Artificial Intelligence-Based Urine Test Data Analysis Device and Method

Abstract

The AI-based urine test data analysis device and method according to the embodiment can precisely analyze color changes and minute differences in urine by utilizing a test strip without a color chart and AI machine learning technology. Furthermore, through the embodiment, the possibility of misdiagnosis and omission is reduced compared to existing methods by rapidly and accurately distinguishing 10 types of test items, and early warning and preventive measures are enabled by analyzing disease-related patterns based on collected data. Additionally, through the embodiment, an individual's health status can be continuously monitored and potential disease risks identified through big data and AI. Moreover, through the embodiment, urine test data is collected and analyzed in real time to provide rapid feedback, and results can be checked immediately upon database construction, thereby reducing existing waiting times for tests.

Description

Technology Field

[0001] The technical concept of the present disclosure relates to an artificial intelligence-based urine test data analysis device and method, and more specifically, to an artificial intelligence-based urine test data analysis device and method that classifies 10 types of test items by analyzing urine test data obtained from an in vitro diagnostic device, such as a test strip without a color chart, using an artificial intelligence model. Background Technology

[0002] Unless otherwise indicated in this specification, the contents described in this section are not prior art for the claims of this application, and are not to be recognized as prior art simply because they are included in this section.

[0003] Urine testing is one of the primary diagnostic methods for assessing physical condition and health issues. Traditional urine tests are performed using the dipstick test. This method involves dipping a test strip into urine to detect specific chemical components (e.g., protein, glucose, pH, ketones, blood); it is quick, simple, and allows for on-site results.

[0004] However, test strips only confirm the presence of specific substances and have low sensitivity to trace components. Furthermore, there is a possibility of false positives and false negatives, and precision is limited because results are judged by human visual inspection.

[0005] Furthermore, tests using test strips, such as color charts, may not provide accurate numerical values ​​and may have low reliability. For example, if the colors of the test strip are not clearly distinguishable, the urine test results cannot be accurately determined. In addition, results from test strip tests can fluctuate depending on environmental conditions such as temperature and humidity. Prior art literature

[0006] 1. Korean Patent Registration No. 10-2354702 (January 19, 2022) 2. Korean Patent Publication No. 10-2023-0155124 (November 10, 2023) The problem to be solved

[0007] An artificial intelligence-based urine test data analysis device and method according to an embodiment collects urine test data from an in vitro diagnostic device including a test strip without a color chart, builds a database in real time, and identifies minute differences in color changes appearing in a profile using artificial intelligence machine learning technology.

[0008] In addition, in the embodiment, 10 types of test items are classified quickly and accurately based on an artificial intelligence machine learning analysis method to derive test results, thereby enabling the prediction and prevention of diseases.

[0009] In addition, the embodiment provides a platform that collects health indicators collected from in vitro diagnostic devices on a mobile basis, accumulates data, and utilizes it to support health management services or systems.

[0010] In addition, the embodiment provides a platform capable of disease prediction and management using big data and artificial intelligence.

[0011] In addition, the embodiment collects individual health indicators based on mobile devices to accumulate data, and enables the accumulated data to be utilized in health management services or systems.

[0012] In addition, the embodiment enables digital healthcare and provides a healthcare platform utilizing big data and artificial intelligence.

[0013] However, the problem to be solved according to one embodiment is not limited only to that mentioned above. means of solving the problem

[0014] An artificial intelligence-based urine test data analysis device according to an embodiment includes a memory for storing at least one instruction; and a processor for performing an operation according to the instruction. The processor collects a learning data set including user biometric data, urine test sheet data, and test result information, and implements an analysis model through learning the collected learning data. The analysis model collects biometric data and a urine test sheet image from a user terminal, detects edges in the urine test sheet image, and determines the color of the internal region of the detected edge to derive a urine test result.

[0015] In addition, the analysis model can detect corners, edges, and inspection areas in a captured image of a urine test strip, correct the orientation of the captured image to generate an analysis file, and identify the color of each inspection area in the generated analysis file.

[0016] In addition, the analysis model can compare the HSV value (Y) of the urine sample with the HSV value (X) of the color discrimination table in order to compare data and derive test results using the HSV color model, and analyze how close it is to the pre-stored reference value (X) for each test item.

[0017] In addition, the analysis model can build a training dataset by clustering and labeling urine test data based on the analysis results and then creating a database.

[0018] In addition, the analysis model can cluster urine test data into general population (normal), urinary tract infection (patient), bladder stones (patient), and diabetes (patient).

[0019] In addition, the analysis model can identify an individual's health status and predict diseases based on said health status by extracting the characteristics of each urine test strip imaging data output from the Conventional Neural Network (CNN) and the Convolutional Neural Network (CNN).

[0020] In addition, the analysis model can predict diseases including diabetes, urinary tract infections, and bladder stones from 10 types of test strips, and calculate the recognition rate and accuracy of the expected diseases and the results of the 10 types of test items.

[0021] In addition, the 10 test strips may include test strips for pH, glucose, protein, ketones, blood, leukocytes, nitrite, bilirubin, urobilinogen, and specific gravity. Effects of the invention

[0022] The artificial intelligence-based urine test data analysis device and method according to the embodiment can precisely analyze color changes and minute differences in urine by utilizing a test strip without a color chart and artificial intelligence machine learning technology.

[0023] In addition, through the examples, the possibility of misdiagnosis and omission is reduced compared to existing methods by rapidly and accurately distinguishing 10 types of test items, and early warning and preventive measures are enabled by analyzing disease-related patterns based on collected data.

[0024] In addition, through the embodiments, it is possible to continuously monitor an individual's health status and identify potential risks of disease using big data and AI.

[0025] In addition, through the embodiments, urine test data is collected and analyzed in real time to provide rapid feedback, and results can be checked immediately upon database construction, thereby reducing the existing waiting time for tests.

[0026] Furthermore, through the embodiments, health data is accumulated and analyzed via a mobile platform to provide personalized health management solutions. Additionally, it enables the derivation of management plans optimized for an individual's lifestyle habits and health status.

[0027] In addition, the digital-based health management platform provided in the embodiment enables the strengthening of the digital healthcare ecosystem, such as remote diagnosis and remote consultation, and allows for effective utilization even in areas with low access to medical care.

[0028] In addition, the embodiment enables data to be linked with other healthcare services by integrating with big data and cloud systems, and provides an integrated platform that can be utilized in various fields such as medical institutions, insurance companies, and research institutes.

[0029] In addition, through the embodiments, human error in the inspection process is reduced through automated artificial intelligence analysis, and efficient inspection is provided compared to existing inspection costs.

[0030] In addition, it saves time for patients and medical institutions by providing rapid results, and enables the long-term accumulation of health indicator data to analyze trends and establish health management strategies for individuals and groups.

