspond to the learning iris image and generate learning input data using the corrected learning iris image.

[0072] For example, referring to FIG. 6, the inner circle area (ICC) and the pupil of the learning iris image have different sizes, and the size of the pupil of the learning iris image and the outermost line of the pupil may change depending on the time of capture, lighting environment, and distance, and thus adjustments are required.

[0073] To this end, the learning data management unit (102a) can correct the size of the iris coordinate system so that the outer line of the outer circle area (OCC) of the iris coordinate system matches the outermost line of the pupil extracted from the learning iris image, and can correct the inner circle area (ICC) to contract or expand so that the inner circle area (ICC) matches the pupil while the outer circle area (OCC) is fixed.

[0074] The learning data management unit (102a) can correct the positions of the central circle area (CC), the outer circle area (ECC), and the outer circle area (OCC) by dividing the separation radius between the inner circle area (ICC) and the outer circle area (OCC) into a first separation distance to a third separation distance according to a predetermined ratio based on the inner circle area (ICC) corrected to match the pupil.

[0075] That is, in the case of the present invention, the iris coordinate system is corrected by utilizing the characteristics of the pupil that contracts and expands according to the lighting environment or environmental changes, thereby supporting accurate matching of the iris coordinate system regardless of the environment in which the iris image was taken.

[0076] In one embodiment, the learning data management unit (102a) can generate first learning input data by comparing a corrected iris coordinate system with a learning iris image, dividing the learning iris image into regions corresponding to each of a plurality of body parts indicated by the iris coordinate system, and matching diseases corresponding to each of the divided regions to the body parts and labeling the corresponding learning iris image.

[0077] In another embodiment, the learning data management unit (102a) may compare the corrected iris coordinate system with the learning iris image, divide the learning iris image into regions corresponding to each of a plurality of body parts indicated by the iris coordinate system, calculate predefined disease-related indices from each of the divided regions, and generate second learning input data including the calculated disease-related indices and the labeled correct answer.

[0078] In the case of the above other embodiment, the method of defining lines, colors, etc. that can be identified in each iris area as disease-related indicators for learning has the advantage of being advantageous for application to a model using a non-image-based artificial neural network and advantageous for implementation with only a small amount of sample data (learning data).

[0079] For example, referring to FIG. 7, the learning data management unit (102a) can calculate the number of 'intrinsic lines', which are lines detected only within each segmented region as disease-related indicators, and calculate the 'gray area ratio', which is the ratio (%) of the area occupied by a predetermined pixel value filling the segmented region (for example, the pixel value may be a pixel value indicating gray or may include pixel values ​​within a predetermined error range from the pixel value) in the segmented region.

[0080] In addition, for each divided area, the number of 'protrusion lines', which are lines that protrude out of the area, can be calculated, and the mean pixel value indicating the average of the pixel values ​​that constitute the area, the variance pixel value indicating the variance of the pixel values ​​that constitute the area, and the median pixel value indicating the median of the pixel values ​​that constitute the area can also be calculated.

[0081] In summary, disease-related indicators may include the gray area ratio, the number of intrinsic lines, the number of protruding lines, the average pixel value, the variance pixel value, and the median pixel value calculated from each segmented area.

[0082] Meanwhile, for certain diseases, the association with the development of the disease is not limited to a single segmented area, but rather can be seen across multiple segmented areas. For example, in the case of hyperlipidemia, as shown in Figures 8a and 8b, a shape filled with mucus is characteristically observed at the bottom of the pupil, and in the case of osteoporosis, as shown in Figures 9a and 9b, a shape filled with mucus may be observed across multiple segmented areas located in the outer direction between the 3 and 5 o'clock positions of the pupil.

[0083] Considering these points, for hyperlipidemia, a predetermined number of segmented regions located in the lower region between the 3 o'clock and 9 o'clock positions can be treated as a single region, and the aforementioned disease-related indicators can be calculated from that region. Similarly, for osteoporosis and other diseases, a predetermined number of segmented regions can be treated as a single region, and disease-related indicators can be calculated from that region.

[0084] The learning data management unit (102a) inputs first learning input data and / or second learning input data into an iris-based disease prediction model (102b), compares candidate diseases obtained as outputs of the iris-based disease prediction model (102b) with correct answers matched to the iris images used to generate the first learning input data or the second learning input data, and can perform supervised learning in a manner of minimizing the error.

[0085] For example, the learning data management unit (102a) can generate a loss function using 'candidate disease' and 'correct answer' as the first and second output values, respectively, and supervise learning the iris-based disease prediction model (102b) so that the result value of the generated loss function is minimized.

