Tibetan medicine-based urine diagnosis intelligent diagnosis method, diagnosis system and diagnosis instrument thereof

Abstract

The invention discloses a Tibetan medicine-based urine diagnosis intelligent diagnosis method, a diagnosis system thereof and a diagnosis instrument, which are characterized in that the Tibetan medicine-based urine diagnosis intelligent diagnosis method is characterized in that a Tibetan medicine-based urine diagnosis method is adopted, urine characteristics and disease syndrome classification is realized based on deep learning, and a Tibetan medicine-based RAG knowledge base is constructed based on a large language model; the functions of intelligent question answering, Tibetan traditional Chinese and western medicine disease control and the like are utilized, detection, diagnosis and service are integrated, and the method comprises the steps of data collection, urine character recognition, syndrome type judgment, diagnosis suggestion generation and the like. The system is composed of a management subsystem, an acquisition subsystem, a training subsystem, a diagnosis subsystem, an RAG subsystem and the like. The diagnostic apparatus comprises a detection device composed of a temperature sensor, a camera, an electronic tongue, an electronic nose module and the like. Compared with the prior art, the Tibetan medicine urine diagnosis and intelligent diagnosis integration system has the advantages that automatic statistics and intelligent analysis are carried out by utilizing metadata and expert marks, the system, the device and the method are integrated, and standardized and intelligent support and inheritance and application of Tibetan medicine are facilitated.

Description

Technical Field

[0001] This invention relates to the field of intelligent analysis and diagnostic technology, specifically to an intelligent diagnostic method, system and instrument for urine diagnosis based on Tibetan medicine. Background Technology

[0002] Traditional Tibetan medicine is an important supplement to the medical field, and Tibetan urine diagnosis is a crucial component of it, especially indispensable for the Tibetan population who rely on Tibetan medicine. Its lack can easily lead to patients receiving inappropriate treatment plans, increasing their financial burden. Currently, the field of urine testing lacks comprehensive digital technologies to examine the patterns of traditional Tibetan medicine urine diagnosis (the nine characteristics of "three times and nine diagnoses," or the odor, appearance, and details of urine samples), making the inheritance and application of Tibetan medicine in a digital context difficult. The traditional Tibetan medicine urine diagnosis method, which classifies urine characteristics according to the three times and nine methods, includes: 1) Color: light yellow, dark yellow, light red, dark red, blue, white, orange, brown, black; 2) Foam: blue and large, yellow and small, dissipates quickly, like saliva, like a rainbow; 3) Odor: none, small, normal, food smell, foul smell; 4) Steam: small and short duration, small and long duration, sometimes large and sometimes small, large; 5) Floating matter: none, thick, thin; 6) Suspended matter: hair-like, cheese-like, horsehair-like, cloud-like, pus-like, fine sand-like, none; 7) Change time: changes begin before the heat dissipates, changes after cooling, changes only begin after the heat dissipates; 8) Change pattern: changes from the edge to the middle, changes from the thick middle to the edge, changes from the edge; 9) Swirling condition: thin, thick.

[0003] In medical diagnosis, especially in the field of urine testing, doctors and researchers often use various testing instruments to analyze urine samples in order to obtain various information about the sample. However, traditional urine testing devices lack the ability to perform multi-factor detection of urine. This means that relying on a single testing method, such as only detecting chemical indicators, fails to capture information about the urine sample's odor, appearance details, etc., leading to incomplete diagnostic results. The odor, color, and transparency of urine play a crucial role in disease diagnosis. Without adequate testing methods, misdiagnosis or missed diagnosis can easily occur, potentially resulting in inappropriate treatment plans for patients. This not only affects treatment outcomes and delays diagnosis but also increases the financial burden on patients, placing both physical and economic stress on them. Furthermore, urine diagnosis is a distinctive diagnostic system in Tibetan medicine, with a history spanning over a thousand years. However, the lack of standardization and intelligent support significantly limits its transmission and application.