[0031] In addition, the embodiments progressively improve accuracy and predictive power through the continuous learning of artificial intelligence algorithms, and enable the monitoring of health status at the community or national level and the establishment of public health policies by analyzing large-scale health data.

[0032] In addition, through the embodiments, it is possible to predict disease outbreaks and respond early, and a healthcare platform utilizing AI and big data can promote the development of new digital medical services and business models.

[0033] The effects obtainable from the exemplary embodiments of the present disclosure are not limited to those mentioned above, and other unmentioned effects can be clearly derived and understood by those skilled in the art to which the exemplary embodiments of the present disclosure belong from the description below. That is, unintended effects resulting from the implementation of the exemplary embodiments of the present disclosure can also be derived by those skilled in the art from the exemplary embodiments of the present disclosure. Brief explanation of the drawing

[0034] FIG. 1 is a drawing showing an artificial intelligence-based urine test data analysis system according to an embodiment. FIG. 2 is a block diagram of a urine test data analysis device according to an embodiment. FIG. 3 is a diagram showing the data processing process of an image correction algorithm according to an embodiment. FIG. 4 is a drawing showing an HSV color model according to an embodiment. FIG. 5 is a diagram showing the data processing process of an analysis model according to an embodiment. FIG. 6 is a diagram illustrating an artificial intelligence-based urine test data analysis process according to an embodiment. Specific details for implementing the invention

[0035] Hereinafter, various embodiments of the present disclosure are described in conjunction with the accompanying drawings. As various embodiments of the present disclosure may be subject to various modifications and may have various forms, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the various embodiments of the present disclosure to specific forms, and it should be understood that they include all modifications and / or equivalents and substitutions that fall within the spirit and scope of the various embodiments of the present disclosure. In relation to the description of the drawings, similar reference numerals have been used for similar components.

[0036] In various embodiments of the present disclosure, terms such as “comprising” or “having” are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0037] In various embodiments of the present disclosure, expressions such as “or” include any and all combinations of the words listed together. For example, “A or B” may include A, may include B, or may include both A and B.

[0038] Expressions such as "first," "second," "first," or "second" used in various embodiments of the present disclosure may modify various components of the various embodiments, but do not limit such components. For example, such expressions do not limit the order and / or importance of such components and may be used to distinguish one component from another.

[0039] When it is mentioned that a component is "connected" or "joined" to another component, it should be understood that the component may be directly connected or joined to the other component, but that a new component may also exist between the component and the other component.

[0040] In the embodiments of the present disclosure, terms such as "module," "unit," "part," etc. are used to refer to a component that performs at least one function or operation, and such component may be implemented in hardware or software, or in a combination of hardware and software. Additionally, a plurality of "modules," "units," "parts," etc. may be integrated into at least one module or chip and implemented as at least one processor, except where each needs to be implemented in specific individual hardware.

[0041] Terms such as those defined in commonly used dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the various embodiments of the present disclosure.

[0042] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the attached drawings.

[0043] FIG. 1 is a diagram showing an artificial intelligence-based urine test data analysis system according to an embodiment.

[0044] Referring to FIG. 1, an artificial intelligence-based urine test data analysis system according to an embodiment may be configured to include a urine test data analysis device (100) and a user terminal (200). In the embodiment, the urine test data analysis device (100) collects a urine test sheet image from the user terminal (200) and analyzes the collected urine test sheet image to derive the user's urine test results. Additionally, it is possible to predict diseases that may occur to the user based on the urine test results and calculate and provide the prediction accuracy.

[0045] In addition, the urine test data analysis device (100) collects urine test images, urine test data analysis results, and user biometric information to build big data, and enables the built big data to be used as training data for an analysis model that analyzes urine test data.

[0046] The user terminal (200) is a user terminal that photographs the urine test sheet after performing a urine test. In an embodiment, the user terminal (200) transmits the photographed urine test sheet image to the analysis device (100) and can receive the urine test results from the analysis device (100). Additionally, the user terminal (200) receives an analysis model implemented in the analysis device (100) and can use it after installing the analysis model.

[0047] The artificial intelligence-based urine test data analysis device and method according to the embodiment collect urine test data from an in vitro diagnostic device including a test strip without a color chart, build a database in real time, and identify minute differences in color changes appearing in the profile using artificial intelligence machine learning technology. Furthermore, the embodiment enables the prediction and prevention of diseases by rapidly and accurately distinguishing 10 types of test items based on an artificial intelligence machine learning analysis method to derive test results. Additionally, the embodiment provides a platform that collects health indicators collected from an in vitro diagnostic device via a mobile basis, accumulates data, and utilizes this data to support or utilize health management services or systems.

[0048] Furthermore, the embodiment provides a platform capable of disease prediction and management utilizing big data and artificial intelligence. Additionally, the embodiment collects individual health indicators via mobile devices to accumulate data, and enables the accumulated data to be utilized in health management services or systems. Furthermore, the embodiment promotes digital healthcare and provides a healthcare platform utilizing big data and artificial intelligence.

[0049] FIG. 2 is a block diagram of a urine test data analysis device according to an embodiment.

[0050] In the embodiment, the urine test data analysis device may be implemented as a server. A server is a computing system that provides services to other computers or devices in a computer network or stores and manages data. The server accepts requests from other computers or devices called clients and provides responses or data in response to those requests. The configuration of the urine test data analysis device illustrated in FIG. 2 is merely a simplified example.

[0051] The communication module (110) can be configured regardless of the mode of communication, such as wired or wireless, and can be configured with various communication networks, such as a Personal Area Network (PAN) or a Wide Area Network (WAN). Additionally, the communication module (110) can operate based on the known World Wide Web (WWW) and may utilize wireless transmission technologies used for short-range communication, such as Infrared Data Association (IrDA) or Bluetooth. For example, the communication module (110) may be responsible for transmitting and receiving data necessary to perform a technique according to one embodiment of the present disclosure.

[0052] Memory (120) may refer to any type of storage medium. For example, memory (120) may include at least one type of storage medium among flash memory type, hard disk type, multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, magnetic disk, and optical disk. Such memory (120) may also constitute the database shown in FIG. 1.

[0053] The memory (120) can store at least one instruction that can be executed by the processor (130). Additionally, the memory (120) can store any form of information generated or determined by the processor (130) and any form of information received by the server (200). Additionally, the memory (120) stores various types of modules, instruction sets, or models.

[0054] The processor (130) can perform technical features according to embodiments of the present disclosure to be described below by executing at least one instruction stored in memory (120). In one embodiment, the processor (130) may be composed of at least one core and may include a processor for data analysis and / or processing, such as a central processing unit (CPU) of a computer device, a general purpose graphics processing unit (GPGPU), or a tensor processing unit (TPU).

