KR20200051289A - Apparatus and method for diagnosing disease based on iris image - Google Patents

Abstract

According to an embodiment of the present invention, provided is a device for diagnosing a disease which can accurately diagnose the disease. The device for diagnosing a disease comprises: a camera configured to photograph eyes of a user; a display; and a processor electrically connected to the camera and the display. The processor obtains an iris image of the user by using the camera, detects a pattern related to the disease from the iris image by analyzing the iris image, and outputs disease information corresponding to the detected pattern by using the display.

Description

A device and method for diagnosing disease based on iris images {APPARATUS AND METHOD FOR DIAGNOSING DISEASE BASED ON IRIS IMAGE}

Embodiments disclosed in this document relate to a technique for diagnosing a disease using an iris image.

Iris analysis can predict the condition of the body's interior, for example, the digestive system, nervous system, musculoskeletal system, and cardiovascular system, and is currently used as a method for medical examination in various countries. The iris has numerous nerve endings, capillaries, and muscle fiber tissue, and is connected to all organs and tissues through the brain and nervous system. In particular, since the nerve endings of the brain are intensively distributed in the iris, the diseased situation due to chemical and physical changes occurring in each tissue and organ in the body may appear in the form of iris, cracks, color, rings, and / or wrinkles. have. Therefore, iridology may serve as a diagnostic indicator for systemic health through analysis of the iris pattern.

Markings appearing on the iris can reflect even the early stages of the disease state, which are difficult to detect by other diagnostic equipment, and thus iris diagnosis is increasingly used in terms of early detection and preventive medicine. However, in spite of the above-mentioned advantages, the accuracy of the iris diagnosis may vary depending on the individual skill of the doctor, the angle at which the iris is photographed, and whether light is irradiated.

In order to solve the above-mentioned problems and the problems raised in this document, the embodiments disclosed in this document aim to provide a disease diagnosis apparatus capable of accurately diagnosing a disease by analyzing an iris image.

The problems of the present invention are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the following description.

The apparatus for diagnosing a disease according to an embodiment of the present invention includes a camera configured to photograph a user's eye, a processor electrically connected to the camera and the display, and the processor acquires an iris image of the user using the camera, By analyzing the iris image, a pattern associated with the disease may be detected in the iris image, and disease information corresponding to the detected pattern may be output using a display.

According to an embodiment, the device may further include a position adjuster connected to the camera and configured to adjust the position of the camera.

According to an embodiment of the present disclosure, the processor may identify the iris region centering on the pupil of the user's eye in the iris image.

According to one embodiment, the processor may acquire an iris image by removing noise from an image photographed using a camera.

According to an embodiment, the device further includes illumination configured to emit light for photographing the camera, and the processor may control at least some of the direction or brightness of the illumination when photographing the camera.

According to one embodiment, the device further includes a proximity sensor, and the processor may recognize the user's proximity to the camera using the proximity sensor when the camera is photographed.

According to one embodiment, the processor may acquire a plurality of images by using a camera, and acquire an iris image by synthesizing the plurality of images.

According to an embodiment, the processor may detect a pattern associated with a disease by converting an iris image into a 3D image including a depth value and analyzing the converted 3D image.

According to an embodiment of the present disclosure, the processor may mask a region related to a disease in an iris image, and convert the masked region into a 3D image.

According to an embodiment, the processor may use a 3D image as an reinforcement learning input value for an analysis model of an iris image.

According to an embodiment, the processor may analyze the iris image using an analysis model of the iris image.

According to an embodiment, the processor may detect a pattern associated with at least some of cancer or dementia in the iris image.

According to an embodiment, the processor may detect at least a portion of a porphyrin pigment, a giant lacuna, or a defect pattern in an iris image.

According to one embodiment, the processor may detect at least some of the beta amyloid precipitate or the toxic protein pattern in the iris image.

According to an embodiment, the processor may detect a color included in the iris in the iris image.

According to one embodiment, the disease information may include information associated with at least some of the findings of the disease, the progress of the disease, or the genetic properties of the user.

According to an embodiment, the processor may output an iris image together with disease information using a display.

According to an embodiment, the processor may output an iris image including a marker for identifying an area in which a pattern associated with a disease is detected using a display.

According to one embodiment, the device further includes a communication circuit configured to communicate with the server, and the processor may transmit the iris image to the server using the communication circuit.