[0004] In summary, the existing methods for collecting training samples for urine diagnosis mainly involve taking multi-view images at a fixed time. However, these methods suffer from problems such as the images failing to cover the entire observation process, failing to cover sufficient sample features, and lacking flexibility. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing an intelligent urine diagnosis method, system, and instrument based on Tibetan medicine. It employs the "three times and nine diagnoses" method of Tibetan medicine urine diagnosis to achieve standardized collection of eight characteristics of urine across three stages. Based on deep learning, it classifies urine characteristics with disease patterns and constructs a Tibetan medicine RAG knowledge base based on a large language model (building a vector knowledge base and intelligent question-answering system based on RAG technology and a large language model). Utilizing intelligent question-answering and comparison functions with Tibetan, Chinese, and Western medicine diseases, it realizes an intelligent urine diagnosis system integrating detection, diagnosis, and service. This effectively enhances the functionality and efficiency of the Tibetan medicine-specific diagnostic system, providing strong technical support for further promoting Tibetan medicine, and possesses good application prospects and development value.

[0006] The specific technical solution for achieving the objective of this invention is: an intelligent diagnostic method for urine based on Tibetan medicine, characterized by the following steps: Step 1: Data Acquisition Information data of images and video sequences of urine samples or vapor from a specified angle are collected from the patient. The urine sample includes nine characteristics: color, foam, odor, vapor, floating matter, suspended matter, time of change, mode of change, and swirling behavior. Step 2: Intelligent Recognition of Urine Characteristics The data collected in step 1 is input into the trained Tibetan medicine urine sample intelligent diagnostic model, and its output is the quantitative phenotypic metadata of the case urine sample. The Tibetan medicine urine sample intelligent diagnostic model is trained using urine sample phenotypic metadata, so that the model can learn to identify and distinguish complex Tibetan medicine urine phenotypic traits from limited labeled samples. Step 3: Tibetan Medicine Syndrome Differentiation Based on the quantitative trait metadata, and using the preset Tibetan medicine syndrome classification rules, at least one Tibetan medicine urine diagnosis syndrome corresponding to the case is determined; Step 4: Generating Auxiliary Diagnostic Suggestions The identified Tibetan medicine urine diagnosis syndrome is converted into a query vector by the RAG module, and a similarity search is performed in its vector knowledge base. The obtained knowledge fragments from Tibetan medicine monographs are combined into an enhanced prompt context, which is then input into the generative model to generate auxiliary diagnostic suggestion text based on Tibetan medicine theory and current case examples. Step 5: Output of diagnostic results Through the human-computer interaction interface of the management module, the quantitative trait metadata, Tibetan medicine urine diagnosis syndrome, and auxiliary diagnostic suggestions are integrated and output to form a complete diagnostic report.

[0007] The Tibetan medicine urine sample intelligent diagnostic model is constructed from a multimodal sample data acquisition module, a RAG module, a generative module, a standardized data management module, and a model training module. It is trained based on a few-shot learning algorithm. This algorithm employs a model framework based on meta-learning or metric learning, utilizing keyframe images and their corresponding metadata in the training dataset to train the model. This enables the model to learn to identify and distinguish complex Tibetan medicine urine characteristics from a limited set of labeled samples. The metadata in the training dataset undergoes multi-dimensional statistical analysis to uncover the inherent correlation between urine characteristics and parameters such as time and temperature. This correlation is then embedded as prior knowledge or used to optimize the model's decision-making logic. The multi-dimensional statistical analysis includes analyzing the probability of occurrence, duration, or evolutionary pattern of different urine characteristics with changes in time and temperature, and establishing a statistical probability model.

[0008] The data acquisition step includes images of steam at specified angles, including a top frontal view of the urine sample, at least one side view, and a side view of the steam; the quantitative morphological metadata in the urine morphology intelligent identification step includes numerical or categorical descriptions of one or more of the following: color, foam, odor, steam, floating matter, suspended matter, change time, change mode, and swirling condition; the preset Tibetan medicine morphology classification rules in the Tibetan medicine morphology determination step are rule-based reasoning models or a lightweight classifier.

