KR102550465B1 - Artificial intelligence-based virtual patient management system - Google Patents

Abstract

An artificial intelligence-based virtual patient management system according to the present invention includes an image library storing lesion images to which diagnostic labels are assigned; An AI learning module for classifying the stored lesion images by disease based on artificial intelligence and constructing stored big data; a virtual patient matching module including a condition setting unit for setting virtual conditions for basic information and lesion information of a virtual patient, and an image extraction unit for extracting a lesion image by searching the big data based on the virtual conditions; and an output module displaying the extracted lesion image.
According to the artificial intelligence-based virtual patient management system of the present invention, based on big data, lesion images matching the virtual conditions of virtual patients can be extracted and provided, so that hospitals can perform staff training using virtual lesion cases It can be done, and it is effective to provide various education to medical students by using virtual lesion cases.

Description

Artificial intelligence-based virtual patient management system}

The present invention relates to an artificial intelligence-based virtual patient management system, and more specifically, is novel and progressive, capable of extracting and providing a lesion image for a virtual patient whose conditions are virtually set based on a lesion image stored in big data. It is about a virtual patient management system.

Radiology, which began with the discovery of X-Ray by Wilhelm Roentgen, is currently playing a very important role in medical diagnosis and treatment. The goal of radiology is to find out the maximum diagnostic information with the minimum amount of radiation, such as X-Ray and ultrasound. CT and MRI are used. Therefore, the need for a radiology method capable of obtaining accurate data while minimizing radiation exposure is attracting attention.

On the other hand, with the recent explosive growth of the smart medical industry worldwide, health care services and information and communication technology are combined to prevent, diagnose, predict and treat individual diseases by using artificial intelligence (AI) based on medical big data. The paradigm of medical care is changing as 'data-centered' digital health is attracting attention.

As such, in the smart medical market, which is the core of the 4th industry, 'virtual patients' have recently been used in various ways. Virtual patient is a state-of-the-art precision medical prediction technology that promotes disease diagnosis, prediction, and treatment direction through artificial intelligence simulation by creating a virtual patient before human treatment. Using machine learning techniques, biometric information and medical images can be combined to create virtual patient data. If enough big data is formed on virtual patients, it is expected to be of great help to patients with various chronic diseases and emergency patients, and the quality of data on these virtual patients is rapidly developing.

As a prior art for this, Korean Patent Publication No. 10-2022-0111212 discloses a 'system and method for generating virtual patient information using machine learning'.

The prior art relates to a virtual patient information system and method using machine learning. Based on the information collected by the information collection unit that collects open and available disease-related information or patient biometric information, a group corresponding to each disease is formed using machine learning, and a group corresponding to each disease is formed. An information generation unit that generates information (data) of patients belonging to each group, a virtual patient group generation unit that creates a new virtual patient group suffering from complex diseases by using machine learning based on patient information (data) of each group, and a virtual patient It is a virtual patient information generation system using machine learning that uses information generators to check conditions and control operations to transmit commands.

In this way, by combining a patient's biometric information and medical images using a machine learning AI model to generate and provide information on a virtual patient at a level similar to that of a real patient, it can be used in the treatment and treatment of general patients as well as emergency patients in the medical field. It is disclosed that there is an advantage that it can help.

Further from this, a new and innovative technology that can predict and provide a lesion image for a virtual patient by utilizing big data storing virtual patient information and actual lesion images to which diagnostic labels have been assigned based on artificial intelligence for virtual conditions. There is a need to provide an advanced virtual patient management system.

Korean Patent Publication No. 10-2022-0111212

The main object of the present invention is to construct big data on lesion images based on artificial intelligence, and based on this, to extract and provide lesion images for patients whose conditions are virtually set.

Another object of the present invention is to apply the extraction and provision of lesion images to prediction of the prognosis of real patients beyond virtual patients.

Another object of the present invention is to further reflect blood test result data in predicting the actual patient's prognosis.