According to an embodiment, the communication circuit further includes a communication circuit configured to communicate with the server, and the processor may acquire an analysis model of the iris image from the server using the communication circuit.

A method of diagnosing a disease based on an iris image according to an embodiment of the present invention includes obtaining a iris image of a user using a camera, detecting a pattern associated with a disease in an iris image by analyzing the iris image, and detected patterns It may include the step of outputting the disease information corresponding to.

According to the embodiments disclosed in the present document, by analyzing an iris image and detecting a pattern associated with a disease, a disease occurring in the body may be diagnosed early in a harmless manner through a simple method.

In addition, the iris image suitable for diagnosis may be obtained by controlling the position of the camera, the brightness and direction of the lighting, etc. when photographing the iris image.

In addition, by using machine learning to provide information about the disease, and by continuously updating the diagnostic model, it is possible to accurately and quickly diagnose the disease.

In addition, various effects that can be directly or indirectly identified through this document may be provided.

Figure 1 shows the appearance of the disease diagnosis apparatus according to an embodiment.
2 is a block diagram showing the configuration of a disease diagnosis apparatus according to an embodiment.
3 is a flowchart illustrating a disease diagnosis method according to an embodiment.
4 is a flowchart illustrating a disease diagnosis method according to an embodiment.
5 is a flowchart illustrating a disease diagnosis method according to an embodiment.
6 is a diagram for explaining a learning model used in a disease diagnosis apparatus according to an embodiment.
7 is a block diagram showing the configuration of a disease diagnosis apparatus according to an embodiment.
8 is a flowchart illustrating a disease diagnosis method according to an embodiment.
9 is a diagram for explaining a learning model used in a disease diagnosis apparatus according to an embodiment.
10 to 14 illustrate exemplary images photographed by a disease diagnosis apparatus according to an embodiment.
15 is a diagram showing a receiver operating characteristic (ROC) curve of a reinforcement learning model used in a disease diagnosis apparatus according to an embodiment.
16 shows exemplary disease information output from the disease diagnosis apparatus according to an embodiment.
In connection with the description of the drawings, the same or similar reference numerals may be used for the same or similar components.

Hereinafter, various embodiments of the present invention will be described with reference to the accompanying drawings. However, this is not intended to limit the present invention to specific embodiments, and should be understood as including various modifications, equivalents, and / or alternatives of embodiments of the present invention.

It should be understood that various embodiments of the document and terms used therein are not intended to limit the technology described in this document to specific embodiments, and include various modifications, equivalents, and / or replacements of the embodiments. Singular expressions may include plural expressions unless the context clearly indicates otherwise. In this document, expressions such as "A or B", "at least one of A and / or B", "A, B or C" or "at least one of A, B and / or C", etc. are all of the items listed together. Possible combinations may be included. Expressions such as "first," "second," "first," or "second," can modify the components, regardless of order or importance, to distinguish one component from another component It is used but does not limit the components. When it is said that one (eg, first) component is “connected (functionally or communicatively)” to another (eg, second) component or is “connected,” the component is the other It may be directly connected to the component, or may be connected through another component (eg, the third component). The size and thickness of each component shown in the drawings are illustrated for convenience of description, and the present invention is not necessarily limited to the size and thickness of the illustrated component.

Hereinafter, a disease diagnosis apparatus according to various embodiments will be described with reference to the accompanying drawings.

Figure 1 shows the appearance of the disease diagnosis apparatus according to an embodiment.

Referring to FIG. 1, the apparatus 100 for diagnosing a disease according to an embodiment includes a main body 100, an input unit 120, a display 130, a camera 140, a proximity sensor 150, and a position adjusting unit 160 ).

The main body 100 may include various hardware configurations (eg, a processor, communication circuit, and memory) employed in a computing device. The body unit 100 may include, for example, a machine learning (eg reinforcement learning) module, and the machine learning module may be mounted on the above-described hardware configuration, or may be configured with separate hardware. The main body 100 may acquire and analyze an image for diagnosis, and provide diagnostic results to the display 130 and / or an external server. The main body 100 may be connected to an external output device such as a printer.

The input unit 120 may be configured to receive input from a user (eg, a doctor). For example, the input unit 120 may include a lever (eg, joystick), a key and / or a button for inputting a direction. By input to the input unit 120, the position of the camera 140, the direction and brightness of the light, etc. can be controlled.

