KR102418399B1 - Dementia an early stage diagnosis platform based on Artificial Intelligence - Google Patents

Abstract

The present invention relates to an AI-based early diagnosis platform for dementia. The present invention includes a learning unit for learning the retinal state of a dementia patient in advance through deep learning, a fundus camera for generating image data by photographing the state of the retina, and an image data collecting unit for collecting and storing image data from the fundus camera; A data analysis requesting unit for requesting analysis of the image data collected from the image data collecting unit and a retinal image photographed through the image data collecting unit are received from the data analysis request unit, receive data from the learning unit, and compare. an image analysis unit that analyzes and transmits feedback to the diagnosis unit, and a selection unit that receives the feedback from the image analysis unit and classifies dementia prevention subjects and selects future dementia predictors and types of dementia from the selection unit It is characterized in that it includes a consulting unit that receives the predicted dementia and provides consulting suitable for the type of dementia.

Description

Artificial intelligence-based early stage diagnosis platform {Dementia an early stage diagnosis platform based on Artificial Intelligence}

The present invention relates to an AI-based early diagnosis platform for dementia, and more particularly, to an AI-based early diagnosis platform for dementia through image analysis of amyloid plaques in the retina inside the eyeball.

Recently, along with the development of medical science, the proportion of the elderly population has been increasing worldwide. In particular, Korea is aging very rapidly compared to other developed countries.

With the rapid aging of the population, the incidence of dementia patients is gradually increasing, and the prevalence of dementia is also rapidly increasing. Also, dementia today is said to be the greatest health enemy that threatens the age of 100.

Dementia is a typical neuropsychiatric disease in which most of the nerve cells of the brain are damaged, resulting in disability, and it is most commonly seen in the elderly. The dementia is largely classified into Alzheimer's dementia, vascular dementia, and other dementias.

As such, dementia is progressive and results in a decline in the sense of balance, and when it gets worse, irreversible decline in daily task performance, time and space judgment, language and communication skills, and abstract thinking ability, and personality changes. It is characterized by a change in judgment and damage to judgment.

Medical institutions that treat such dementia have been diagnosing dementia in advance using medical devices such as MRI or PET as medical equipment for diagnosis of dementia. there is reality

In addition, due to the nature of dementia, it is difficult to detect early before cognitive function declines.

Therefore, in recent years, early diagnosis of dementia at low cost and the development of pain-free diagnostic equipment and diagnostic methods have been continued.

The most common form of dementia is Alzheimer's disease, a progressive neurodegenerative disorder marked by memory loss, cognitive and behavioral stability. The cause of the disease is not yet known, but as a result of analyzing the brain tissues of patients after death, amyloid plaques composed of beta-amyloid peptide (Aβ) between neurons and hyperphosphorylated tau protein filaments in neurons The accumulation of neurofibrillary tangles formed by a Lee VM et al, Neuron 1999;24:507-510; Selkoe DJ JAMA 2000;283:1615-1617].

Therefore, studies on a method for diagnosing dementia using such amyloid plaques have been actively conducted recently.

Korean Patent Publication No. 2017-0073557 Korean Patent Publication No. 2018-0002234 Korean Patent Publication No. 2020-0007244

Therefore, an object of the present invention is to be able to provide an AI-based early diagnosis platform for dementia that can be prevented in advance by analyzing amyloid plaque images in the retina found in Alzheimer's disease or dementia at an early stage.

In addition, the purpose of this is to provide an AI-based early diagnosis platform for dementia that can reduce social costs because it can delay the progression of dementia by enabling detection of dementia before cognitive function declines at an older age.

In addition, it aims to provide an AI-based early diagnosis platform for dementia that can help build a healthy social foundation by diagnosing and preparing for dementia diseases that commonly occur in the elderly in advance.