[0009] A Tibetan medicine-based intelligent urine diagnosis system is characterized by comprising: a management subsystem, a data acquisition subsystem, a training subsystem, a diagnostic subsystem, and a RAG subsystem. These subsystems are interconnected and work collaboratively to form an intelligent diagnostic model for Tibetan medicine urine samples, generating preliminary diagnostic results based on phenotypic metadata. The management subsystem includes at least: patient identity information, medical history, and diagnostic records, as well as case management of urine sample images, videos, and related metadata. Case management includes: user registration, login, and access control, and coordination of data flow and operations between subsystems. The data acquisition subsystem consists of an image acquisition module and an odor acquisition module, acquiring multimodal urine sample data and transmitting it to the training, diagnostic, and management subsystems for unified management. The training subsystem tags the images or videos of urine samples acquired by the image acquisition module with metadata and automatically incorporates the odor data acquired by the odor acquisition module into the metadata, based on the images or videos of the urine samples and metadata. The data is used to train a Tibetan medicine urine sample intelligent diagnostic model. The trained model is then tested and evaluated to assess its diagnostic effectiveness. The RAG subsystem includes a knowledge base construction module, a document processing module, and a diagnostic enhancement module. The knowledge base construction module constructs a knowledge base from accessed multi-source Tibetan medicine monographs and electronic documents. The document processing module parses and preprocesses the electronic documents, including text extraction, word segmentation, noise reduction, and standardization, to form structured text data. Based on natural language processing technology, the structured text data is transformed into high-dimensional vectors to construct a searchable vector knowledge base. The diagnostic enhancement module transforms the case's urine morphology metadata and preliminary diagnostic results into query vectors, performs similarity searches in the vector knowledge base, and combines the retrieved Tibetan medicine knowledge fragments with the current case information to form an enhanced context. This enhanced context is then input into the trained Tibetan medicine urine sample intelligent diagnostic model for reasoning, generating auxiliary diagnostic suggestions based on Tibetan medicine theory. These suggestions are then returned to the user interface of the diagnostic subsystem or management subsystem for final decision-making reference.

[0010] A smart urine diagnostic instrument based on Tibetan medicine includes a diagnostic instrument body and a detection space formed by an openable diagnostic instrument body cover. The diagnostic instrument body body contains a container tray, a fixing component, a cooling fan, a fan cover, and LED bracket lights. One side of the inner wall has a three-port socket, a USB connector, and a first and second camera for image capture. The other side of the inner wall has a fixing bracket. The container tray has multiple slots for placing urine sample cups and an electronic tongue module. An electronic nose module is fixedly connected to the fixing bracket. The fan cover is located on the top of the diagnostic instrument body cover. The LED bracket lights are located on both sides of the inner wall of the diagnostic instrument body, with two sets of LED bracket lights arranged symmetrically. The diagnostic instrument body cover contains a cooling fan and a temperature sensor, with the temperature sensor located at half the length of the diagnostic instrument body cover and close to one set of LED bracket lights. LED spotlights are installed on the fixing component.

[0011] The illumination areas of the first camera and the second camera are intersecting and set at an angle.

[0012] The main body of the diagnostic instrument box is made of non-transparent material. A handle is provided on one side of the outer wall, and a background plate is provided on one side of the inner wall below the fixed bracket to enhance the contrast of image acquisition. The background plate has at least two different colored background areas to enhance the contrast of image acquisition.

[0013] The installation height of the fastener is greater than the height of the urine sample cup.

[0014] Compared with the prior art, the present invention has the following beneficial technical effects and significant technical progress: 1) Multimodal data acquisition to achieve diagnosis of steam and odor properties. The traditional Tibetan medicine urine diagnosis theory of "three times and nine diagnoses" involves nine sample characteristics: color, foam, odor, steam, floating matter, suspended matter, time of change, mode of change, and swirling behavior. Observing the characteristics of urine at different stages (color, foam, odor, and steam) is particularly important for the accuracy of traditional Tibetan medicine urine diagnosis. Currently, no work can comprehensively cover the detection of sample characteristics, especially steam, which is one of the key characteristics, and there is no feasible intelligent detection method. This invention designs a steam-view imaging method for samples and acquires gustatory information, and uses electronic tongue, electronic nose, and other devices to realize the related device design, which is of great importance to improving the completeness of intelligent urine diagnosis.

[0015] 2) The video capture method is used, which contains more complete urine sample information. Traditional Tibetan medicine's urine diagnosis is based on the "Three Times, Nine Diagnoses" theory, observing the urine sample as a complete and continuous process from the hot urine stage to the cool urine stage. Current methods for collecting training samples for urine diagnosis primarily involve simultaneously capturing multi-view images at a fixed moment. Problems with this method include, but are not limited to, the inability to capture the entire observation process, insufficient sample features, and a lack of flexibility. The training method of this invention involves capturing video and using expert opinions for differentiated labeling, strictly controlling and acquiring metadata such as time and temperature throughout the entire process, and then automatically statistically and intelligently analyzing the acquired metadata. The advantage of this training method is that the acquired image set is distributed throughout the entire effective urine diagnosis process, containing sufficient sample characteristic information, rather than images from a single moment.