The display 130 may be configured to output images photographed by the camera 140 and diagnosis result information. The display 130 may be configured in various ways, such as LCD or LED, and may also include a touch sensor.

The camera 140 may be configured to photograph an eyeball of a user (eg, a patient). For example, the two cameras 140 may be disposed at positions capable of simultaneously photographing the left and right eyes, respectively.

The proximity sensor 150 may be configured to detect a user's approach when photographing. The proximity sensor 150 may be disposed at a position adjacent to the camera 140 to detect proximity when the user approaches the camera 140.

The position adjusting unit 160 may be configured to adjust the position of the camera 140. The position adjusting unit 160 may be connected to a portion where the camera 140 is mounted, and may move in the vertical direction according to an input to the input unit 120 to adjust the position of the camera 140. The position adjusting unit 160 may be configured to move in various directions as well as up and down directions.

2 is a block diagram showing the configuration of a disease diagnosis apparatus according to an embodiment.

Referring to FIG. 2, the apparatus 200 for diagnosing a disease according to an embodiment includes a camera 210, a proximity sensor 220, a lighting 230, a position adjusting unit 240, a display 250, and a communication circuit 260 ) And a processor 270. The disease diagnosis device 200 of FIG. 2 may be the same device as the device 100 of FIG. 1.

The camera 210 may be configured to photograph a user's eye. The camera 210 may be configured to simultaneously photograph the user's left and right eyes.

The proximity sensor 220 may be configured to detect a user's approach when photographing using the camera 210. The proximity sensor 220 may be disposed at a position adjacent to the camera 210.

The lighting 230 may be configured to emit light for imaging the camera 210. The brightness and direction of the light 230 may be variously adjusted to obtain an image suitable for diagnosis.

The position adjusting unit 240 may be configured to adjust the position of the camera 210. The position adjusting unit 240 may be physically connected to a portion where the camera 210 is disposed. The camera 210 may be moved by the movement of the position adjusting unit 240, and the position of the camera 210 may be adjusted to a position suitable for photographing a user's eye.

The display 250 may be configured to output visual information to a user (eg, a doctor and / or patient, etc.). The display 250 may output an image acquired by the camera 210, an image processed by the processor 270, and a diagnosis result.

The communication circuit 260 may be configured to communicate with an external server. The communication circuit 260 may transmit and receive data to and from an external server through various wired or wireless communication methods.

The processor 270 may be electrically connected to the camera 210, the proximity sensor 220, the lighting 230, the position adjusting unit 240, the display 250, and the communication circuit 260. The processor 270 may control the camera 210, the proximity sensor 220, the lighting 230, the position adjusting unit 240, the display 250, and the communication circuit 260, and perform various data processing and calculations. It can be done. The processor 270 may execute instructions stored in a memory (not shown) or perform the following operations using stored data and a learning model. The processor 270 may be implemented in one integrated form, or may be implemented in two or more separate configurations.

According to one embodiment, the processor 270 may acquire an iris image of the user using the camera 210. The processor 270 may recognize the user's proximity to the camera 210 using the proximity sensor 220 when the camera 210 is photographed. When the user is not close, the processor 270 may adjust the position of the camera 210 so that the camera 210 is close to the user's eye using the position adjusting unit 240. The processor 270 may control at least a part of the direction or brightness of the light 230 when the camera 210 is photographed to obtain an image suitable for analysis.

According to an embodiment of the present disclosure, the processor 270 may identify the iris region based on the pupil of the user's eye in the captured image. When the iris area is identified, the processor 270 may obtain an iris image by removing noise (eg, the remaining areas except the iris) from the captured image.

According to an embodiment of the present disclosure, the processor 270 may acquire a plurality of images by using the camera 210, and acquire an iris image by synthesizing the plurality of images. In some cases, it may be difficult to photograph the entire area of the iris with one shot. In this case, after performing a plurality of photographing, the processor 270 may acquire an image including the entire area of the iris by synthesizing the acquired image and the like.

According to an embodiment, the processor 270 may detect a pattern associated with a disease by converting an iris image into a 3D image including a depth value and analyzing the converted 3D image. The processor 270 may analyze a two-dimensional image and convert it into a three-dimensional image in order to derive a more accurate analysis result, and then analyze the three-dimensional image. The processor 270 may mask a region associated with a disease in an iris image and reduce the masked region to a 3D image in order to reduce the computation amount for 3D transformation. The processor 270 may use the 3D image as a reinforcement learning input value for the analysis model of the iris image. By using the pre-processed image as an input value for learning, the accuracy of analysis using a model can be improved.