[0086] Here, a typical technician can choose to use one of the known loss functions such as Mean Squared Error (MSE), Binary Cross-Entropy, Categorical Cross-Entropy, Huber Loss, etc.

[0087] Here, supervised learning means continuously adjusting the values ​​of hyper parameters that constitute the iris-based disease prediction model (102b) until the result value of the loss function is minimized. Hyper parameters are the number of layers that constitute the neural network, the number of neurons in each layer, etc., and since anyone who is a general engineer who can handle artificial neural networks can understand them, a detailed explanation is omitted.

[0088] In addition, with regard to the method of adjusting hyperparameters so that the result value of the loss function is minimized, various optimization algorithms for fast minimization, such as Gradient Descent, Stochastic Gradient Descent, and Momentum, are known, so ordinary technicians can use them.

[0089]

[0090] In one embodiment, when an image such as the first learning input data is used as input data, the iris-based disease prediction model (102b) may be an artificial neural network based on a CNN (Convolutional Neural Network).

[0091] When the input data is composed of disease-related indicators for each segmented area, such as the second learning input data, the iris-based disease prediction model (102b) may be an artificial neural network based on LSTM (Long short time memory).

[0092] The structure of artificial neural networks based on CNN or LSTM is distributed and disclosed in various forms of structures and pre-defined libraries, so that ordinary engineers can easily implement it by setting parameters using public data such as Google's Tensorflow. Therefore, a description of the detailed structure of the specific artificial neural network is omitted.

[0093]

[0094] FIG. 10 is a conceptual diagram illustrating a process of predicting a disease by simultaneously using two types of iris-based disease prediction models according to one embodiment.

[0095] Referring to FIG. 10, the iris-based disease prediction model (102b) may be composed of a first artificial neural network (CNN) based on a convolutional neural network (CNN) and a second artificial neural network (LSTM) based on a long short time memory (LSTM).

[0096] The first artificial neural network (CNN) can be pre-supervised learning using the first learning input data, and the second artificial neural network (LSTM) can be pre-supervised learning using the second learning input data.

[0097] The learning data management unit (102a) can generate the first input data using an iris image (i.e., an image to be subject to disease prediction) received from the user terminal (200) instead of the learning iris image in the same manner as the process of generating the first learning input data, and can generate the second input data using an iris image (i.e., an image to be subject to disease prediction) received from the user terminal (200) instead of the learning iris image in the same manner as the process of generating the second learning input data.

[0098] The learning data management unit (102a) can input the generated first input data into a first artificial neural network (CNN) to obtain a first candidate disease, and input the second input data into a second artificial neural network (LSTM) to obtain a second candidate disease. At this time, the learning data management unit (102a) can obtain a 1-1 incidence probability indicating an incidence probability corresponding to the first candidate disease and a 1-2 incidence probability indicating an incidence probability of a disease corresponding to the next priority of the first candidate disease as an output of the first artificial neural network (CNN), and can obtain a 2-1 incidence probability indicating an incidence probability corresponding to the second candidate disease and a 2-2 incidence probability indicating an incidence probability of a disease corresponding to the next priority of the second candidate disease as an output of the second artificial neural network (LSTM).

[0099] The disease prediction unit (104) can determine the predicted disease by selecting one of the first candidate disease and the second candidate disease.

[0100] For example, the disease prediction unit (104) can determine the first candidate disease as the predicted disease if the selection criterion value calculated using the following mathematical formula is a positive number, and can determine the second candidate disease as the predicted disease if the selection criterion value is a negative number.

[0101]

[0102] In the above mathematical expression 1, RV is the selection criterion value, P1 is the 1-1 occurrence probability, P2 is the 2-1 occurrence probability, SP1 is the 1-2 occurrence probability, and SP2 is the 2-2 occurrence probability.

[0103] According to the embodiment according to FIG. 10, by operating in a manner of selecting one of the candidate diseases according to each of the two neural networks, the prediction accuracy of each of the two neural networks for each disease type can be increased, and in particular, since the difference between the highest incidence probability and the next-ranked invention probability in the neural network is reflected as a correction value in the mathematical expression 1 above, in a prediction situation where the incidence probability of the disease is particularly higher than that of other diseases, the possibility of selecting the disease as the predicted disease increases, so there is an advantage of minimizing the prediction error.