In order to solve this problem, the present invention provides a learning unit that learns the retinal state of a dementia patient in advance through deep learning, a fundus camera that generates image data by photographing the state of the retina, and collects and stores image data from the fundus camera A data analysis requesting unit requesting analysis of the image data collected from the image data collecting unit and the image data collecting unit, and receiving the image of the retina photographed through the image data collecting unit from the data analysis requesting unit, the learning unit An image analysis unit that receives and compares and analyzes data from the diagnosis unit, and an image analysis unit that delivers feedback to the diagnosis unit, receives the feedback from the image analysis unit, and classifies dementia prevention subjects to select future dementia predictors and types of dementia It is characterized in that it includes a consulting unit that receives the dementia predictor from the department and the screening unit and provides consulting suitable for the type of dementia.

In addition, the fundus camera is connected to the image processing unit and the image processing unit for processing the image transmitted from the image sensor and the image sensor for imaging the retina and an illumination module for irradiating light to the retina is formed to irradiate the retina and the processed A camera module for transmitting an image to the image data collection unit, a graphic user interface for displaying an image photographed by the image sensor, a memory for storing an image of the image photographed by the image sensor, and the retina when photographing an image, the It characterized in that it comprises a control unit for controlling the fundus camera.

In addition, the control unit is characterized in that the control so as to automatically search for a defocus with the best sharpness for the eyeball, and obtain an image of the eye for the corresponding defocus as the reference image.

In addition, in the photographing of the eyeball, near-infrared rays are irradiated to the retina through the pupil to obtain a signal (light) reflected from another position of the retina, and a tomographic analysis of the retina is performed in a non-invasive way to check the state of the retina. It is characterized by tomography (OCT).

In addition, the learning unit is characterized in that the early diagnosis of dementia by comparing and analyzing the image of the amyloid plaque in the retina of the dementia patient and the retina image data from the image data.

And, in the early diagnosis algorithm using the AI-based dementia early diagnosis platform, the step of learning the retinal image with dementia using a deep learning model through the learning unit and the retina of the person who is trying to diagnose dementia using the fundus camera In the step of photographing and deep-learning by visualizing the retinal images taken in the above step in real time, and comparing and analyzing the amyloid plaques of the retinal images with dementia obtained through the learning unit with the captured retinal images, the grade It characterized in that it comprises the step of classifying the risk of dementia in consideration of the amyloid plaque grade graded in the step of grading and the step of classifying the dementia risk person classified in the step, and consulting a specialist with the risk of dementia classified in the step.

Therefore, the present invention is effective in providing an AI-based early diagnosis platform for dementia that can diagnose and detect dementia at an early stage by implementing a deep learning model for early diagnosis of dementia by analyzing amyloid plaque images in the retina based on artificial intelligence there will be

In addition, since dementia-related health care can be continuously performed with steady interest through early diagnosis and prediction of dementia, there is an effect that can greatly reduce dementia-related national and social costs.

In addition, there is an effect that allows the patient to diagnose without pain by performing the amyloid plaque image analysis in the retina as a non-invasive method.

1 is a block diagram of a fundus camera;
2 is a block diagram of an image processing unit;
3 is a view showing the photographing of the patient's eye using a fundus camera.
4 is a block diagram of an AI-based early diagnosis platform for dementia according to the present invention.
5 is a screen picture of a mobile phone sent to a person classified as a dementia risk person.
6 is a flowchart illustrating a method for early detection of dementia using an AI-based dementia early diagnosis platform.
7 is a block diagram of a second embodiment of the present invention;
8 is a block diagram of a platform for early diagnosis of dementia according to a second embodiment of the present invention.
9 is a flowchart of a method for analyzing a retinal image for early diagnosis of dementia with artificial intelligence according to a second embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, in adding reference numerals to the components of each drawing, it should be noted that the same components are given the same reference numerals as much as possible even though they are marked on different drawings.

In addition, in the following description of the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, since the terms used in the present application are only used to describe specific embodiments, it is not intended to limit the present invention, and the singular expression means a plural expression unless the context clearly indicates otherwise. we want to leave

Before describing the present invention, the description of terms appearing in the present specification will be briefly described.