According to an embodiment, the processor 270 may detect a pattern associated with a disease in the iris image by analyzing the iris image. For example, the processor 270 may analyze the iris image using an analysis model of the iris image. The processor 270 compares a pattern included in the iris image with a pre-stored pattern (for example, a learned pattern), and when the pattern included in the iris image corresponds to a pre-stored pattern, detects the pattern as a pattern associated with the disease Can be.

According to an embodiment, the processor 270 may detect a pattern associated with at least a portion of cancer or dementia in the iris image. For example, the processor 270 may detect at least a portion of a porphyrin pigment, a giant lacuna, or a defect pattern in an iris image. For another example, the processor 270 may detect at least a portion of a beta amyloid precipitate or a toxic protein pattern in an iris image. For another example, the processor 270 may detect a color included in the iris in the iris image. Based on the above-described pattern and color, the processor 270 may detect a user's disease (eg, cancer or dementia). Since the disease can be detected only by the image of the iris, simple diagnosis is possible, and the disease at an early stage can be detected.

According to an embodiment, the processor 270 may output disease information corresponding to the detected pattern using the display 250. Disease information may include, for example, information associated with at least some of the findings of the disease, the progress of the disease, or the genetic nature of the user. According to an embodiment, the processor 270 may output an iris image together with disease information using the display 250. For convenience of the user, the processor 270 may output an iris image including a marker for identifying an area where a pattern associated with a disease is detected.

According to one embodiment, the processor 270 may transmit the iris image to the server using the communication circuit 260. The iris image provided as a server may be used as learning data for an analysis model of the iris image. The processor 270 may acquire an analysis model of the iris image from the server using the communication circuit 260. The accuracy of diagnosis by the device can be improved by providing training data for the analysis model and updating the analysis model.

3 is a flowchart illustrating a disease diagnosis method according to an embodiment.

Hereinafter, it is assumed that the disease diagnosis apparatus 200 of FIG. 2 performs the process shown in the flowchart. In addition, in the description associated with the flowchart, operations described as being performed by the disease diagnosis apparatus may be understood to be controlled by the processor 270 of the disease diagnosis apparatus 200.

Referring to FIG. 3, in operation 310, the apparatus for diagnosing a disease may acquire an iris image of a user using a camera. For example, the apparatus for diagnosing a disease may acquire an iris image by photographing a user's eye and preprocessing the photographed image.

In step 320, the apparatus for diagnosing a disease may detect a pattern associated with a disease in the iris image by analyzing the iris image. For example, the disease diagnosis apparatus may analyze an iris image using an analysis model using reinforcement learning, and detect a pattern that appears when a disease occurs in a user's body in the iris image.

In operation 330, the disease diagnosis apparatus may output disease information corresponding to the detected pattern. For example, the apparatus for diagnosing a disease may provide information about the disease to the user when the iris includes a pattern that appears when the disease occurs.

4 is a flow chart for explaining a disease diagnosis method according to an embodiment, and describes in detail an image acquisition step using a camera and illumination.

According to one embodiment, the apparatus for diagnosing a disease may obtain an image suitable for analysis by adjusting the position of the camera, the brightness and direction of illumination, and may remove noise from the acquired image.

 Referring to FIG. 4, in operation 410, the apparatus for diagnosing a disease may acquire an image using a camera. For example, the apparatus for diagnosing a disease may acquire an image by photographing a user's eye.

In step 420, the disease diagnosis apparatus may determine whether the entire area of the iris is detected. For example, the apparatus for diagnosing a disease may recognize whether the entire area of the iris is included in the image by recognizing the shape of the iris included in the image.

If the entire area of the iris is not detected, in step 430, the disease diagnosis device may reposition the camera. For example, the disease diagnosis device may adjust the height of the camera according to the input by the input device.

If the entire area of the iris is detected, in step 450, the disease diagnosis device may photograph the iris including the pupil. According to an embodiment, before the iris is photographed, in step 440, the disease diagnosis device may control illumination to remove noise due to reflection and refraction of light. For example, the disease diagnosis device may control the brightness and direction of lighting.