In order to achieve the above object, an artificial intelligence-based virtual patient management system according to the present invention includes an image library storing lesion images to which diagnostic labels are assigned; An AI learning module for classifying the stored lesion images by disease based on artificial intelligence and constructing stored big data; a virtual patient matching module including a condition setting unit for setting virtual conditions for basic information and lesion information of a virtual patient, and an image extraction unit for extracting a lesion image by searching the big data based on the virtual conditions; and an output module displaying the extracted lesion image.

Furthermore, the system includes a lesion image input unit that receives a lesion image to which a diagnosis label for the patient is assigned from a management terminal, and a condition input unit that receives basic information of the patient and current conditions for the lesion information from the management terminal. a patient information input module to; And, an AI prediction module for extracting a lesion image for a future time point by searching the big data based on the diagnosis label assigned to the lesion image of the patient and the current condition, wherein the output module includes the extracted It is characterized in that it includes a function of displaying a lesion image for a future time point.

In addition, the system includes a test library storing result data for each item of a blood test to which a diagnostic label is assigned, wherein the AI learning module classifies and stores the stored lesion images and result data by disease Big data is constructed, and the patient information input module further includes a result data input unit for receiving result data of the patient from a management terminal, and the AI prediction module includes a diagnosis label assigned to the lesion image of the patient and the patient information input module. It is characterized in that a lesion image for a future time point is extracted by searching the big data based on the current condition and the result data.

According to the artificial intelligence-based virtual patient management system of the present invention,

1) Based on big data, by extracting and providing lesion images that match the virtual conditions of virtual patients, hospitals can conduct staff training using virtual lesion cases, and medical students can also use virtual lesion cases. to provide a variety of education,

2) Based on the patient's lesion image, it is possible to predict and determine the prognosis of future lesions through actual cases, and at this time, a search is performed based on big data to track the lesion image that has the highest match with the patient. It makes it possible to observe the prognosis,

3) In predicting the lesion prognosis for the patient's future time point, it is possible to further reflect not only the lesion image but also the blood test result data, thereby increasing the prediction accuracy.

1 is a schematic configuration diagram of a virtual patient management system of the present invention.
Figure 2 is a block diagram showing the detailed configuration of the virtual patient management system of the present invention.
3 is a conceptual diagram showing an example of a lesion image according to the present invention;

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The accompanying drawings are not drawn to scale, and like reference numbers in each drawing indicate like elements.

1 is a schematic configuration diagram of a virtual patient management system of the present invention.

Referring to FIG. 1 , the artificial intelligence-based virtual patient management system of the present invention is characterized by including a management terminal 20 and a main server 10.

The management terminal 20 is a terminal that performs functions such as inputting lesion images or setting virtual conditions. Such management terminal 20 may be a PC installed in a hospital in general, or a hospital formed by a group of PCs installed in a hospital. Could be a server. Furthermore, it goes without saying that the management terminal 20 is provided with a display, so that the lesion image extracted through a configuration to be described later can be output through the display installed in the management terminal 20 .

The main server 10 is a subject that provides learning and extraction functions for lesion images based on artificial intelligence in the virtual patient management system of the present invention, and the subject of the present invention is the main server 10. can

The main server 10 has a central processing unit (CPU) and a hardware-based memory and storage means such as a hard disk, and a program that can be executed in the central processing unit, that is, software is installed and the software can be executed. A series of detailed configurations of the software will be described later as units such as 'modules', 'parts', and 'parts'.

Configurations such as 'modules' or 'units' or 'interfaces' or 'parts' are software or hardware such as FPGAs or ASICs executed via CPU and memory in a state installed and stored in the storage means of the main server 10 means the work composition of

At this time, the configuration of 'module' or 'unit' or 'interface' is not limited to hardware, and may be configured to be in an addressable storage medium or configured to reproduce one or more processors.

As an example, 'module' or 'part' or 'interface' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, and attributes. fields, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays and variables.

Functions provided by these 'modules' or 'units' or 'interfaces' may be combined into a smaller number of components and 'units' or 'modules', or may be combined into additional components and 'units' or 'modules'. can be further separated.

In addition, the main server 10 means all types of hardware devices including at least one processor, and may be understood as encompassing software configurations operating in the corresponding hardware devices according to embodiments.