First, the meaning of the ocular fundus will be described. The fundus is the inner surface of the eye, opposite to the lens, and the retina, optic disc, macula, central concave retina, and posterior pole are included in the fundus. A fundus camera is a medical camera that takes pictures of the condition of the eyeball through the pupil. That is, it is an instrument equipped with a fundus spectacle and a camera.

And, the term deep-learning will be described. Artificial intelligence is the ability to think and act like humans by mimicking the human brain and neural networks. An example of using the deep learning is a model in which artificial intelligence draws a scene seen by a human eye as it is.

Many machine learning algorithms such as artificial neural networks have appeared regarding how to distribute data, and the deep learning is a technology used to armament or classify data using such artificial neural networks. The core of deep learning is to classify and divide data by a computer as if a human classifies objects by discovering a certain pattern among numerous data through prediction through classification.

1 is a block diagram of a fundus camera, FIG. 2 is a block diagram of an image processing unit, FIG. 3 is a diagram illustrating photographing of a patient's eye using a fundus camera, and FIG. 4 is artificial intelligence-based dementia according to the present invention It is a block diagram of an early diagnosis platform; A configuration diagram of a second embodiment of the present invention, FIG. 8 is a block diagram of a dementia early diagnosis platform according to a second embodiment of the present invention, and FIG. 9 is a retina for early diagnosis of dementia according to a second embodiment of the present invention It is a flowchart of a method of analyzing an image with artificial intelligence.

Hereinafter, first, with reference to the drawings, the fundus camera 100 for photographing the state of the fundus and retina of the eye will be described. The fundus camera 100 is an instrument for fundus examination, and fundus examination is to observe abnormal findings of the fundus caused by local eye diseases and systemic diseases through the pupil that has become mydriatic (the pupil is enlarged to 4 mm in diameter or more). is a test that

As shown in Figure 1, the fundus camera 100 is an illumination module 110 formed with an irradiator for irradiating light to the retina, an image sensor 120 that captures the retina and converts an optical signal into an electrical signal, an image It consists of a processing unit 130 , a camera module 140 , a graphic user interface 150 , a memory 160 , and a control unit 170 . Furthermore, a manipulation unit (not shown) including a capture button (not shown) for capturing an image is also formed.

The illumination module 110 for irradiating light to the retina may include an irradiator 111 and a lighting control unit 112 . The irradiation unit 111 may include a light source for generating light. For example, the light source may be a high-brightness LED.

Through the lighting control unit 112, it is possible to control whether or not to irradiate the light and the brightness of the light. The lighting control unit 112 may be connected to the image processing unit 130 .

The camera module 140 includes the image sensor 120 and the image processing unit 130 , and transmits the processed retinal image to the image data collection unit 220 .

The image sensor 120 captures a human retina or pupil, and includes a lens (not shown), and may image the image sensor 120 by enlarging the retina. The image sensor 120 includes a lens (not shown), and a CMOS sensor is preferable.

The image sensor 120 illuminates the retina through the pupil with near-infrared rays to obtain a signal (light) reflected from another location of the retina, and conducts a tomographic analysis of the retina in a non-invasive way. It is characterized in that the optical coherence tomography (OCT) to look at the state of the retina. In addition, it has a painless effect on the patient.

An anti-reflection coating layer (not shown) may be formed on the image sensor 120 to prevent interference by light reflected from the retina of the eyeball.

An image processing unit 130 for processing the image by receiving the retina image photographed from the image sensor 120 is formed.

The image processing unit 130 processes an image of a retina image, and may be connected to the image sensor 120 to receive an image photographed from the image sensor 120 .

As shown in FIG. 2 , the image processing unit 130 includes an image preprocessor 131 , an image analysis unit 132 , and an image compression unit 133 .

The image preprocessor 131 may convert the photographed retinal image into a color image. Also, pixel defects or distortion of the video image may be corrected.