In step 460, the apparatus for diagnosing a disease may search the iris region centering on the pupil. For example, the apparatus for diagnosing a disease may recognize the location of the pupil by performing image search, and detect an iris region located in the periphery around the pupil.

In step 470, the disease diagnosis device may remove noise from the image. For example, the apparatus for diagnosing a disease may determine and remove an image portion outside the iris region as noise.

In operation 480, the apparatus for diagnosing a disease may acquire an iris image from which noise is removed. Through the steps 410 to 480, the apparatus for diagnosing a disease may acquire an iris image suitable for analysis.

When it is difficult to detect the entire area of the iris through one shot, the disease diagnosis apparatus may acquire an iris image by synthesizing a plurality of images obtained through a plurality of shots. For example, the apparatus for diagnosing a disease may synthesize images after photographing the upper and lower parts of the eye.

5 is a flowchart for explaining a disease diagnosis method according to an embodiment, and a pre-processing step of converting a 2D image into a 3D image by adding a depth value will be described in detail.

According to an embodiment, the apparatus for diagnosing a disease may perform a transformation to classify an iris image in three dimensions by adding a depth value to a two-dimensional image. Thus, the disease diagnosis apparatus can accurately recognize soft tissues represented by a specific protein form, and distinguish them from other types of tissues. For example, beta-amyloid deposits and toxic protein patterns associated with dementia can be confused with surrounding tissue, which can be avoided by converting a two-dimensional image into a three-dimensional image.

Referring to FIG. 5, in operation 510, the apparatus for diagnosing a disease may receive an iris image. Here, the iris image may be an image obtained in step 480.

In operation 520, the apparatus for diagnosing a disease may mask a region associated with the disease among iris images. The disease diagnosis device may add a depth value to the masked area.

In operation 530, the apparatus for diagnosing a disease may normalize two-dimensional values of the iris image. For example, the apparatus for diagnosing a disease may normalize 2D data of an image and convert it into a value between 0 and 1. In operation 540, the apparatus for diagnosing a disease may perform a vector operation. For example, the apparatus for diagnosing a disease may generate a depth value by projecting an imaginary value that moves an image located on a plane in a depth direction. The vector operation can be performed by circulating all the pixels of the masked buffer. In operation 550, the apparatus for diagnosing a disease may generate a depth map using the result value of the vector operation.

In operation 560, the apparatus for diagnosing a disease may generate a 3D image by adding a depth map to the 2D image. Thereby, the pretreatment can be completed.

The pre-processed data may be used as an input value of the learning model, and the learning model may process data by reinforcement learning.

6 is a diagram for explaining a learning model used in a disease diagnosis apparatus according to an embodiment.

Referring to FIG. 6, the reinforcement learning model may be used for cluster analysis, association analysis, and network analysis. For example, the reinforcement learning model is a method in which an agent defined in a certain environment recognizes a current state and selects an action or a sequence of actions to maximize reward among selectable actions. In the reinforcement learning model according to an embodiment, an action may be determined by referring to the meaning of the association analysis primarily classified based on the input 3D image. At this time, information representing the current image area situation is defined as a state, and an environment is formed to constitute a layer (Layer: learning phase-distinct layer). Agents derive rewards as a result. At this time, there may be cases where the data of the already trained model is not sufficient. In the case of reinforcement learning, the value of the future is maximized on the basis of the reward by action when the past results are not sufficient. Efficient results can be achieved by learning and making self-study decisions. Therefore, reinforcement learning can be a very useful tool in medical image analysis. When the above-described process is repeated, a reward generated as an output may be recycled as an input by forming a virtuous cycle structure by feedback.

7 is a block diagram showing the configuration of a disease diagnosis apparatus according to an embodiment.

Referring to FIG. 7, the image acquisition unit 710 may acquire an image by photographing a user's eye. The diagnostic area converter 720 may perform pre-processing to generate an image to be analyzed. The reinforcement learning discrimination unit 730 may analyze an image using the learning model 740. The learning model 740 may be updated through the reinforcement learning discrimination unit 730. The feedback module 750 may recycle the result value derived from the reinforcement learning discrimination unit 730 as an input. The transmission and output module 760 may provide the result value to the user, or may provide it to an external server.

8 is a flowchart for explaining a disease diagnosis method according to an embodiment, and detailed steps of diagnosing a disease by analyzing a preprocessed image.