For example, a computing device as an example of a server may be understood as including all of smart phones, tablet PCs, desktops, laptops, and user clients and applications running on each device, but is not limited thereto.

Therefore, the system of the present invention will be described in more detail based on each component constituting the artificial intelligence-based virtual patient management system.

2 is a block diagram showing the detailed configuration of the virtual patient management system of the present invention, and FIG. 3 is a conceptual diagram showing an example of a lesion image of the present invention.

2 and 3, the artificial intelligence-based virtual patient management system of the present invention includes an image library 100, an AI learning module 200, a virtual patient matching module 300, and an output module 400. It is characterized by doing.

The image library 100 stores lesion images to which diagnostic labels have been assigned. Here, the assigned diagnostic labels refer to information about the patient who provided the corresponding lesion image, for example, the age of the patient who provided the corresponding lesion image, In addition to gender, height, weight, etc., information on diseases and lesions included in the corresponding lesion image may be included.

Therefore, the name of the disease, the location of the lesion, the size of the lesion, the progress, or other references may be included, and such a diagnosis label is basically a diagnosis or biopsy report of the patient who provided the image of the lesion, other medical tests or In the state of including the diagnosis record as lesion information, it can be said that it includes basic information (age, sex, height, weight, etc.) of the patient and further includes the date of diagnosis.

The AI learning module 200 performs a function of classifying stored lesion images by disease based on artificial intelligence and constructing stored big data. In this case, based on the stored lesion image and the label assigned to the lesion image, artificial intelligence, preferably a machine learning model, is used to classify the lesion image according to each disease and provide a function of constructing stored big data.

At this time, in classifying the lesion image by disease according to the lesion image and the diagnostic label assigned to the lesion image, the analysis of the diagnostic label must be thoroughly performed. In this case, classifying the lesion image by disease based on the diagnostic label Of course, it is more desirable to make a more detailed classification according to the size or location of the lesion, the degree of progression of the disease, and the age and gender of the patient.

Therefore, classifying lesion images by disease by analyzing the information included in the diagnostic label can be said to form lesion images into groups corresponding to each disease. At this time, each information included in the diagnostic label correlates with each other. If there is a correlation, it can be grouped into the same group, and if there is no correlation, it can be grouped into another group.

Furthermore, in order to increase learning efficiency in learning and building big data through the AI learning module 200, machine learning including ensemble techniques are used to learn lesion images and diagnostic labels included in lesion images, and based on this, It is possible to build big data by classifying lesion images by disease.

Here, the ensemble technique refers to combining multiple models to generate multiple classifiers in Ensemble Learning, creating multiple different classifiers for a given training data set, and determining the results of these classifiers using a voting method or a weighted vote. It can be said that it is to improve the learning performance as it performs classification by combining in a different way.

The virtual patient matching module 300 sets virtual conditions of the virtual patient, searches big data built through the above-described AI learning module 200 based on the set virtual conditions, and generates a lesion image that meets the virtual conditions of the virtual patient. It provides a function of extracting, and includes a condition setting unit 310 and an image extraction unit 320 for this purpose.

The condition setting unit 310 sets virtual conditions for the basic information and lesion information of the virtual patient, and here, the virtual patient means a patient with conditions set virtually, not an actual patient. Here, the basic information may include gender, age, height, and weight, and the lesion information may include the name of the disease, the location of the lesion, and the size of the lesion.

For example, a virtual patient is a virtual condition, and the basic information can be set as female, age 34, height 163 cm, weight 52 kg, lesion information is advanced gastric cancer, location is lower 1/3 (center of mid vulva, posterior wall), mesodifferentiated type It can be set to tubular adenocarcinoma, 2.8cm*2.6cm long shape, muscle layer (pT2) invasion, lymphovascular invasion, and pT2/pN1 stage.

Therefore, when a virtual condition including basic information and lesion information is set, it can be said that a virtual patient with basic information having a disease corresponding to the corresponding lesion information is created.