[0016] 3) Metadata statistical analysis helps to continuously optimize intelligent urine collection strategies. This invention can automatically acquire and learn the temperature or time patterns of various urinary diagnostic characteristics at different stages by utilizing metadata and expert tags. The learning results of these patterns can be used to adjust the automated sample collection strategy of the intelligent diagnostic process in real time. This work is beneficial to the standardization and accuracy of the intelligent urinary diagnostic process.

[0017] 4) Integrated System This invention is a Tibetan medicine intelligent urine diagnosis professional integrated system that integrates systems, devices, training and diagnostic methods, which is conducive to standardization and intelligent support, as well as the inheritance and application of Tibetan medicine. Attached Figure Description

[0018] Figure 1 Flowchart of intelligent urine diagnosis method; Figure 2 This is a schematic diagram of the structure of an intelligent urine diagnosis system; Figure 3 This is an image of the intelligent urine diagnostic instrument. Figure 4 A schematic diagram showing the installation of a smart urine diagnostic instrument; Figure 5 This is the main view of the intelligent urine diagnostic instrument. Figure 6 This is a schematic diagram of the structure of an intelligent urine diagnostic instrument; Figure 7 for Figure 4 A schematic diagram of node a. Detailed Implementation

[0019] The technical solution and effects of the present invention will be further described in detail below through specific embodiments. Example

[0020] See Figure 1A smart urine diagnosis method based on Tibetan medicine includes constructing a Tibetan medicine urine sample smart diagnosis model consisting of a multimodal sample data acquisition module, a RAG module, a generative module, a standardized data management module, and a model training module, to achieve smart diagnosis of case urine. The smart urine diagnosis method specifically includes the following steps: 1) Data Collection Images and video sequences of urine samples or vapor from a specified angle are collected, along with taste information data of the urine sample received by a taste sensor. The specified angles include the top frontal view of the urine sample, at least one side view, and the side view of the vapor.

[0021] 2) Intelligent recognition of urine characteristics The collected images, video sequences, or taste information data are input into the trained Tibetan medicine urine sample intelligent diagnostic model to obtain the quantitative morphological metadata of the case urine sample output by the model. The quantitative morphological metadata includes one or more numerical or categorical descriptions of color, foam, odor, steam, floating matter, suspended matter, change time, change mode, and swirling situation.

[0022] 3) Tibetan Medicine Syndrome Differentiation Based on the quantitative trait metadata obtained from the intelligent urine trait identification step, at least one Tibetan medicine urine diagnosis syndrome corresponding to the case is determined according to the preset Tibetan medicine syndrome classification rules. The preset Tibetan medicine syndrome classification rules are rule-based reasoning models or a lightweight classifier.

[0023] 4) Generation of auxiliary diagnostic suggestions The Tibetan medicine urine diagnosis syndrome determined in the Tibetan medicine syndrome determination step is input into the RAG module, which converts the Tibetan medicine urine diagnosis syndrome into a query vector and performs a similarity search in its vector knowledge base to obtain relevant Tibetan medicine monograph knowledge fragments. The Tibetan medicine urine diagnosis syndrome and the retrieved knowledge fragments are combined into an enhanced prompt context, which is then input into the generative model. The model receives the auxiliary diagnostic suggestion text output by the generative model, which is based on Tibetan medicine theory and the current case symptom syndrome. The information input into the RAG module includes some or all of the quantitative trait metadata obtained in the urine trait intelligent recognition step.

[0024] 5) Output of diagnostic results Through the human-computer interaction interface of the management subsystem, the quantitative trait metadata, Tibetan medicine urine diagnosis syndrome, and auxiliary diagnostic suggestion text are integrated and output to form a complete diagnostic report.