According to an embodiment of the present invention, the disease diagnosis apparatus finds a part having a semantic relation of an image in a trained model and indexes the similar items by grouping them, and classifies the similar items, and detects the protein pattern corresponding to the disease in the similar items first. The backtracking technique is applied to backtrack the accuracy of the detected pattern, and a post-processing step is fed back to input from the stochastic conclusion.

Referring to FIG. 8, when an iris image is input in step 805, in step 810, the disease diagnosis apparatus may perform a reinforcement learning algorithm.

To this end, the disease diagnosis apparatus performs a reinforcement learning algorithm using learning data including cancer diagnosis learning data and senile disease (dementia) learning data in step 815, and generates an iris diagnosis area learning model in step 820. can do. The disease diagnosis apparatus may generate a model (for example, a reinforcement learning model) by learning in advance a diagnosis area of an image in order to automatically distinguish an area to be diagnosed and perform a diagnosis operation. The learning model can be generated based on the learning iris image and the diagnosis site image, and learns patterns such as porphyrin pigment spots, fever, and defects in advance for cancer diagnosis, and is assigned to toxic proteins for diagnosis of senile disease (dementia). You can learn the corresponding pattern. In addition, the lesion signs of the iris have a different color according to the determination area, and an analysis process according to the color may be included as a submodule. The trained model can be used in step 810.

More precisely, Alzheimer's dementia is known to be caused by an excessive increase in the protein "beta amyloid" in the brain. When the concentration of beta amyloid increases, nerve cells in the brain are destroyed, and memory is eventually erased. Therefore, when diagnosing Alzheimer's dementia by brain biopsy or Positron Emission Tomography (PET), which can confirm protein distribution, beta amyloid is used as a major measure of disease diagnosis, that is, a biomarker. However, in elderly people over 80 years of age with normal cognitive function, the probability of developing amyloid PET is reduced to 25%. On the other hand, there is a need for early diagnosis regardless of age, since premature Alzheimer's disease and hereditary Alzheimer's disease dementia occur before age 65. Therefore, depending on the initial or progression of Alzheimer's, beta-amyloid proteins can be accumulated in the body, and senile disease (dementia) can be detected using the fact that certain protein patterns affect the iris.

In step 825, the disease diagnosis device may classify (index) the semantic association of the image. The agent constituting the state becomes an analysis data image and derives the reward as a result, and when the data of the pre-trained model is not sufficient, the Q-Value is calculated based on the result of the decision (action). One output, Q-Value and pixel classification can be derived even when a certain probability is reached.

In steps 830 and 835, the disease diagnosis device uses indexing results to detect beta-amyloid deposits and toxic protein patterns produced by Alzheimer's in iris tissues, and to detect porphyrin pigment spots, cavities, and defect patterns caused by cancer. Can be. In steps 830 and 835, the apparatus for diagnosing diseases attempts to analyze patterns by stacking layers of patterns, for example, 200 layers. In repeated experiments, excellent performance was shown around 200 floors, calculation time was excessive for the 250 floors and higher, and there was no improvement in the performance of the diagnostic results. On the other hand, about 150 pattern layers showed a result of about 0.576% on half of the Q-value used in the ROC (receiver operating characteristic) curve.

In operation 840, the apparatus for diagnosing a disease may apply a technique for tracking the accuracy of a pattern. Step 840 is a step performed primarily to add reliability to the result after the pattern analysis is completed, and performs backtracking to improve and verify the accuracy of the detected pattern. This is the process of evaluating the analysis information at the stage before outputting the data based on the primary findings. As a classifier of reinforcement learning and deep learning models, it is applied to verify its performance and results through the ROC curve. There is a problem in that an error is entered in the process of human judgment, the reading result is inconsistent, or inefficiency occurs. Therefore, two scales, sensitivity and specificity, can be used for the accuracy and efficiency of the analysis results. Sensitivity is the rate at which a person who has a disease is judged to have “disease”, and specificity is the rate at which a person without a disease is judged to have “no disease”.

In step 845, the disease diagnosis device may update the model through feedback. For example, the disease diagnosis device may reuse the Q-value derived as a final result as an input.

The disease diagnosis apparatus may display the iris diagnosis result and the image in step 850, transmit the iris data to the server in step 855, and output the result in step 860 using a printer.