The image extraction unit 320 performs a function of extracting a lesion image matching the virtual condition of a virtual patient by searching big data based on set virtual conditions. In other words, it provides a function to search big data based on the virtual condition and extract the lesion image with the diagnostic label that has the highest match rate with the virtual condition.

Accordingly, among the lesion images stored in the image library 100, a lesion image having a diagnosis label having the highest concordance rate with the virtual condition may be extracted, and a lesion image having a condition virtually similar to that of the virtual patient is extracted through this.

The output module 400 provides a function of displaying the extracted lesion image. As described above, the output module 400 can output the extracted lesion image through the display provided in the management terminal 20, or a separate image installed in the hospital. It is also possible to output the extracted lesion image through the display.

According to the virtual patient management system of the present invention, by extracting and providing a lesion image matching the virtual patient's virtual condition based on big data, the hospital can perform staff training using virtual lesion cases. Of course, it is possible to provide various education to medical students by using virtual cases of lesions.

In addition, when extracting a lesion image, a reference time point of a virtual condition can be set, and how the lesion included in the lesion image can change until a specific time point (future time point) after the reference point can be predicted. To this end, the present invention The virtual patient management system of may further include additional components.

To this end, the virtual condition first set by the condition setting unit 310 may be a virtual condition at a reference point in time for the basic information and lesion information of the virtual patient. Here, the reference point of view may be a current point of view or a point of view virtually set through the management terminal 20 . Most simply, the reference point in time may be the current point in time.

In this case, the image extraction unit 320 can firstly extract a lesion image based on a corresponding reference time point, predict a future time point of the reference time point, and secondly extract a lesion image for a future prognosis of a virtual patient. The image extraction unit 320 may include a current image extraction part 321 and a future image extraction part 322 .

The current image extraction part 321 performs a function of first extracting the lesion image by searching the big data based on the virtual condition of the reference point in time, wherein the lesion image is first extracted through the image extraction part Since it is the same as the basic configuration and function of the above-described image extraction unit 320, a detailed description thereof will be omitted.

The future image extraction part 322 searches the big data based on the diagnostic label of the primary extracted lesion image, and performs a function of secondary extraction of a lesion image for a specific future point in time after the reference point in time. , The diagnostic label of the primary extracted lesion image is identified, and based on the date the lesion image was taken, the lesion information included in the lesion image, and the information on the patient who provided the lesion image, a specific time point after the reference point, the future Performs the function of secondary extraction of the lesion image for the viewpoint.

This is based on the fact that lesion images are taken repeatedly to monitor the lesion periodically, rather than temporarily taking a lesion image only at a specific point in time when a lesion occurs in general. Referring to this, it can be said that the patient who provided the corresponding lesion image performs secondary extraction of the lesion image repeatedly taken at a future point in time. Here, the secondaryly extracted lesion image may be a lesion image after 3 months, a lesion image after 6 months, a lesion image after 12 months, and the like, compared to the reference time point.

Therefore, according to the current image extraction part 321 and the future image extraction part 322, the first extracted lesion image based on the virtual condition and the first extracted lesion image based on the diagnosis label of the corresponding lesion image Secondary extraction of lesion images at a future time point.

In other words, the primary extracted lesion image based on the virtual conditions of the reference point in time and the secondary extracted lesion image of a future point in time after the reference point are extracted together.

Through this, it is possible to predict how the lesion included in the primary extracted lesion image will change in the future through the secondary extracted lesion image. In this case, the output module 400 outputs the primary extracted lesion image and the secondary extracted lesion image. By displaying the extracted lesion image, it is possible to predict how the lesion included in the primary extracted lesion image may change in the future.

Furthermore, the virtual patient management system of the present invention goes beyond extracting a lesion image that meets the virtual condition of a virtual patient from big data to predict and provide a lesion image of the patient at a future point in time based on a lesion image of an actual patient. It is possible, for this purpose, the system of the present invention may further include a patient information input module 500 and an AI prediction module 600.

The patient information input module 500 performs a function of receiving a lesion image of the patient and the current condition of the patient from the management terminal 20, and includes a lesion image input unit 510 and a condition input unit 520 for this purpose. .