[0025] The training of the Tibetan medicine urine sample intelligent diagnostic model includes: few-shot learning training and statistical regularity analysis training. The few-shot learning training employs a model framework based on meta-learning or metric learning, utilizing keyframe images and their corresponding metadata. The model is trained using a few-shot learning algorithm, enabling it to learn to identify and distinguish complex Tibetan medicine urine diagnostic traits from a limited set of labeled samples. The output of the trained model is the prediction result of the urine diagnostic trait metadata for the input urine sample image. The statistical regularity analysis training specifically includes: analyzing the probability, duration, or evolution pattern of different urine diagnostic traits changing with time and temperature, and establishing a statistical probability model. This statistical regularity analysis training performs multi-dimensional statistical analysis on the metadata in the training dataset, uncovering the inherent correlation between urine diagnostic traits and parameters such as time and temperature, and embedding this statistical regularity as prior knowledge or using it to optimize the model's decision logic. The output of the model trained in these steps is the prediction result of the urine diagnostic trait metadata for the input urine sample image. Example

[0026] See Figure 2A Tibetan medicine-based intelligent urine diagnosis system includes: a management subsystem, a data acquisition subsystem, a training subsystem, a diagnostic subsystem, and a RAG subsystem. These subsystems are interconnected and work collaboratively to form an intelligent Tibetan medicine urine sample diagnosis model for case urine samples, generating preliminary diagnostic results based on phenotypic metadata. The management subsystem manages basic case information, including at least patient identity information, medical history, and diagnostic records; it also manages urine sample images, videos, and related metadata. Case management includes user registration, login, and access control, as well as connecting and coordinating the data acquisition and training subsystems. The system manages data flow and operations between the system, diagnostic subsystem, and RAG subsystem. The acquisition subsystem, composed of an image acquisition module and an odor acquisition module, transmits the acquired multimodal case urine data to the training subsystem or diagnostic subsystem for further processing, and manages the acquired data uniformly through the management subsystem. The training subsystem tags the images or videos of urine samples acquired by the image acquisition module with metadata and automatically incorporates the odor data acquired by the odor acquisition module into the metadata. Based on the images or videos of the urine samples and the metadata, the intelligent diagnostic model for Tibetan medicine urine samples is trained. The trained model is then verified and tested. To evaluate the diagnostic effectiveness of the model; the diagnostic subsystem is used to call the acquisition subsystem to collect image, video, and odor data of urine samples from cases; it calls the Tibetan medicine urine sample intelligent diagnostic model to perform intelligent diagnosis on the urine sample images or videos of cases, outputs a series of phenotypic metadata of the cases, and generates preliminary diagnostic results based on the phenotypic metadata; the RAG subsystem includes: a knowledge base construction module, a document processing module, and a diagnostic enhancement module; the knowledge base construction module constructs a knowledge base from the accessed multi-source Tibetan medicine monograph electronic documents; the document processing module parses and preprocesses the electronic documents, including: text extraction, word segmentation, The system denoises and standardizes data to form structured text data. Based on natural language processing technology, the structured text data is transformed into high-dimensional vectors to construct a searchable vector knowledge base. The diagnostic enhancement module transforms the case's urine morphology metadata and preliminary diagnostic results into query vectors, performs similarity retrieval in the vector knowledge base, and combines the retrieved Tibetan medicine knowledge fragments with the current case information to form an enhanced context. This context is then input into the trained Tibetan medicine urine sample intelligent diagnostic model for reasoning, generating auxiliary diagnostic suggestions based on Tibetan medicine theory. These suggestions are then returned to the user interface of the diagnostic subsystem or management subsystem for final decision-making reference.

[0027] The intelligent urine diagnosis system of the intelligent diagnostic system includes the following steps: multimodal sample data acquisition, expert labeling and metadata construction, standardized data management, and model training. The multimodal sample data acquisition involves collecting time-series video data of urine samples from cases. Simultaneously, a temperature sensing module records the timestamp and real-time temperature data corresponding to the time-series video data. The expert labeling and metadata construction receives analysis and labeling information input by Tibetan medicine experts based on the time-series video data through the human-computer interaction interface of the management subsystem. The labeling information includes at least: a diagnostically representative keyframe image extracted from the time-series video data, the video timestamp corresponding to the keyframe image, the real-time temperature data recorded at that moment, and the urine diagnosis characteristic analysis results made by experts based on Tibetan medicine theory for the keyframe image. The timestamp, temperature data, and urine diagnosis characteristic analysis results corresponding to the keyframe image are collectively defined as the metadata of the keyframe image. The urine diagnosis characteristic analysis results include, but are not limited to, qualitative or quantitative descriptions of one or more of the following characteristics of urine: color, foam, odor, vapor, floating matter, suspended matter, change time, change mode, and swirling behavior. Example