[0055] This processor (130) can train a neural network or model designed in a machine learning or deep learning manner. To this end, the processor (130) can perform calculations for training the neural network, such as processing input data for training, extracting features from input data, calculating errors, and updating the weights of the neural network using backpropagation. Additionally, the processor (130) can perform inference for a specific purpose using a model implemented in an artificial neural network manner.

[0056] In an embodiment, the processor (130) collects a training data set including user biometric data, urine test strip data, and test result information, and implements an analysis model through training the collected training data. To this end, the processor (130) constructs a training data set based on biometric data and urine test strip data collected from various users. This training data set includes information reflecting the health status of individual users, and this information is used to train an analysis model for disease prediction and diagnosis. Specifically, the processor (130) collects biometric data (e.g., body temperature, heart rate, weight, etc.) from users in real time and collects data on various test items such as pH, glucose, protein, and ketones of urine obtained through urine test strips. In addition, it includes existing test result data (presence of disease, health status evaluation, etc.) to improve the accuracy of the training data set. Subsequently, the processor (130) processes missing values, outliers, etc., from the collected data to convert them into a form suitable for analysis, analyzes correlations between data, and derives patterns between specific variables to increase the efficiency of model training.

[0057] Additionally, the processor (130) integrates biometric data, urine test strip data, and test result information to form a multidimensional learning dataset. In the embodiment, the dataset anonymizes individual user data and maintains data balance to facilitate generalization of the learning model. Subsequently, the processor (130) trains an artificial intelligence-based analysis model using the learning dataset. In the embodiment, the analysis model learns data patterns by utilizing machine learning or deep learning algorithms (e.g., neural networks, random forests, etc.). In the embodiment, the processor (130) improves the accuracy and reliability of the analysis model through iterative learning and verification processes, and the optimized model can be used to predict diseases or derive test results by receiving user biometric data and urine test strip data as input. In the embodiment, the analysis results of the urine test data are provided to the user, and user feedback and additional data are reflected in a new learning dataset to continuously update the model.

[0058] In addition, the analysis model collects biometric data and images of urine test strips from a user terminal, detects edges in the images of the urine test strips, and determines the color of the internal areas of the detected edges to derive urine test results. To this end, the analysis model collects biometric data from the user in real time through a user terminal (e.g., smartphone, tablet, etc.). In the embodiment, the biometric data includes various indicators reflecting the user's health status, such as body temperature, heart rate, and blood pressure. Furthermore, the analysis model acquires images of the urine test strips using the camera of the user terminal, and provides guidelines to the user (e.g., shooting angle, lighting conditions) during the shooting process to ensure optimal image quality. Subsequently, the analysis model applies an edge detection algorithm to analyze the acquired urine test strip images. At this stage, edge detection includes the process of accurately identifying the boundaries of color zones within the urine test strip to define the area to be analyzed. In the embodiment, the analysis model distinguishes the color zones of the urine test strip (e.g., test items such as pH, glucose, and protein) through edge detection, and this is utilized as basic data for result analysis by test item. Subsequently, the analysis model converts the color of the detected edge interior region into an HSV or RGB color model and analyzes it as quantitative data. In addition, it derives the value of each test item by comparing the color data with a predefined reference color chart (e.g., a color discrimination table). For example, if a specific color value is close to the reference color of the glucose test strip, it is determined as the result of that item.

[0059] Subsequently, the analysis model synthesizes the color analysis results of each test item to derive urine test results. For example, it evaluates the user's health status by summarizing results such as urine pH, glucose concentration, and the presence of protein. The test results derived in the embodiment are transmitted to the user terminal in real time and provided to the user in a visual and easy-to-understand format (e.g., graphs, text descriptions). In the embodiment, the test results are linked with a health management application on the user terminal to accumulate personal health data and support long-term health management plans based on this data. Furthermore, continuous updates and improvements to the analysis model can be pursued based on user feedback regarding test accuracy, additional symptoms, etc.

[0060] In addition, the analysis model detects corners, edges, and test areas in the captured image of the urine test strip, corrects the orientation of the captured image to generate an analysis file, and identifies the color of each test area in the generated analysis file. Figure 3 is a diagram illustrating the data processing process of an image correction algorithm according to an embodiment. Referring to Figure 3, the analysis model uses a corner detection algorithm (e.g., Harris Corner Detection, FAST) to detect rectangular or special shaped corners of the urine test strip. In the embodiment, corners are utilized to identify the overall structure of the urine test strip and to precisely determine the location of the test area. Additionally, the analysis model detects edges within the image to identify the boundary lines of the test area. In the embodiment, algorithms such as Canny edge detection or a Sobel filter are used to clearly separate each test item area of ​​the urine test strip.

[0061] In addition, the analysis model defines the test area of ​​the urine test strip based on corner and edge information. In the embodiment, each test area (e.g., pH, glucose, protein, etc.) is extracted in the form of a square or rectangle, and each area corresponds to a test item. During the process of defining the test area, the analysis model removes noise or distortion within the image to secure highly reliable area data. Furthermore, the analysis model checks for tilt or skewed orientation of the urine test strip based on the corner detection results. For instance, the analysis model aligns the image by applying geometric transformations (e.g., rotation, scaling). Specifically, the analysis model can correct the tilt relative to the reference axis using OpenCV's Hough Line Transform. Additionally, orientation correction ensures that each test item on the urine test strip is accurately positioned at the reference location.

[0062] Subsequently, the analysis model extracts color data for each inspection area based on the corrected image to generate an analysis file. Additionally, it saves the color information (Hue, Saturation, Value, or RGB values) of each inspection item into an analysis file. In the embodiment, the analysis file includes the location, color value, and result status of each item, and the file format can be generated in a structure suitable for analysis and linkage, such as JSON, XML, or CSV.

[0063] Subsequently, the analysis model identifies the color of the test area. In the embodiment, the analysis model converts the color of each test area into an HSV or RGB color model to quantitatively analyze it, and determines the test item result by comparing the color value with a predefined color chart (color discrimination table). For example, if the color of the pH area is closest to "120, 0.5, 0.8," the analysis model determines it as "pH 7.0." Additionally, in the embodiment, average value calculation or color clustering is applied to ensure consistency of color values. Subsequently, the analysis model synthesizes the results of each test item from the analysis file to generate a urine test result. In the embodiment, the analysis result is output in a visual format (e.g., graph, diagram) that is easy for the user to understand, or it can be linked with a health management system.