The lesion image input unit 510 receives a lesion image for a specific patient from the management terminal 20, and the corresponding lesion image must also be assigned a diagnostic label.

The condition input unit 520 performs a function of receiving input of the patient's basic information and current conditions for the lesion information from the management terminal 20, and receives detailed lesion information and basic information of the patient as well as the diagnosis label. that made it possible If all items included in the lesion information are included in the diagnosis label, the lesion information may be replaced with the contents included in the diagnosis label, of course.

That is, among the contents included in the diagnosis label, contents corresponding to the lesion information may be input as lesion information, and if the basic information of the patient is also included in the diagnosis label, contents corresponding to the basic information among the contents included in the diagnosis label It can be input as this basic information.

The AI prediction module 600 performs a function of extracting a lesion image for a specific time point in the future by searching the big data based on the diagnosis label assigned to the patient's lesion image and the set current condition.

Here, extracting a lesion image for a specific time point in the future means, as in the above-described future image extraction part 322, by searching the big data based on the diagnosis label assigned to the patient's lesion image and the current condition, It is to extract a lesion image for a future time point, which is a later time point.

More specifically, after primary extraction of a lesion image having a diagnostic label most similar to the patient's diagnostic label and current condition, and then referring to the diagnostic label of the primary extracted lesion image, the patient who provides the primary extracted lesion image is a future viewpoint. It is to extract the image of the lesion photographed on .

Here, the future point of view may refer to a point of time after the point of time when the primary extracted lesion image was taken. For example, it may be 3 months, 6 months, or 12 months after the primary extracted lesion image is taken.

In this case, if there are multiple primary extracted lesion images, there may be multiple lesion images for future time points. Therefore, a lesion image (primarily extracted lesion image) having a diagnostic label similar to that of the patient is first extracted, and further Based on the diagnosis label of the primary extracted lesion image, a lesion image taken at a future time point, i.e., a later time point from the time of the imaging, can be extracted.

Therefore, the lesion image at the future time point may be a lesion image after 3 months, a lesion image after 6 months, and a lesion image after 12 months compared to the current time point, and the number of lesion images extracted first may be equal to or more than that of course. .

Accordingly, when a lesion image for a future viewpoint is output in this way, the lesion image for a future viewpoint extracted through the output module 400 can be displayed.

Therefore, as a result, when the patient's lesion image is input from the management terminal 20 and the current condition is entered, the diagnosis label included in the lesion image and the current condition are analyzed to obtain the most consistent match with the patient's lesion image among the lesion images stored in the big data. High, similar lesion images are primarily extracted, and based on the diagnostic label of the primary extracted lesion image, the lesion image provided by the patient who provided the primary extracted lesion image at a later time point (future time point) is searched for and secondary extraction is performed. will be provided.

In other words, if the patient input from the management terminal 20 is a patient to be managed, the lesion image most similar to the patient to be managed is extracted from among the lesion images stored in the big data based on the lesion image of the patient to be managed, and the extracted lesion It is possible to provide a lesion image of a patient corresponding to the image at a future point of view, so that the future lesion image of the patient to be managed can be predicted.

Through this, it is possible to predict and provide the prognosis of the patient's lesion image at the future point of view, and based on the patient's lesion image, the prognosis of the future lesion can be predicted and judged through actual cases, as well as big data at this time. By performing a search based on , it is possible to observe the follow-up prognosis for the lesion image that has the highest concordance with the patient.

Furthermore, at this time, in extracting the lesion image of the future viewpoint, at least one lesion image of the future viewpoint may be extracted. Several lesion images can be produced, such as a lesion image after 6 months and a lesion image after 12 months. Therefore, it can be said that a plurality of lesion images of a future viewpoint are generally extracted.

In this case, at a future point in time, which is a later point in time based on a specific point in time, the lesion may be worse or alleviated than at a specific point in time. Therefore, the image of the lesion at a future point of view may vary greatly depending on the deterioration/remission of the lesion.

In order to reflect this and provide a lesion image at a future viewpoint, the AI prediction module 600 of the present invention may include a basic extraction unit 610, an aggravated image designation unit 620, and a relief image designation unit 630. there is.