9 is a diagram for explaining a learning model used in a disease diagnosis apparatus according to an embodiment, and describes a reinforcement learning classification process for generating medical supplementary information using a reinforcement learning model and its results in detail.

Referring to FIG. 9, when a patient's left and right iris images are input, the disease diagnosis apparatus may perform a reinforcement learning algorithm based on already learned iris models. The reinforcement learning model first establishes an “environment”, which can begin with the Markov decision process. For example, the apparatus for diagnosing a disease may define information representing a current image area condition entered as an input as a state, and configure an environment. The disease diagnosis device may constitute a layer. The agent constituting the state can derive Q-Value and pixel classification. This process can be used as a result of semantic analysis.

10 to 14 illustrate exemplary images photographed by a disease diagnosis apparatus according to an embodiment.

10 shows an iris image in which no disease pattern has been detected. The iris image shown in FIG. 10 does not include a disease-related pattern.

11 shows an iris image in which two disease patterns are detected, and the first disease pattern 1110 may correspond to dementia. The second disease pattern 1120 may correspond to prostate cancer. The first disease pattern 1110 and the second disease pattern 1120 may be detected by analyzing the iris image, and a diagnostic result may be output based on the detected pattern.

12 shows the pattern represented by the three-dimensional image. The disease diagnosis device may add a depth value to the area 1310 indicated in yellow in FIG. 13. The first pattern 1210 and the second pattern 1220 may be detected by using the 3D image. By attempting three-dimensionalization, it can be used to analyze layers of disease patterns produced by sediments caused by diseases.

14 shows an image in which an identification mark 1410 is displayed in an area in which a pattern corresponding to a disease is detected among iris images. The identification marker 1410 allows the user to easily identify the pattern corresponding to the disease.

Table 1 discloses findings mapped to the results of the analysis. If it is normal among the classifications in Table 1, the normal size and color of the readings and the sigmoid pattern stop at the 200th floor, and no information is displayed in the area and characteristics.

In the case of senile disease and prostate cancer, corresponding pattern information was detected in the Sigmoid 180 and 220 layers, respectively, and it can be seen that unique values corresponding to areas, characteristics, and locations appear. In particular, the color value for each region and property can be used as an auxiliary value for determining the trend variation of a specific protein, and the 3D depth value indicates how deep the depth layer of each disease is. Bend and wrinkle values are used to indicate the state of each region, which can also be used as auxiliary values.

By category Reading pattern Area, characteristics location Remark normal Normal size and Colors Sigmoid 200Layer: Stop None None - Senile disease Alzheimer's (protein pattern) Sigmoid
180Layer Findings: Color: normal,
3D Depth Layer: 0.98 point
Winding pattern:
10-20
Wrinkle pattern:
80 or more Left eye
: Top center,
Senile disease
Alzheimer's,
Specific protein
Pattern detection cancer Cancer Sigmoid
220Layer
Findings: Color: Dark black,
3D Depth Layer: 0.96 point
Winding pattern:
12-20
Wrinkle pattern:
85 or more Left eye
: Bottom center,
Right eye Prostate cancer,
Defect detection

15 is a diagram showing a receiver operating characteristic (ROC) curve of a reinforcement learning model used in a disease diagnosis apparatus according to an embodiment.

Referring to FIG. 15, the graph includes a curve visualizing sensitivity and specificity at the same time. The performance of the diagnostic technique can be quantified by calculating the area under the curve (AUC). If the AUC value is 1, it is a 100% perfect judgment method, and as a result of analyzing two patterns of cancer and dementia, the result was found to be over 0.9%. Specific values are listed in Table 2.

AUC (area under curve) Sensitivity (T = 0.5) Specificity (T = 0.5) cancer 0.94 95.6% 94.7% dementia 0.92 94.5% 93.5%

16 shows exemplary disease information output from the disease diagnosis apparatus according to an embodiment.

Referring to FIG. 16, the apparatus for diagnosing a disease according to an embodiment may display the progression of the disease. The degree of disease progression can be quantified and shown in bar graph form. For example, 0 to 150 indicate an initial state of progress of the invention, 160 to 170 indicate a patient's perceivable state, 160 to 180 indicate a state in which an outbreak can be determined from data such as CT or MRI, and 190 to 200 are serious The level indicates the progression of the onset. If the disease diagnosis apparatus according to an embodiment is used, a pattern may be detected and diagnosed from 150 before CT or MRI.