[0064] In addition, the analysis model utilizes the HSV color model to compare data and derive test results by sequentially comparing the HSV value (Y) of the urine sample with the HSV value (X) of the color discrimination table, and For each test item, the degree of proximity to a pre-stored reference value (X) is analyzed. Fig. 4 is a diagram showing an HSV color model according to an embodiment. Referring to Fig. 4, the analysis model processes a captured image of a urine test strip to collect color information of the urine sample and extracts HSV values ​​for each test item area. The HSV values ​​represent hue, saturation, and value, respectively, and quantitatively express the color data of the urine sample. The HSV values ​​(Y) extracted in the embodiment are used as input data for comparison with the reference values ​​(X) corresponding to each test item of the urine test strip. Additionally, the analysis model constructs a color discrimination table. In the embodiment, the color discrimination table includes pre-defined HSV reference values ​​(X) for each test item (pH, glucose, protein, etc.) of the urine test strip. The reference values ​​(X) consist of HSV values ​​corresponding to various states (e.g., normal, borderline, abnormal) and are constructed for accurate discrimination of each state.

[0065] In the embodiment, the analysis model sequentially compares the HSV value (Y) extracted from the urine sample with the reference value (X) of the color discrimination table. The comparison is performed by calculating the difference between each HSV channel (H, S, V) and quantitatively evaluating the similarity between Y and X through Euclidean distance or other distance measurement algorithms. In the embodiment, the reference value (X) with the smallest calculated distance is identified to determine the corresponding state. Subsequently, the analysis model returns the test state corresponding to the most similar reference value (X) as a result of the comparison between Y and X. For example, if the HSV value (Y) of the urine sample is closest to the HSV reference value (X) corresponding to "pH 7.0" in the color discrimination table, the result of the corresponding test item is determined to be "pH 7.0". Additionally, the same process is repeated for each test item to derive the final test result. Afterward, the analysis model transmits the derived test result to the user terminal and visually displays the result (e.g., graph, diagram, text) so that the user can easily understand it. In the embodiments, the analysis results are stored in a cloud or local database and utilized for long-term health management and data analysis. Furthermore, the analysis model continuously updates the color discrimination table and distance measurement algorithm based on user feedback and additional data to improve accuracy. Additionally, the model's flexibility is enhanced by dynamically modifying the table when new states or reference values ​​occur.

[0066] In addition, the analysis model constructs a training dataset by clustering and labeling urine test data based on the analysis results and storing it in a database. To this end, the analysis model collects images of urine test strips and related biometric data from users. Subsequently, the urine test data is converted into a quantified form (e.g., Hue, Saturation, Value) and metadata is added. In the embodiment, the metadata includes additional health data such as the user's age, gender, lifestyle habits, and medical history, and the analysis model collects this metadata to reinforce the context of the training data. Afterward, the analysis model ensures the quality of the training data by processing missing values ​​or abnormal data. Furthermore, the analysis model normalizes the various ranges of urine test items to improve efficiency in clustering and model training. For example, the analysis model uses Principal Component Analysis (PCA) to reduce data dimensionality, remove noise, and enable learning by focusing on key patterns.

[0067] Furthermore, the analysis model selects an appropriate clustering algorithm to group the data. For example, the analysis model may choose one of the following: the K-means algorithm, which groups data based on a fixed number of clusters; DBSCAN, a density-based clustering suitable for handling outliers; or Hierarchical Clustering, which identifies relationships between data through a hierarchical structure. Subsequently, the analysis model sets key variables to be used for clustering (e.g., pH, glucose concentration, protein presence, etc.) and applies the clustering algorithm to classify data of similar characteristics into the same group.

[0068] In the embodiments, the analysis model clusters urine test data into categories such as general population (normal individuals), urinary tract infection (patients), bladder stones (patients), and diabetes (patients). In the embodiments, the analysis model applies a suitable algorithm to cluster the urine test data. For example, it clusters by setting a fixed number of clusters (e.g., 4) based on the data distribution, or effectively handles outliers by clustering based on data density. Additionally, it can analyze hierarchical relationships between data. In the embodiments, the analysis model sets the number of clusters according to four predefined groups (general population, urinary tract infection, bladder stones, and diabetes). Subsequently, it defines the health status of the corresponding group by analyzing the key characteristic values ​​(pH, glucose concentration, white blood cell count, etc.) of each cluster. For example, the general population (normal individuals) is set to have a pH of 6–7 and negative for glucose and protein, while urinary tract infection is set to positive for white blood cells and detectable nitrite. Furthermore, bladder stones are defined by low pH (acidity) and a high specific gravity above a certain level, while diabetes is defined by glucose positivity and the detection of ketone bodies. Subsequently, the analysis model clearly defines the results by assigning labels such as "Normal," "Urinary Tract Infection," "Bladder Stone," and "Diabetes" to each cluster. Additionally, the analysis model receives new urine test data, performs clustering and labeling, and predicts and diagnoses the likelihood of disease based on the clustering results. Moreover, the analysis model databases the clustered and labeled data to utilize as a training dataset. Finally, the analysis model continuously updates the clustering model through user feedback and additional data.

[0069] FIG. 5 is a diagram illustrating the data processing process of an analysis model according to an embodiment. Referring to FIG. 5, in the embodiment, the analysis model analyzes the color of each area of ​​the microscopic examination paper. In the embodiment, the analysis model analyzes the characteristics of each cluster and assigns a label. For example, labels such as Normal, Borderline, and Abnormal may be assigned. In the embodiment, the label may be determined by referring to expert opinions or existing disease databases. Additionally, the analysis model trains a classification model (SVM, Decision Tree, etc.) using existing data and labels, and automatically labels new data.

[0070] Subsequently, the analysis model stores data including urine test data, clustering results, and labels in a relational database (RDB) or NoSQL database. In the embodiment, data fields may include user ID, test date, test items (pH, glucose, etc.), HSV values, cluster ID, labels, etc. Additionally, the analysis model manages new data in an integrated manner with the existing database. In the embodiment, the analysis model can re-evaluate and update clusters and labels based on data changes.

[0071] Subsequently, the analysis model constructs a training dataset and generates cluster and label-based datasets. In the embodiment, the analysis model constructs a training dataset by utilizing clustering and labeling results, and prepares the input data required for the learning algorithm by including a cluster ID and a label for each data point. Additionally, the analysis model ensures that normal and abnormal data are included in a balanced manner to improve the generalization performance of the learning model. Furthermore, the analysis model separates the data into training, validation, and test datasets for use in evaluating model performance and trains based on the training dataset. Additionally, the analysis model periodically verifies the cluster and label results to continuously improve the quality of the training dataset.

[0072] In the embodiments, data is organized through clustering and labeling to optimize model training, and new patterns can be discovered by analyzing similarity between data through clustering. Additionally, the predictive power of the analysis model is enhanced through labeled data, and the database and model can be flexibly updated even when new data is added.