The basic extractor 610 can also be referred to as the basic component of the AI prediction module 600, and searches the big data based on the diagnosis label given to the patient's lesion image and the current condition to search for a plurality of lesion images for a future viewpoint. Provides a function to extract . Since the configuration and functions of the basic extractor 610 can be referred to the AI prediction module 600 described above, a detailed description thereof will be omitted.

The deteriorating image designation unit 620 performs a function of designating a lesion image having a diagnosis label worse than the patient's current condition among the extracted lesion images as the deteriorating image. Deterioration here refers to all cases in which the lesion can be judged to have deteriorated based on the diagnosis of the hospital, such as an increase in the size of the lesion, increased infiltration of the lesion, metastasis of the lesion, or additional formation of a lesion in a different location from the original location. includes This can be determined based on the information included in the diagnostic label included in the lesion image.

In other words, it can be said that the diagnosis label of the extracted lesion image is compared with the patient's current condition, and only the lesion image in a state worse than the patient's current condition is re-extracted and designated as an aggravated image.

The palliative image designation unit 630 performs a function of designating, as a palliative image, a lesion image having a diagnosis label more palliative than that of the patient's current condition among a plurality of extracted lesion images. All cases that can be judged to be alleviated by hospital diagnosis, such as small size, shallow invasion, or disappearance of the lesion, are included. This can also be determined based on the information included in the diagnostic label included in the lesion image.

In other words, it can be said that the diagnosis label of the extracted lesion image is compared with the patient's current condition, and only the lesion image in a more relaxed state than the patient's current condition is re-extracted and designated as the alleviated image.

Therefore, when the aggravated image and the ameliorated image are designated in this way, the aggravated image and the ameliorated image can be displayed through the output module 400, so that the patient's lesion image is expected to deteriorate in the future lesion image (aggravated image), Of course, it can be used for staff training or patient treatment based on future lesion images (relief images) in which the image of the patient's lesion is expected to be alleviated.

As another configuration, the patient's blood test data can be further used to improve prediction accuracy in predicting and extracting a lesion image at a future point in time by searching big data based on the diagnosis label included in the patient's lesion image and the current condition. .

First, the system of the present invention includes the test library 700 and can store result data for each item of a blood test. At this time, there are no restrictions on the items that can be included in the blood test, but in general, the items that can be included in the blood test are white blood cell count, red blood cell count, hemoglobin count, hematocrit, platelets, mean red blood cell volume, blood sugar level, serum creatinine level, lipid level Even if there are items not mentioned, if they can be identified through a blood test, they are not limited.

Therefore, result data for each item obtained by performing a blood test is stored in the test library 700, and preferably, a diagnostic label may be assigned to the result data of the corresponding blood test. In this case, the diagnosis label may refer to the above description, and the name of a disease suspected through the corresponding blood test result may be assigned to the result data as a diagnosis label.

Furthermore, when blood test result data is stored in the test library 700 in this way, the AI learning module 200 constructs big data by classifying and storing the stored lesion images and result data by disease when constructing big data. can As described above, it can be said that lesion images and result data are classified and stored by disease based on the diagnosis label.

More specifically, at this time, in classifying the lesion image and result data by disease, the analysis of the diagnosis label must be thoroughly performed. Of course, it is also possible to analyze the causal relationship between the lesion image and the result data by analyzing the lesion image classified as a disease and the result data.

Therefore, classifying the lesion image and result data by disease by analyzing the information included in the diagnostic label can be said to form the lesion image and result data into a group corresponding to each disease. It is possible to determine whether information is correlated with each other, group it into the same group if there is a correlation, and group it into another group if there is no correlation.

If big data is constructed in this way, it is possible to further refine the above-described configuration of predicting a future lesion image based on a patient's lesion image input through result data. To this end, the patient information input module 500 uses the result data The patient's blood test result data may be received from the management terminal 20 including the input unit 530, and the AI prediction module 600 may use the diagnostic label assigned to the patient's lesion image and the current condition and the patient's lesion image. Based on the resulting data, it is possible to search the big data and extract a lesion image for a future time point.