In addition, the image pre-processing unit 131 corrects the image transmitted from the image sensor 120 into a more accurate image, and superimposes the image in the image analysis unit 132 to perform image analysis more easily. have.

In addition, the image compression unit 133 having a function of compressing the retinal image preferably includes a chip for video stream encoding and photo encoding.

The image of the retina of the eye processed through the image processing unit 130 as described above, that is, the image of the image photographed in the graphic user interface (150) is displayed through the display (not shown), This is to be controlled to be output to the display by the controller 170 .

The graphic user interface 150 is made of a GUI (Graphic User Interface) so that the user of the fundus camera 100 can operate, and selection of a shooting mode and image display are possible through the fundus camera 100 .

An image of an image such as a retina or a pupil displayed through the graphic user interface 150 is stored through the memory 160 .

In addition, a control unit 170 for controlling the illumination module 110 , the image sensor 120 , and the image processing unit 130 , which are components of the fundus camera 100 , is formed when capturing an image of the retina of the eyeball.

The control unit 170 controls the lighting module 110 to turn on the photographing light source for a short time when a capture button (not shown) is pressed, and at this time, through the image signal output from the image sensor 120, The image is controlled to be stored in the memory 160 . Furthermore, the control unit 170 adjusts the setting value of the image sensor 120 so that the image of the retina without distortion is captured.

FIG. 3 is a diagram illustrating photographing of an eyeball (not shown) of a person diagnosed with dementia 1 using the fundus camera 110 . As shown, the retina of the person diagnosed with dementia 1 is photographed using the fundus camera 100 .

The fundus camera 100 configured as described above receives light incident on the retina through the pupil, obtains a signal reflected from another position of the retina, examines the condition of the fundus, and can also perform autofluorescence fundus photography.

The auto-fluorescence is a non-invasive imaging technology that evaluates the retinal condition by imaging lipofuscin generated from retinal pigment epithelial cells with a confocal laser ophthalmoscope for patients with retinal disease. it means. This is used for early diagnosis of macular degeneration (a disease in which a change occurs in the macula, the center of the retina, which deteriorates vision and makes it difficult to see the desired area).

In other words, 'autofluorescence fundus imaging', which is used for the purpose of diagnosing and using retinal lesions by imaging retinal lesions in retinal diseases, is a method of irradiating 488nm laser wavelength to patients with suspected retinal diseases that are difficult to fluorescein angiography. Then, lipofuscin, which emits light from the retinal pigment epithelium, is photographed with a confocal scanning laser ophthalmoscope to diagnose retinal lesions.

Since autofluorescence fundus imaging is a non-invasive method of imaging using autofluorescence properties without injection of fluorescent substances, it is a safe technology with a low risk of causing harm to the patient, and additional lesions that cannot be confirmed by fundus imaging can be found. .

Hereinafter, the AI-based early diagnosis platform 200 for dementia according to the present invention will be described with reference to the drawings.

4 is an AI-based early diagnosis platform 200 for dementia according to the present invention. As shown, the dementia early diagnosis platform 200 includes a fundus camera 100 that generates image data by photographing the state of the retina, and a learning unit 210 that learns the retina state of a dementia patient in advance through deep learning. ), an image data collection unit 220 , a data analysis request unit 230 , an image analysis unit 240 , a selection unit 250 , and a consulting unit 260 .

The learning unit 210 trains a deep learning model in advance on the retinal image of a patient with dementia.

The learning unit 210 is to pre-learning by using the retina image of the dementia patient with dementia as a learning model.

In general, as a result of analyzing the postmortem brain tissue of dementia patients, amyloid plaques formed by the accumulation of beta-amyloid peptide between nerve cells and neurofibrillarytangles formed by hyperphosphorylated tau protein filaments in nerve cells has been reported to exist.