The embodiments of the present invention have been described in more detail with reference to the accompanying drawings, but the present invention is not necessarily limited to these embodiments, and may be variously modified without departing from the technical spirit of the present invention. . Therefore, the embodiments disclosed in the present invention are not intended to limit the technical spirit of the present invention, but to explain, and the scope of the technical spirit of the present invention is not limited by these embodiments. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. The scope of protection of the present invention should be interpreted by the claims below, and all technical spirits within the equivalent range should be interpreted as being included in the scope of the present invention.

Claims (21)

A camera configured to photograph the user's eye;
display; And
And a processor electrically connected to the camera and the display,
The processor,
Acquiring the iris image of the user using the camera,
By analyzing the iris image, the pattern associated with the disease is detected in the iris image,
And outputting disease information corresponding to the detected pattern using the display. According to claim 1,
A diagnostic device for disease based on the iris image, further comprising a position control unit connected to the camera and configured to adjust the position of the camera. According to claim 1,
The processor,
In the iris image, characterized in that to identify the iris area around the pupil of the user's eye, the apparatus for diagnosing a disease based on the iris image. According to claim 1,
The processor,
A device for diagnosing a disease based on an iris image, characterized by acquiring the iris image by removing noise from an image photographed using the camera. According to claim 1,
Further comprising illumination configured to emit light for shooting of the camera,
The processor,
A device for diagnosing a disease based on an iris image, characterized in that at least a part of the direction or brightness of the light is controlled when the camera is photographed. According to claim 1,
Further comprising a proximity sensor,
The processor,
A device for diagnosing a disease based on an iris image, characterized in that when the camera is photographed, the user's proximity to the camera is recognized using the proximity sensor. According to claim 1,
The processor,
Acquiring a plurality of images using the camera,
A device for diagnosing a disease based on an iris image, characterized by obtaining the iris image by synthesizing the plurality of images. According to claim 1,
The processor,
Converting the iris image into a three-dimensional image containing depth values,
A device for diagnosing a disease based on an iris image, characterized by detecting a pattern associated with the disease by analyzing the converted 3D image. The method of claim 8,
The processor,
Masking an area related to the disease among the iris images,
Device for diagnosing a disease based on an iris image, characterized in that the masked area is converted into the 3D image. The method of claim 8,
The processor,
Device for diagnosing a disease based on an iris image, characterized in that the 3D image is used as a reinforcement learning input value for an analysis model of the iris image. According to claim 1,
The processor,
And analyzing the iris image using the analysis model of the iris image. According to claim 1,
The processor,
In the iris image, characterized in that for detecting a pattern associated with at least a portion of cancer or dementia, a device for diagnosing a disease based on the iris image. According to claim 1,
The processor,
A device for diagnosing a disease based on an iris image, characterized by detecting at least a portion of a porphyrin pigment, a giant lacuna, or a defect pattern in the iris image. According to claim 1,
The processor,
A device for diagnosing a disease based on an iris image, characterized by detecting at least a portion of a beta amyloid precipitate or a toxic protein pattern in the iris image. According to claim 1,
The processor,
A device for diagnosing a disease based on an iris image, characterized by detecting a color included in the iris in the iris image. According to claim 1,
The disease information, the diagnosis of the disease based on the iris image, characterized in that it comprises information related to at least some of the findings of the disease, the progress of the disease or the genetic properties of the user. According to claim 1,
The processor,
Device for diagnosing a disease based on an iris image, characterized by outputting the iris image together with the disease information using the display. According to claim 1,
The processor,
Device for diagnosing a disease based on an iris image, characterized by outputting the iris image including a marker for identifying an area where a pattern associated with the disease is detected using the display. According to claim 1,
Further comprising a communication circuit configured to communicate with the server,
The processor,
Device for diagnosing a disease based on an iris image, characterized by transmitting the iris image to the server using the communication circuit. According to claim 1,
Further comprising a communication circuit configured to communicate with the server,
The processor,
Device for diagnosing a disease based on an iris image, characterized by obtaining an analysis model of the iris image from the server using the communication circuit. Obtaining an iris image of the user using a camera;
Detecting a pattern associated with a disease in the iris image by analyzing the iris image; And
And outputting disease information corresponding to the detected pattern.

Images