Here, extracting a lesion image for a specific time point in the future means, as in the future image extraction part 322 described above, based on the diagnosis label assigned to the patient's lesion image, the current condition, and the patient's result data, the big data is searched for, and a lesion image for a future time point, which is later than the current time point, is extracted.

More specifically, the patient's diagnostic label on big data. After primary extraction of the lesion image of the group with the highest similarity with the current condition and result data, a lesion image taken at a future time point by the patient who provided the primary extracted lesion image is extracted by referring to the diagnosis label of the primary extracted lesion image. is to pay

Here, the future point of view may refer to a point of time after the point of time when the primary extracted lesion image was taken. For example, it may be 3 months, 6 months, or 12 months after the primary extracted lesion image is taken.

In this case, if there are multiple primary extracted lesion images, there may be multiple lesion images for future time points. Therefore, a lesion image (primarily extracted lesion image) having a diagnostic label similar to that of the patient is first extracted, and further Based on the diagnosis label of the primary extracted lesion image, a lesion image taken at a future time point, i.e., a later time point from the time of the imaging, can be extracted.

Therefore, the lesion image at the future time point may be a lesion image after 3 months, a lesion image after 6 months, and a lesion image after 12 months compared to the current time point, and the number of lesion images extracted first may be equal to or more than that of course. .

Accordingly, when a lesion image for a future viewpoint is output in this way, the lesion image for a future viewpoint extracted through the output module 400 can be displayed.

According to the detailed configuration of the AI prediction module 600, it is possible to further reflect not only the lesion image but also the result data of the blood test in predicting the lesion prognosis for the future time point of the patient, thereby increasing the prediction accuracy.

In addition, in predicting a patient's prognosis, it is possible to go beyond reflecting the blood test result data to reflect the patient's lifestyle, correct the result data, and predict the future prognosis accordingly. To this end, patient information is input first. The module 500 may further include a life pattern input unit 540 .

The life pattern input unit 540 performs a function of receiving life pattern data about drug compliance, sleep habits, drinking habits, eating habits, exercise amount, and stress of the patient from the management terminal 20 .

At this time, drug compliance may include the name, dosage, and dosage cycle of the drug to be taken, sleep habits may include average bedtime and average sleep time and wake up time, and drinking habits may include drinking cycle and amount of alcohol. Eating habits include meal time, meal amount, and type of food consumed, exercise amount may include exercise cycle, exercise amount, and type of exercise performed, and stress is collected through consultation with patients. It may include a numerical value of the level of stress felt by the patient or the current level of stress of the patient as determined by the doctor through consultation with the patient.

Furthermore, the lifestyle pattern data may further include various additional factors that may be included in the lifestyle pattern in addition to drug compliance, sleep habits, drinking habits, eating habits, exercise amount, and stress.

Such life pattern data is basically input through the management terminal 20, but in order for the lifestyle pattern data to be input through the management terminal 20, a doctor consults or treats the patient and asks questions about the lifestyle pattern data. After the patient is provided with a question and answer sheet, the patient answers the question and answer sheet, and then the completed questionnaire or question and answer sheet is input as life pattern data.

Therefore, if the patient's life pattern is grasped according to the life pattern data, the AI prediction module 600 can correct the resulting data based on the life pattern data through the data correction unit 640. This can also be said to predict future result data through current life pattern data. For example, if exercise is low, lipid levels, cholesterol levels, and blood sugar levels can be increased, and if the drinking cycle is short and the amount of alcohol is high, liver levels , It can increase cholesterol levels, and it is possible to increase lipid levels, cholesterol levels, and blood sugar levels even when exercise is low, and it is possible to increase blood sugar levels when eating habits are irregular and undesirable nutrients continue.

Through this, it is possible to perform prediction for the future by correcting the input patient result data based on the change in result data expected based on the life pattern data. In this case, the AI prediction module 600 is applied to the patient's lesion image. It is possible to extract a lesion image for a specific time point in the future by searching the big data based on the assigned diagnostic label and the corrected result data.