As a main component of such amyloid plaques, there may be amyloid beta (Aβ or Abeta). The amyloid beta refers to a peptide of 36 to 43 amino acids that is critically involved in Alzheimer's disease found in the brain of an Alzheimer's patient. The peptide is derived from amyloid precursor protein (APP), which is degraded by beta secretase and gamma secretase to produce amyloid beta (Aβ).

Therefore, when the learning unit 210 learns the image of the retina in which the amyloid plaques of dementia patients are detected as data, it learns the correlation between the input value (an image representing the brain activity pattern) and the output value (the correct answer photo) by itself. It will create an image similar to what you saw. By learning such a large amount of data in advance, the learning unit 210 can distinguish the retinal state of the dementia patient by itself.

An image data collection unit 220 for collecting and storing data of images photographed from the fundus camera 110 is formed.

The image data collection unit 220 serves to collect all image data of the retina photographed through the fundus camera 100 . In addition, in conjunction with the memory 150 of the fundus camera 100 described above, it also serves to store image data from the memory 150 .

In addition, the image data collection unit 220 has one image format among a color image, a gray image, a binary image, and the like in collecting the image data information.

The data analysis request unit 230 requests the image analysis unit 240 to analyze the image data collected from the image data collection unit 220 , and receives data from the image data analysis unit 220 to analyze the image. It serves as a medium for transmitting to the unit 240 .

The image analysis unit 240 receives the image of the retina photographed through the image data collection unit 220 from the data analysis request unit 230 and receives the image and data of the retina with dementia from the learning unit 210 . Received and compared with each other, common or overlapping, etc. are analyzed, and feedback therefor is transmitted to the selection unit 250 .

Then, the selection unit 250 receives the feedback from the image analysis unit 240 and classifies the dementia prevention target with a high probability of dementia in the future.

Therefore, it is to classify subjects who are expected to have dementia due to a high probability of having dementia in the future, and to select and diagnose the type of dementia corresponding thereto.

The consulting unit 260 receives the dementia predictor from the selection unit 250 and provides consulting suitable for the type of dementia. The consulting unit 260 classifies each dementia predictor into a dementia risk group, informs them of the amyloid plaque level by means of transmission such as mail or e-mail, and explains the type of dementia expected in the future. to send out information about In addition, it is also possible to use a mobile phone text message or transmission means such as Kakao Talk.

Hereinafter, the operation of the AI-based dementia early diagnosis platform 200 according to the present invention will be described.

First, it is to learn in advance the deep learning model of artificial intelligence through the learning unit 210 with the accumulated image of the retina with dementia.

This is to learn the analysis of amyloid plaques generated in the retina in advance due to dementia so that the learning unit 210 determines and recognizes that the deep learning model is a dementia patient by itself. Accordingly, the learning unit 210 is able to fully recognize the image of the retina affected by dementia by learning in advance.

Next, the retina is photographed using the fundus camera 100 for a person who is expected to have dementia or a person who wants to diagnose dementia early.

The image data collection unit 220 that receives the retinal image through the fundus camera 100 is the retina through the data analysis request unit 230 in order to check whether the retinal image is likely to suffer from dementia. transmit the video of

Then, the data analysis request unit 230 requests the image analysis unit 240 to analyze the received retina images.

The image analysis unit 240 analyzes and grades the images of the retina for amyloid plaques.

This is to analyze the images of the retina by analyzing the amyloid plaques, digitizing them with a certain index and scoring them, for example, 0-25 points are grade 1, 26-50 points are grade 2, 51-75 points are grade 3, 76-100 points. The point is divided into 4 grades, and from the 3rd grade onwards, images with a high risk of dementia are determined.

For reference, the determination of the score is to generate an evaluation grade score through consultation and supervision of medical experts such as doctors.

If the image analysis unit 240 analyzes or finds that the score belongs to the dementia risk group, the selection unit 250 shows the image of the retina in which future dementia is expected through the grade determined through the image analysis unit 240 . Selecting the appropriate examiner one by one.