[0074] In addition, the analysis model extracts the characteristics of each urine test strip image data output from a Conventional Neural Network (CNN) and a Convolutional Neural Network (CNN) to identify an individual's health status and predict diseases based on said health status. To this end, a urine test strip image is captured using a user device (smartphone, tablet, etc.), and a high-resolution image is secured to enhance the performance of the analysis model. The analysis model performs image normalization (e.g., brightness correction, resizing) and noise removal. Subsequently, the analysis model separates the test area of ​​the urine test strip (e.g., edge detection, region segmentation) to extract only the area to be analyzed.

[0075] In the embodiment, convolutional layers process the input urine test strip image through the convolutional layers of a CNN to extract low-level features (e.g., edges, textures, colors). Additionally, the analysis model learns suitable features by adjusting the filter size (e.g., 3x3, 5x5) and reduces the feature map through pooling layers to increase computational efficiency and preserve important features. In the embodiment, Max Pooling or Average Pooling can be used. Furthermore, the analysis model learns low-level features (e.g., colors, shapes) and high-level features (e.g., test patterns, color changes) through a multilayer CNN.

[0076] Subsequently, the analysis model utilizes feature maps generated from the CNN output to reflect the status of each test item (pH, glucose, protein, etc.), and constructs a model that correlates urine test data with individual health status through Fully Connected Layers. Afterward, it calculates the probability of specific disease states using functions such as the Softmax function to identify the individual's health status.

[0077] In the embodiment, the analysis model classifies normal, borderline, and abnormal states based on the output data of the CNN. Subsequently, it determines abnormalities in the health status by detecting pattern changes in specific test items (e.g., excessive pH change, glucose detection).

[0078] In addition, the analysis model inputs the extracted feature data into an existing disease prediction model to derive results. For example, urinary tract infections (UTIs) can be detected by leukocyte and nitrite patterns, and diabetes can be detected by glucose and ketone bodies. Kidney stones can be set by low pH and high specific gravity. Furthermore, the analysis model presents the likelihood of a specific disease occurring as a probability value (e.g., 85% probability) and enables the simultaneous prediction of multiple diseases by utilizing the multi-task learning capabilities of CNNs.

[0079] Furthermore, the analysis model provides users with health status and disease prediction results in a visual format (e.g., graphs, charts) and stores prediction results and user feedback data for use in model training. Additionally, the analysis model suggests warning messages and preventive measures for users at high risk of disease.

[0080] In the embodiment, the limitations of existing urine testing methods are overcome through high-precision analysis using CNN, and various diseases can be simultaneously predicted based on urine test strip image data to reduce time and costs.

[0081] Furthermore, the analysis model predicts diseases including diabetes, urinary tract infections, and bladder stones from 10 types of test strips, and calculates the recognition rate and accuracy of the expected diseases based on the results of the 10 test items. To this end, data on pH, glucose, protein, ketone bodies, blood, leukocytes, nitrite, bilirubin, urobilinogen, and specific gravity are collected from the 10 types of test strips. Subsequently, the analysis model normalizes the data for each test item to transform it into a format suitable for the model, and ensures data quality by handling missing values ​​and removing outliers. Then, the analysis model derives features from the 10 test items that are significant for disease prediction. For example, diabetes is identified based on glucose and ketone body levels, while urinary tract infections are identified based on leukocyte, nitrite, and pH levels. Additionally, bladder stones are identified based on pH and specific gravity.

[0082] In addition, the analysis model learns disease prediction by utilizing machine learning or deep learning models (e.g., Random Forest, XGBoost, CNN) and uses a multi-class classification model to simultaneously predict multiple diseases including diabetes, urinary tract infections, and bladder stones.

[0083] In the embodiment, the analysis model takes the values ​​of each test item as input and calculates the probability of disease occurrence. For example, the analysis model determines diabetes if glucose > 180 mg / dL or ketone bodies are detected, and determines urinary tract infection if leukocytes are positive, nitrites are detected, or pH increases. Additionally, it can determine bladder stones if pH is 5 or lower or specific gravity increases. Subsequently, the analysis model derives predicted probability values ​​for each disease (e.g., diabetes 85%, urinary tract infection 90%). Then, the prediction results are compared with actual data (labeled health data) to calculate accuracy, precision, recall, F1 score, etc.

[0084] In addition, the analysis model evaluates the contribution of each test item (pH, glucose, etc.) to disease prediction. In the examples, SHAP (SHapley Additive exPlanations) or the Feature Importance algorithm can be utilized to calculate contribution scores for each item.

[0085] The analysis model analyzes the correlation between the value of each item and the predicted disease outcome, and visually provides the user with the probability of disease occurrence and the recognition rate of each disease (e.g., diabetes 85%, urinary tract infection 90%, bladder stones 70%).

[0086] In addition, the analysis model expresses the reliability of the prediction results numerically (e.g., prediction accuracy 92%) and helps users understand by highlighting test items and results that are important for specific diseases.

[0087] In the embodiment, the processor (130) collects a training data set containing data on 10 items of a urine test strip, urine test results, and user information. Subsequently, the collected training data is trained using machine learning or deep learning techniques to implement an analysis model capable of predicting the user's health status and potential diseases by analyzing images of the urine test strip. The analysis model provided in the embodiment can be used to evaluate the user's health status through accurate analysis of each item of the urine test strip (e.g., pH, protein, glucose, etc.) and to predict the likelihood of developing specific diseases (e.g., diabetes, kidney disease, etc.). Additionally, the accuracy of the analysis results can be improved by utilizing user information (e.g., age, gender, body mass index, etc.). The purpose of this analysis model is to provide data that medical professionals can refer to for diagnosis or to help the user manage their own health status.

[0088] In addition, the analysis model derives the probability of occurrence for each disease and the user's condition based on the analysis results of the captured images of the test strips. To this end, the analysis model utilizes the captured images of urine test strips as input data to evaluate the probability of occurrence for each disease and the user's overall health status. In the embodiment, the analysis model requires the captured images of the urine test strips to undergo a pre-analysis processing step. This includes image distortion correction, noise removal, and color and contrast correction. This improves image quality and enables the analysis model to recognize accurate data. Subsequently, the analysis model extracts color or pattern information corresponding to each item on the urine test strip (e.g., pH, protein, glucose, etc.) based on the pre-processed images. The feature extraction step quantitatively analyzes each item on the test strip using computer vision techniques and deep learning algorithms. Furthermore, the analysis model derives values ​​for each item by comparing the extracted feature data with predefined reference values ​​(e.g., normal range or warning level). In this process, the user's basic information (e.g., age, gender, body mass index, etc.) is also considered to increase the precision of the analysis results.