In other words, in searching for a specific lesion image through the above-described AI prediction module 600 and extracting a lesion image at a future time point of a patient who inputs the corresponding lesion image, corrected result data rather than the input result data is used, so that the patient's It is to further reflect the patient's lifestyle in predicting future lesion images.

According to this configuration, future prognosis is predicted by reflecting the patient's lifestyle as well as the patient's lesion image and blood test result data, thereby predicting the lesion image at a future point in time based on lifestyle, one of the important variables in the prognosis of lesions. It provides the effect of increasing the accuracy for .
As described above, the configuration and operation of the artificial intelligence-based virtual patient management system according to the present invention are expressed in the above description and drawings, but this is only an example and the spirit of the present invention is not limited to the above description and drawings. And, of course, various changes and modifications are possible within a range that does not deviate from the technical spirit of the present invention.

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

10: main server 20: management terminal
100: image library 200: AI learning module
300: virtual patient matching module 310: condition setting unit
320: image extraction unit 321: current image extraction part
322: future image extraction part 400: output module
500: patient information input module 510: lesion image input unit
520: condition input unit 530: result data input unit
540: life pattern input unit 600: AI prediction module
610: basic extraction unit 620: deterioration image designation unit
630: alleviation image designation unit 640: data correction unit
700: inspection library

Claims (8)

As an AI-based virtual patient management system,
an image library storing lesion images with diagnostic labels;
An AI learning module for classifying the stored lesion images by disease based on artificial intelligence and constructing stored big data;
A condition setting unit for setting virtual conditions of a reference point in time for basic information and lesion information of a virtual patient, and a current image extraction part for first extracting a lesion image by searching the big data based on the virtual condition of the reference point in time A virtual image extraction unit including a future image extraction part that searches the big data based on the diagnosis label of the primary extracted lesion image and secondaryly extracts a lesion image for a future time point, which is a specific time point after the reference point in time. patient matching module;
Characterized in that, the virtual patient management system comprising a; output module for displaying the firstly extracted lesion image and the secondly extracted lesion image. delete According to claim 1,
The system,
a patient information input module including a lesion image input unit that receives a lesion image to which a diagnosis label for the patient is assigned from a management terminal, and a condition input unit that receives basic information of the patient and current conditions for lesion information from the management terminal; and,
An AI prediction module for extracting a lesion image for a future time point by searching the big data based on the diagnosis label assigned to the lesion image of the patient and the current condition,
The output module,
Characterized in that it comprises a function of displaying the lesion image for the extracted future time point, the virtual patient management system. According to claim 3,
The AI prediction module,
A basic extraction unit for extracting a plurality of lesion images for a future time point by searching the big data based on the diagnosis label assigned to the lesion image of the patient and the current condition;
A deteriorating image designation unit designating a lesion image having a diagnosis label worse than the current condition of the patient among the extracted plurality of lesion images as an aggravated image; and
A relief image designation unit for designating a lesion image having a diagnosis label more alleviated than the patient's current condition among the extracted plurality of lesion images as a relief image;
The output module,
Characterized in that it comprises a function of displaying the deterioration image and alleviation image, virtual patient management system. According to claim 3,
The system,
A test library that stores result data for each item of a blood test to which a diagnostic label is assigned;
The AI learning module,
Constructing big data in which the stored lesion image and the resultant data are classified by disease and stored,
The patient information input module,
Further comprising a result data input unit for receiving result data of the patient from a management terminal,
The AI prediction module,
The virtual patient management system, characterized by extracting a lesion image for a future time point by searching the big data based on the diagnosis label assigned to the lesion image of the patient, the current condition, and the result data. According to claim 5,
The patient information input module,
A life pattern input unit for receiving lifestyle pattern data on drug compliance, sleep habits, drinking habits, eating habits, exercise amount, and stress of the patient from the management terminal;
The AI prediction module,
A data correction unit for correcting the resultant data based on the life pattern data;
The AI prediction module,
The virtual patient management system, characterized by extracting a lesion image for a specific time point in the future by searching the big data based on the diagnostic label assigned to the lesion image of the patient and the corrected result data. delete delete

Images