The selection unit 260 uses the above-described scores, for example, in a grade 1 safety group with 0 to 25 points, a grade 2 semi-safe group with 26 to 50 points, a grade 3 risk group with a score of 51 to 75 points, and 4 for a grade 3 risk group with a score of 76 to 100 points. The screening is done in the same way as the grade high-risk group sign.

As shown in FIG. 5 , the consulting unit 260 provides a checkup for dementia risk check, dementia consultation check, and location-based service to the examinee who has taken the image of the retina in which dementia is expected selected through the selection unit 250 , based on the location-based service. The goal is to prepare for dementia in advance by making it possible to search for nearby hospitals, etc.

In addition, a separate notification such as mail or text message is also possible. Therefore, it is preferable that the consulting unit 260 link with each other so that an application related to the examinee's mobile phone is downloaded.

Therefore, it is possible to diagnose dementia early by performing screening using the dementia early diagnosis platform 200 according to the present invention.

Referring to FIG. 6 , it is a flowchart illustrating an early diagnosis algorithm using an AI-based early diagnosis platform for dementia. A description that overlaps with the above-described embodiment will be omitted to some extent.

First, the retinal image with dementia is pre-learned using a deep learning model through the learning unit 210 (step 1).

Using the fundus camera 100, the retina of those suspected of dementia or those who wish to diagnose dementia in advance are photographed (second step).

Visualizing the retinal images taken in the second step in real time to deep-learning (third step). That is, the retina is photographed for examinees diagnosed with dementia and prepared for visual analysis.

The retinal images of the examinees photographed through the fundus camera 100 and the amyloid plaques of the retinal images with dementia obtained through the learning unit 210 are compared and analyzed and graded (step 4).

Amyloid plaques are analyzed and graded on the photographed retinal images. The fourth step is to quantify and score the amyloid plaques with a certain index, which is to generate an evaluation grade score through consultation and supervision of medical experts such as doctors.

Therefore, it is a method of early diagnosis of dementia by comparing and analyzing the image of the amyloid plaque in the retina of the dementia patient in the learning unit 210 and the retina image data from the image data.

In consideration of the amyloid plaque grade obtained in the above step, the risk of dementia is classified (step 5), and the person at risk of dementia classified in the step is consulted with a medical professional such as a doctor (step 6). In other words, it is to prepare a plan to prevent dementia in advance by delivering data explaining the risk of dementia and other appropriate prescriptions or prevention rules to the examinees who are selected as those at risk for dementia.

Hereinafter, a second embodiment of the platform 200 for early diagnosis of dementia of the present invention will be described with accompanying drawings. In this second embodiment, the user terminal 300 is used instead of the fundus camera 100, and the user terminal 300 may be an electronic device including a smart phone, a tablet PC, a laptop, and the like.

7 shows the user terminal 300 and the dementia diagnosis system S, which are components of the dementia early diagnosis platform 200 according to the second embodiment of the present invention.

The user terminal 300 may include a camera, and the camera may include a dedicated camera for iris authentication, or a dedicated lens for iris authentication may be attached to the camera.

Accordingly, when an eye of a diagnostician is photographed using the user terminal 300 , a high-resolution screen or image can be secured for a specific region including the iris and retina.

8 is a block diagram of the dementia early diagnosis platform 200 according to the second embodiment. The user terminal 300, the image capturing unit 310 and the image preprocessing unit 320 which are components of the dementia diagnosis system S. , is provided through the neural network learning unit 330 and the early dementia diagnosis unit 340 .

The image capturing unit 310 is a part that captures an eye image through a user terminal 300 such as a smart phone, and may acquire a captured image by using an iris authentication exclusive lens that can be attached to the front of the smart phone. The photographing of the iris is for identity authentication.

The image preprocessor 320 may extract only a retinal region necessary for early diagnosis of dementia from the eye image captured by the image capturing unit 310 . First, using the JPEG2000 library, only a region-of-interest (RoI) can be extracted from an eye image. Here, the region of interest refers to the minimum region required to extract the retina.