[0089] Furthermore, the analysis model, which is a trained machine learning or deep learning model, calculates the probability of occurrence for each disease based on the analysis results for each item. For example, it provides predicted probabilities such as an 85% probability of developing diabetes or a 70% probability of developing kidney disease under specific conditions. Additionally, in the embodiment, the analysis model comprehensively evaluates the predicted probability of disease occurrence and the analysis results for each item to derive the user's overall health status. For example, it provides status information such as "possibility of mild kidney function decline," "within normal range," or "immediate additional examination required." In the embodiment, the analysis results are provided in a visualized form to make them easy for the user to understand. For example, graphs, charts, color codes, etc., are used to clearly indicate the results for each item and the probability of disease occurrence. Furthermore, if additional necessary measures or expert consultation are required, they can be recommended.

[0090] Furthermore, the analysis model identifies trends over time by accumulating the analysis results of the user's urine test strip images, the probability of disease occurrence, and the user's condition. By accumulating the analysis results derived from the user's urine test strip images, the probability of disease occurrence, and user condition information over time, the analysis model provides the functionality to track changes in health status and identify trends. To this end, the analysis model stores each user's urine test strip analysis results (e.g., pH, protein concentration, glucose concentration, etc.), the predicted probability of disease occurrence based on these results, and user condition information in a database. The data is stored with time stamps, allowing for the tracking of changes between test points.

[0091] In the embodiments, data is stored in chronological order based on the user's unique ID, and results are classified and structured by item. Changes in the value of specific items (e.g., pH, protein concentration) are calculated over time, and average values, maximum / minimum values, and rates of change are derived. Additionally, the analysis model tracks temporal changes in the probability of disease occurrence to derive information such as, for example, "the probability of developing diabetes has increased by 20% over the past 6 months." Subsequently, it determines whether the health status is improving, maintaining, or deteriorating based on the status assessment information. Furthermore, the analysis model visualizes the trends of change over time using graphs, line charts, heatmaps, etc., and provides this information to the user. For example, numerical changes by item, disease probability, and overall status may be provided. In the embodiments, numerical changes by item may be represented as a line graph using a time axis (x) and a value axis (y), and disease probability may be represented as a line graph comparing trends for each disease. The overall status may be represented by changes in health status indicated by color codes or icons.

[0092] In addition, the analysis model provides a graph that predicts future conditions based on trend data (e.g., predicted condition after 3 months using the prediction model). Subsequently, the analysis model provides notifications to the user or medical professional if specific items exceed normal ranges or a worsening trend is detected based on the analysis results. In the embodiment, advice or recommended measures to improve health conditions are provided based on the trend of change. For example, recommended measures such as, "Protein levels are showing an increasing trend. Increase fluid intake and additional testing are recommended," may be provided.

[0093] Additionally, the processor (130) utilizes time-accumulated data to train a machine learning / deep learning model and uses it to predict future health conditions or the likelihood of disease occurrence. This model enables more accurate results to be derived based on personalized data.

[0094] In addition, the analysis model converts the analysis results of accumulated user test sheet images, the probability of disease occurrence, and the user's status into visualization elements including graphs.

[0095] In the embodiments, the analysis model organizes accumulated data into a visualizable format. Each item (e.g., pH, protein concentration), disease probability, and user status are arranged along with time stamps. In the embodiments, the analysis model selects appropriate graphs considering the data type and user understanding. For instance, changes by item are selected to be visualized as line graphs, while disease probability is selected to be visualized as bar graphs or stacked area graphs. User status is selected to be visualized as a heatmap or icon-based.

[0096] Furthermore, the analysis model represents the temporal change of items using a line graph that utilizes a time axis (x-axis) and specific item values ​​(y-axis). The line graph clearly shows the upward or downward trends of the values. In the embodiment, the analysis model visualizes the probability of each disease occurrence along the time axis. It uses bar graphs to compare changes by disease or utilizes stacked area graphs to visually represent the overall probability. Additionally, the analysis model represents the status assessment as a heatmap. Color codes (e.g., Normal: Green, Mild Abnormality: Yellow, Severe: Red) are applied based on the time axis (x-axis) and key statuses (y-axis). Moreover, the analysis model can emphasize the long-term trends of the data by adding linear regression or curved trend lines to each graph. Furthermore, the analysis model utilizes machine learning models to include projected future data in the graphs. For example, future data can be represented by dotted lines or other colors. Additionally, the analysis model integrates all graphs into a single screen to allow users to intuitively grasp all the data. For instance, a line graph showing changes by item is placed at the top, and a bar graph showing the probability of disease occurrence is placed in the middle.

[0097] A user status heatmap can be placed at the bottom. Additionally, the analysis model is implemented using a responsive design for its graphs and visualizations, ensuring easy access for users across various devices such as PCs, tablets, and smartphones. Furthermore, a feature is added that allows users to view detailed data in a pop-up window when they click on a specific item.

[0098] Furthermore, if the analysis model detects items indicating an abnormal state as a result of analyzing the captured image of the test strip, it increases the analysis sensitivity of said abnormal items by a certain percentage in subsequent test analyses. In the embodiment, the analysis model analyzes the captured image of the urine test strip to detect abnormal states by comparing them with the normal range for each item. If a specific item (e.g., protein concentration, glucose concentration, etc.) exceeds the normal range or is identified as being in a borderline state, the analysis sensitivity for that item is increased in the following manner. To this end, the analysis model sets parameters for adjusting the analysis sensitivity for items identified as abnormal. For example, the reference value of the image analysis algorithm is adjusted to enable more precise detection of color changes, concentration differences, or subtle pattern changes of the item during the image processing process. Subsequently, the weight of the item is increased in the machine learning-based analysis model, so that the model places greater emphasis on learning and predicting data for that item. This improves the analysis accuracy for items in abnormal states.

[0099] In addition, for items identified as abnormal, the analysis model extracts detailed data more precisely by increasing the resolution of captured images or performing magnified image analysis on specific areas. For example, if protein concentration increases, it captures subtle changes by performing magnified analysis on the color change area corresponding to that item.

[0100] In addition, the analysis model is set to increase the sensitivity for anomaly items by a certain percentage (e.g., 10% to 50%). This percentage can be dynamically adjusted based on the severity of the anomaly and the characteristics of the item. For example, the sensitivity is increased by 10% for minor anomalies, and 50% sensitivity is applied for severe anomalies.

[0101] Furthermore, the analysis model compares the results of subsequent tests performed with adjusted sensitivity with previous results to evaluate changes in the status and improvement of the relevant item. Through this process, the analysis model continuously learns and can maintain or further adjust personalized sensitivity settings. In the embodiment, if the abnormal condition persists or worsens based on the results of the sensitivity enhancement analysis, a warning is provided to the user, and additional testing or expert consultation is recommended. By detecting abnormal conditions early and evaluating them more precisely in subsequent tests, the embodiment enables the provision of rapid and accurate health management information to the user.