[0104] Meanwhile, it is not necessarily limited to the embodiment according to FIG. 10, and it is also possible for the disease prediction unit (104) to determine a candidate disease obtained using any one of the first and second artificial neural networks as a predicted disease.

[0105]

[0106] FIG. 11 is a diagram illustrating an example of a hardware configuration of a server (100) for predicting a disease through AI-based iris analysis according to one embodiment.

[0107] Referring to FIG. 11, the server (100) may include at least one processor (110) and a memory (120) that stores instructions that instruct the at least one processor (110) to perform at least one operation.

[0108] Here, at least one operation may be interpreted as including an operation or function of the functional units included in the server (100) described with reference to FIGS. 1 to 10.

[0109] Here, at least one processor (110) may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed.

[0110] The memory (120) may be composed of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (120) may be one of a read-only memory (ROM) and a random access memory (RAM).

[0111] The server (100) may further include a storage device (160) that stores temporary data, input data, intermediate processing data, output data, etc. for performing at least one of the above operations. The storage device (160) may be a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or various memory cards (e.g., a micro SD card).

[0112] In addition, the server (100) may include a transceiver (130) that performs communication via a wireless network. The server (100) may further include an input interface device (140), an output interface device (150), etc. Each component included in the server (100) may be connected by a bus (170) and communicate with each other.

[0113]

[0114] The methods according to the present invention may be implemented in the form of program instructions that can be executed by various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either singly or in combination. The program instructions recorded on the computer-readable medium may be those specifically designed and constructed for the present invention, or may be known and available to those skilled in the computer software art.

[0115] Examples of computer-readable media may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions may include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate with at least one software module to perform the operations of the present invention, and vice versa.

[0116] Additionally, the above-described method or device may be implemented by combining all or part of its configuration or function, or may be implemented separately.

[0117] Although the present invention has been described above with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various modifications and changes may be made to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below.

[0118]

[0119] As described above, the present invention can be widely used in the healthcare industry because it can predict diseases through iris analysis.

[0120]

[0121] 100: A server for predicting diseases through AI-based iris analysis.

[0122] 101: Image Preprocessing

[0123] 102: Prediction Model Learning Section

[0124] 102a: Learning Data Management Department

[0125] 102b: Iris-based disease prediction model

[0126] 103: Prediction Model Tuning Section

[0127] 104: Disease Prediction Department

Claims

1. As a server for predicting diseases through AI-based iris analysis, Preprocessing of iris images received from user terminals Image preprocessing unit; A prediction model learning unit that supervises learning of an iris-based disease prediction model using pre-collected learning data; A prediction model tuning unit that performs additional learning on the supervised-learned iris-based disease prediction model; and A disease prediction unit that determines a predicted disease corresponding to the iris image using a candidate disease obtained from the additionally learned iris-based disease prediction model and transmits disease prediction information including the determined predicted disease to a user terminal; A server for predicting diseases through AI-based iris analysis.

2. In claim 1, The above image preprocessing unit filters the iris image using a predefined image filter, In the filtered iris image, two points corresponding to the outer corners of the eyes belonging to the left and right edge areas of the eyes are searched, and a reference line, which is a straight line connecting the two searched points, is created. Rotate the iris image by a predetermined angle so that the generated reference lines are parallel, Detecting the outline in the rotated iris image, By extracting the pupil area from the outermost circular portion of the detected outline, an iris image to be analyzed is obtained. A server for predicting diseases through AI-based iris analysis.

3. In claim 2, The above prediction model learning unit is, A learning data management unit that corrects the collected learning data using an iris coordinate system and performs supervised learning on the iris-based disease prediction model using the corrected learning data; and an iris-based disease prediction model that predicts a candidate disease based on the iris image, The above iris coordinate system includes an inner circle region (ICC) located at the most central part of the pupil region, a central circle region (CC) sharing the same center point as the inner circle region (ICC) and spaced apart from the radial direction by a predetermined first distance, an outer circle region (ECC) sharing the same center point as the central circle region (CC) and spaced apart from the radial direction by a predetermined second distance, and an outer circle region (OCC) sharing the same center point as the outer circle region (ECC) and spaced apart from the radial direction by a predetermined third distance to form the outermost outline line of the pupil region. The first region between the inner circumferential region (ICC) and the central circumferential region (CC) corresponds to the intestines including the transverse colon, small intestine, cecum, duodenum, and ascending colon. The second region between the outer circumference region (ECC) and the outer circumference region (OCC) is the region corresponding to the skin, lymphatic circulation, and cerebral blood vessels. A server for predicting diseases through AI-based iris analysis.

Images