The neural network learning unit 330 includes a DB (reference numerals omitted) for storing image data of the iris and retina captured by the image capturing unit 310 and a neural network model learning from the stored data.

In the information to be stored in the DB, bio information corresponding to personal information, unique identification information (resident registration number, etc.), and password must be encrypted and stored.

Since the image data of the iris and retina corresponds to bio-information, it must be encrypted and then stored. When storing data, the DB encryption method will be preferable for the encryption method. In the case of encryption, an encryption key must be set, and the encryption key uses a code value extracted from iris and retina image data, is encrypted using the corresponding encryption key, and then stored in the DB.

Next, using the stored retinal image data, an AI neural network for early diagnosis of dementia symptoms made in advance can be learned.

Early diagnosis of dementia through retinal images The artificial intelligence neural network should be trained to be classified/detected/segmented for dementia. Therefore, the present artificial intelligence neural network for early diagnosis of dementia can pre-learning the part of the neural network that classifies the prognostic symptoms of dementia for more accurate classification during learning. Since the artificial neural network to be used for the classification part occupies a large part of the overall neural network, the part is made as lightweight as possible so that it can be performed smoothly on mobile devices.

Hereinafter, the learning method of the artificial intelligence neural network will be described.

Pre-training is performed on the classification part, which is to perform pre-training by analyzing the amyloid-plaque in the retina of Alzheimer's patients with dementia. As the learning method, the brain tissue of patients after death was analyzed, and beta-amyloid peptides were accumulated between nerve cells to form amyloid plaques and neurofibrillary tangles formed by hyperphosphorylated tau protein filaments in nerve cells. ) to pre-learning about the existence of After performing the preceding learning, with the extracted specific map, it is possible to learn to analyze future dementia.

And, the dementia early diagnosis unit 340 grades the amyloid plaques of each dementia diagnosis target based on the photographed retinal images using the retina images that have been pre-processed with the artificial intelligence neural network learned by the neural network learning unit 330 . If the number exceeds a certain level, it is identified as a risk group for future dementia and the subject is notified.

9 is a flowchart of a method of analyzing a retinal image for early diagnosis of dementia with artificial intelligence according to a second embodiment of the present invention. Descriptions overlapping with those described above will be omitted to some extent.

A captured image obtained by photographing the user's retina with the user terminal 300 may be obtained (Step 1), and the captured image may be pre-processed to extract a Region of Interest from the user terminal 300 (Step 2). .

When extracting the region of interest, for example, a method of extracting only the region of interest is used because the region that is not extracted as the region of interest in the captured image is unnecessary. Since data is not exhausted, more accurate diagnosis can be made.

In addition, the resolution of the extracted ROI is extended by the user terminal 300 (Step 3). In Step 3, since the output images are each divided into a patch unit, which is one of the image recognition units, it can be converted into a single high-resolution image through a merging process. That is, the input low-resolution retinal image is divided into a specific patch size, and the resolution is further improved by inputting the divided image and extracting the reconstructed and merged output image.

The region of interest to be transmitted from the user terminal 300 to the dementia diagnosis system S is compressed using a library for reducing the amount of data (Step 4). In Step 4, compression is performed using a compressed file library (CFL), and only the region of interest is extracted.

User authentication is performed through the iris image of the user terminal 300 (step 5). Accordingly, the identity of the owner of the user terminal 300 can be clearly identified through the image of the iris.

The user terminal 300 encrypts the data of the iris image and the retina image to be stored in the dementia diagnosis system S (Step 6).

Based on the data stored in the dementia diagnosis system (S), an artificial neural network to perform early diagnosis of dementia is learned (Step 7), and the artificial neural network learned in the user terminal 300 as described above is used to predict future dementia. Predict the probability of occurrence and notify each subject of this (Step 8).