[0102] In addition, the analysis model collects lifestyle information, including the food consumed by the user and sleep duration, and identifies the correlation between the collected lifestyle information and the results of urine test data analysis. To this end, the analysis model collects lifestyle information such as the user's food intake, sleep duration, and physical activity levels, and derives correlations with the results of urine test data analysis based on this information, thereby more precisely evaluating the user's health status and suggesting directions for improvement. In the embodiment, the analysis model collects the food consumed by the user, sleep duration, and other lifestyle information through various methods. Food intake information is obtained through diet records entered by the user, photo-based food recognition, or linked health management applications. Sleep information is collected through wearable devices, smartphone applications, or data manually entered by the user. Subsequently, the item-specific analysis results derived from urine test data (e.g., pH, protein, glucose concentration, etc.) and the collected lifestyle information are stored in a database, and the association between the two datasets is mapped based on time stamps. For example, it analyzes the correlation between the intake of a high-protein diet on a specific date and an increase in protein concentration in the urine.

[0103] Furthermore, the analysis model derives correlations by analyzing collected lifestyle information and urine test data using machine learning algorithms or statistical techniques. To this end, data preprocessing is performed to improve data quality through handling missing values, removing outliers, and standardization, and correlation coefficients are calculated. In the examples, Pearson correlation coefficients, Spearman rank correlation coefficients, etc., are used to quantify the correlation between each piece of lifestyle information and urine test item. Subsequently, the analysis model detects patterns between the data by utilizing clustering or regression analysis techniques. Finally, the results of the correlation analysis are visualized using graphs, heatmaps, or network diagrams so that users can easily understand them. For example, the impact of specific food intake on changes in urine test results is expressed through color intensity or line thickness.

[0104] In the embodiment, the analysis model evaluates the impact of the user's lifestyle habits on urine test data based on the analysis results and provides customized advice for improvement. For example, it provides feedback such as, "Recent intake of a high-sodium diet has increased sodium levels in the urine. It is recommended to reduce sodium intake."

[0105] In the embodiment, as the user's lifestyle information and urine test data are continuously updated, the analysis model learns new data and continuously improves the accuracy of correlation evaluation. This enables increased precision in personalized health management.

[0106] Furthermore, the analysis model validates the model design by comparing profiles indicating specificity and significance between normal individuals and patients with diabetes, urinary tract infections, and bladder stones, with weights indicating the influence of representative markers. To this end, the analysis model collects urine test data from normal individuals and subjects belonging to each disease group. The data includes results for each item on the urine test strip (e.g., pH, protein, glucose concentration, etc.) and basic user information (e.g., age, gender, medical history, etc.). In the examples, the collected data is classified into a normal group and disease groups (diabetes, urinary tract infection, bladder stones), and reference data is established to analyze the specific characteristics and significance of each group. Subsequently, representative markers specifically observed in each disease group are derived. For example, an increase in glucose concentration is derived as a marker for diabetes, and an increase in white blood cell count is derived as a marker for urinary tract infections. Changes in pH and crystal formation are derived as markers for bladder stones. Finally, the analysis model assigns weights to evaluate the influence of the representative markers within the analysis model. Initial weights are set based on training data, and the impact of the weights on disease group classification is compared. Subsequently, the analysis model statistically analyzes the profile differences between the normal group and the disease group. In the embodiments, the analysis model can analyze profile differences between the normal group and the disease group through analysis of variance (ANOVA), correlation analysis, and ROC curve analysis. Analysis of variance (ANOVA) verifies whether the values ​​of each representative marker show a statistically significant difference between groups. Correlation analysis calculates the correlation between the representative marker and the disease group classification to evaluate the contribution of the marker to disease diagnostic accuracy. ROC curve analysis evaluates diagnostic accuracy based on the sensitivity (true positive rate) and specificity (true negative rate) of the representative marker.

[0107] Furthermore, the analysis model compares performance changes based on the weighting of representative markers in the analysis model for disease group classification, and evaluates the classification accuracy, sensitivity, and specificity of the model with modified weights. In the examples, the impact of weights on the disease group classification results is quantified to verify whether the model design is optimal. Additionally, the analysis model validates model performance by dynamically adjusting weights to more clearly distinguish the differences between the normal group and the disease group. Subsequently, the analysis results are visualized to represent the profile differences between the normal group and each disease group using heatmaps, graphs, or clustering diagrams.

[0108] Cross-validation is performed to evaluate the reliability of the final model design, confirming consistent performance across both training and test data. In the examples, the validated model is utilized to analyze new urine test data, and the accuracy and efficiency of diagnosis are evaluated based on the model's classification results.

[0109] The model is continuously improved by reflecting additional clinical data or user feedback, and the optimization of representative markers and weights is iteratively performed. In the examples, the specificity and significance between healthy individuals and specific disease groups are quantitatively evaluated, and the analysis model design is validated based on this to enhance the accuracy and reliability of disease diagnosis.

[0110] Hereinafter, we will look at FIG. 6. The artificial intelligence-based urine test data analysis method illustrated in FIG. 6 can be performed by an artificial intelligence-based urine test data analysis device (100) including a processor (130).

[0111] Meanwhile, FIG. 6 is merely illustrative, and the concept of the present invention is not to be interpreted as being limited to that illustrated in FIG. 6. For example, each step may be configured in a different order than that illustrated in FIG. 6, at least one of the steps illustrated in FIG. 6 may not be performed, or one or more steps not illustrated in FIG. 6 may be additionally performed.

[0112] Below, an artificial intelligence-based urine test data analysis method will be described in turn. Since the operation (function) of the artificial intelligence-based urine test data analysis method according to the embodiment is essentially the same as the function of the artificial intelligence-based urine test data analysis system, descriptions that overlap with FIGS. 1 to 5 will be omitted.

[0113] FIG. 6 is a diagram illustrating an artificial intelligence-based urine test data analysis process according to an embodiment. Referring to FIG. 6, in step S110, a training data set including user biometric data, urine test strip data, and test result information is collected, and in step S120, an analysis model is implemented through training the collected training data.

[0114] In step S130, the analysis model collects biometric data and a urine test strip image from the user terminal, and in step S140, detects edges in the urine test strip image and determines the color of the area inside the detected edges to derive a urine test result.

[0115] The artificial intelligence-based urine test data analysis device and method according to the embodiment can precisely analyze color changes and minute differences in urine by utilizing a test strip without a color chart and artificial intelligence machine learning technology.

[0116] In addition, through the examples, the possibility of misdiagnosis and omission is reduced compared to existing methods by rapidly and accurately distinguishing 10 types of test items, and early warning and preventive measures are enabled by analyzing disease-related patterns based on collected data.

[0117] In addition, through the embodiments, it is possible to continuously monitor an individual's health status and identify potential risks of disease using big data and AI.

[0118] In addition, through the embodiments, urine test data is collected and analyzed in real time to provide rapid feedback, and results can be checked immediately upon database construction, thereby reducing the existing waiting time for tests.