In addition to the above-described embodiment, the present invention accumulates big data indicating the possibility of dementia according to the location and shape of the dementia-related area, and based on the big data, the degree of the possibility of dementia according to the location and shape of the disease area is learned and determined, and according to the degree of possibility of future diagnosis of dementia, a subject with a high probability of future dementia, including amyloid plaque level, may be notified in real time through the user terminal 300 .

Those skilled in the art to which the present invention pertains as described above will understand that the present invention may be implemented in other specific forms without changing the technical spirit or essential features of the present invention. Therefore, it should be understood that the above-described embodiments are illustrative and not restrictive.

The scope of the present invention is indicated by the appended claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

1: Dementia diagnosis target 100: Fundus camera
110: lighting module 111: irradiation unit
112: lighting control unit 120: image sensor
130: image processing unit 140: camera module
150: graphical user interface (GUI) 160: memory
170: control unit 200: dementia early diagnosis platform
210: learning unit 220: image data collection unit
230: data analysis request unit 240: image analysis unit
250: selection unit 260: consulting unit
300: user terminal 310: video recording unit
320: image preprocessing unit 330: neural network learning unit
340: Dementia early diagnosis unit S: Dementia diagnosis system

Claims (6)

In the AI-based early diagnosis platform for dementia,
a learning unit that learns in advance the retinal state of dementia patients through deep learning;
A fundus camera for generating image data by photographing the state of the retina;
an image data collection unit for collecting and storing image data from the fundus camera;
a data analysis request unit for requesting analysis of the image data collected from the image data collection unit;
an image analysis unit receiving the retinal image photographed through the image data collection unit from the data analysis request unit, receiving data from the learning unit, comparing and analyzing, and delivering feedback to the selection unit;
a selection unit that receives the feedback from the image analysis unit and classifies dementia prevention subjects to select future dementia patients and types of dementia;
Includes a consulting unit that receives the dementia predictor from the selection unit and provides consulting suitable for the type of dementia,
The fundus camera is
an illumination module formed with an irradiator for irradiating light to the retina;
an image sensor that captures the retina and converts an optical signal into an electrical signal;
an image processing unit for processing the retinal image transmitted from the image sensor;
a camera module comprising the image sensor and the image processing unit to transmit the processed image of the retina to the image data collection unit;
a graphical user interface for displaying the processed video image;
a memory for storing the video image displayed by the graphic user interface;
When taking the image of the retina, including a control unit to store the retina image in the memory through the image signal output from the image sensor,
The fundus camera receives a signal that is incident on the retina through the pupil to obtain a signal reflected from another location of the retina, and examines the condition of the fundus,
The control unit adjusts the setting value of the image sensor so that the fundus camera takes an image of the retina without distortion,
The fundus camera injects light into the retina through the pupil and obtains a signal reflected from another location of the retina to examine the condition of the fundus by autofluorescence fundus photography,
The learning unit learns the image of the retina in which the amyloid plaque of the dementia patient is detected as data, and learns the correlation between the input value, which is an image representing the brain activity pattern, and the output value, which is the correct answer photo, to generate an image similar to that seen by a person, ,
Early diagnosis of dementia The artificial intelligence neural network classifies, detects, and divides dementia, classifies the prognostic symptoms of dementia,
The prior learning for the classification part analyzes amyloid plaques in the retina of Alzheimer's patients with dementia, and the prior learning analyzes the postmortem brain tissue of the patient, and beta-amyloid peptides are accumulated between nerve cells to form amyloid plaques and Pre-learning about the existence of a nerve fiber mass formed by hyperphosphorylated tau protein filaments in nerve cells,
Using the pre-learned artificial intelligence neural network, the amyloid plaque of a subject diagnosed with dementia is graded based on the photographed retinal image using the retina image, and if a certain number is exceeded, it is identified as a future dementia risk group and notified to the subject An artificial intelligence-based early diagnosis platform for dementia, characterized in that delete delete delete delete delete

Images