[0119] Furthermore, through the embodiments, health data is accumulated and analyzed via a mobile platform to provide personalized health management solutions. Additionally, it enables the derivation of management plans optimized for an individual's lifestyle habits and health status.

[0120] In addition, the digital-based health management platform provided in the embodiment enables the strengthening of the digital healthcare ecosystem, such as remote diagnosis and remote consultation, and allows for effective utilization even in areas with low access to medical care.

[0121] In addition, the embodiment enables data to be linked with other healthcare services by integrating with big data and cloud systems, and provides an integrated platform that can be utilized in various fields such as medical institutions, insurance companies, and research institutes.

[0122] In addition, through the embodiments, human error in the inspection process is reduced through automated artificial intelligence analysis, and efficient inspection is provided compared to existing inspection costs.

[0123] In addition, it saves time for patients and medical institutions by providing rapid results, and enables the long-term accumulation of health indicator data to analyze trends and establish health management strategies for individuals and groups.

[0124] In addition, the embodiments progressively improve accuracy and predictive power through the continuous learning of artificial intelligence algorithms, and enable the monitoring of health status at the community or national level and the establishment of public health policies by analyzing large-scale health data.

[0125] In addition, through the embodiments, it is possible to predict disease outbreaks and respond early, and a healthcare platform utilizing AI and big data can promote the development of new digital medical services and business models.

[0126] Meanwhile, the methods according to the various embodiments of the present invention described above can be implemented in the form of an application or software program that can be installed on an existing electronic device.

[0127] In addition, the whole or part of the method may be composed of multiple software function modules and implemented on an operating system (OS). Alternatively, each step may be composed of a single software function module, or each step may be combined to form a single software function module and implemented on an operating system. Therefore, even if all of the embodiments of the present disclosure are not implemented as a single software function module, if multiple software function modules implement each step of the present disclosure and multiple software function modules are implemented on a single operating system, it can be understood that the method of the present disclosure has been implemented.

[0128] In addition, the methods according to the various embodiments of the present invention described above can be implemented solely through software upgrades or hardware upgrades of existing electronic devices. Furthermore, the various embodiments of the present invention described above can also be performed through an embedded server equipped in an electronic device or an external server of the electronic device.

[0129] Meanwhile, according to one embodiment of the present invention, the various embodiments described above may be implemented as software comprising instructions stored on a computer-readable recording medium using software, hardware, or a combination thereof. In some cases, the embodiments described herein may be implemented as the processor itself. According to the software implementation, embodiments such as the procedures and functions described herein may be implemented as separate software modules. Each of the software modules may perform one or more functions and operations described herein.

[0130] Meanwhile, a computer or a similar device may include a device according to the disclosed embodiments, which is capable of calling instructions stored from a storage medium and operating according to the called instructions. When said instructions are executed by a processor, the processor may perform a function corresponding to said instructions directly or by using other components under the control of said processor. The instructions may include code generated or executed by a compiler or an interpreter.

[0131] A computer-readable recording medium may be provided in the form of a non-transitory computer-readable recording medium. Here, "non-transitory" simply means that the storage medium does not contain a signal and is tangible, without distinguishing whether data is stored semi-permanently or temporarily on the storage medium. In this context, a non-transitory computer-readable medium refers to a medium that stores data semi-permanently and is readable by a device, rather than a medium that stores data for a short moment, such as registers, caches, or memory. Specific examples of non-transitory computer-readable media may include CDs, DVDs, hard disks, Blu-ray discs, USBs, memory cards, and ROMs.

[0132] As described above, exemplary embodiments have been disclosed in the drawings and specification. Although specific terms have been used to describe the embodiments in this specification, they are used only for the purpose of explaining the technical concept of this disclosure and are not intended to limit the meaning or the scope of this disclosure as defined in the claims. Therefore, those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of this disclosure should be determined by the technical concept of the appended claims.

Claims

Claim 1 A urine test data analysis device comprising: a memory storing at least one instruction for artificial intelligence-based urine test data analysis; and a processor performing an operation according to said instruction, wherein the processor collects a learning data set including user biometric data, urine test sheet data, and test result information, implements an analysis model through learning the collected learning data, and the analysis model collects biometric data and a urine test sheet image from a user terminal, detects edges in the urine test sheet image, and determines the color of the internal region of the detected edge to derive a urine test result. Claim 2 A urine test data analysis device according to claim 1, wherein the analysis model detects corners, edges, and test areas in a urine test strip image, corrects the orientation of the image to generate an analysis file, and identifies the color of each test area in the generated analysis file. Claim 3 In paragraph 2, the above analysis model utilizes an HSV color model to compare data and derive test results, sequentially comparing the HSV value (Y) of a urine sample with the HSV value (X) of a color discrimination table, and analyzing how close it is to a pre-stored reference value (X) for each test item. This is a urine test data analysis device. Claim 4 In paragraph 3, the above analysis model is a urine test data analysis device that constructs a training data set by clustering and labeling urine test data according to the analysis results and then creating a database. Claim 5 In paragraph 4, the above analysis model is a urine test data analysis device that clusters urine test data into general public (normal person), urinary tract infection (patient), bladder stone (patient), and diabetes (patient). Claim 6 In claim 4, the above analysis model is a urine test data analysis device that extracts characteristics of each urine test sheet imaging data output from a Conventional Neural Network (CNN) and a Convolutional Neural Network (CNN), identifies an individual's health status, and predicts diseases according to said health status. Claim 7 In claim 6, the above analysis model predicts diseases including diabetes, urinary tract infection, and bladder stones from 10 types of test strips, and calculates the recognition rate and accuracy of the 10 types of test item result values ​​and the expected diseases, a urine test data analysis device. Claim 8 A urine test data analysis device according to claim 7, wherein the above 10 types of test strips include test strips for pH, glucose, protein, ketones, blood, leukocytes, nitrite, bilirubin, urobilinogen, and specific gravity. Claim 9 A method for analyzing data using an artificial intelligence-based urine test data analysis device, comprising: a step of collecting a learning data set including user biometric data, urine test strip data, and test result information, and implementing an analysis model through learning the collected learning data; a step in which the analysis model collects biometric data and a urine test strip image from a user terminal; and a step of detecting edges in the urine test strip image and determining the color of the internal region of the detected edges to derive a urine test result. Claim 10 A computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises one or more instructions, and when the instructions are executed by a computing device having one or more processors, the computing device performs the steps of: collecting a learning data set including user biometric data, urine test strip data, and test result information, and implementing an analysis model through learning the collected learning data; the analysis model performs the steps of: collecting biometric data and a urine test strip image from a user terminal; and detecting an edge in the urine test strip image and determining the color of the area inside the detected edge to derive a urine test result.

Images