[0028] See Figures 3-4A smart urine diagnostic instrument based on Tibetan medicine includes a main body 11 and an openable cover 12. The main body 11 and the cover 12 form a detection space. The main body 11 has a feed inlet with a container tray 13. One side of the inner wall of the main body 11 has a three-port socket 14, a USB interface 15, and a first camera 16 and a second camera 17 for image capture. An electronic tongue module 18 is mounted on the container tray 13. A fixing member 19 is located inside the main body 11, and an LED spotlight 21 is mounted on the fixing member 19. A cooling fan 22 is located at the bottom of the cover 12, and a fixing bracket 23 is located on one side of the main body 11. Specifically, the detection space formed by the main body 11 and the cover 12 provides a stable environment for urine sample testing, reducing interference from external factors. The feed inlet facilitates sample placement, and the container tray 13 holds the sample container. The three-port socket 14 provides power to the internal electrical equipment of the instrument, allowing connection to any power-consuming equipment (such as the first camera 16, the second camera 17, etc.). The USB interface 15 facilitates data transmission and connection to external devices. The first camera 16 and the second camera 17 are used to acquire images of the urine sample from different angles, providing a data basis for subsequent analysis of the urine's color, transparency, and other appearance characteristics. The electronic tongue module 18 analyzes and detects the taste characteristics of the urine, obtaining information related to its chemical composition. The fixing component 19 secures the LED spotlight 21, which illuminates the urine sample, allowing the camera to capture clearer images. The cooling fan 22 reduces the heat generated inside the instrument during operation, ensuring stable operation of all components. The mounting bracket 23 supports and secures other detection components, ensuring their position remains accurate and facilitating detection.

[0029] The main body 11 of the diagnostic instrument is made of a non-transparent material. A background plate 24 for enhancing image acquisition contrast is provided on one side of the inner wall of the main body 11, and the background plate 24 is located below the fixed bracket 23. Specifically, the non-transparent material of the main body 11 avoids interference from external light during the detection process, ensuring the stability and controllability of the ambient light during detection. The background plate 24 provides a clear contrast with the urine sample, helping the first camera 16 and the second camera 17 to better capture the detailed features of the urine sample, improving the quality of image acquisition and thus enhancing the accuracy of subsequent image analysis. In this device, the background plate 24 is two colors and is located opposite the first camera 16 and the second camera 17. The background plate 24 is installed below the fixed bracket 23, which neither affects the normal operation of the fixed bracket 23 and its components, nor does it interfere with the normal operation of the fixed bracket 23, while providing a suitable background for urine sample image acquisition.

[0030] The container tray 13 is equipped with a urine sample cup 25. A mounting bracket 23, with its carrying end positioned above the urine sample cup 25, has a third camera 101 at one end and an electronic nose module 26 fixedly connected to it. Specifically, the urine sample cup 25 holds the urine sample and serves as the carrier for the entire testing process. The carrying end of the mounting bracket 23, positioned above the urine sample cup 25, provides a suitable detection position for the third camera 101 and the electronic nose module 26. The third camera 101 can capture images of the urine sample from above, obtaining image information from the top view of the urine sample. This image is then combined with images captured by the first camera 16 and the second camera 17 to achieve comprehensive image recording of the urine sample. The electronic nose module 26 can detect the odor information emitted by the urine. By analyzing the odor components, it assists in determining the health status of the urine sample, further enriching the testing dimensions and providing more evidence for disease diagnosis.

[0031] The diagnostic instrument box cover 12 has a fan cover 27 at the top center. LED bracket lights 28 for illuminating the inside of the box are located on both sides of the inner wall of the main body 11, and the two sets of LED bracket lights 28 are symmetrically arranged. Specifically, the fan cover 27 protects the cooling fan 22 from foreign objects entering and affecting its normal operation; it also acts as a guide, making the airflow smoother. The two symmetrically arranged LED bracket lights 28 provide uniform and sufficient light to the testing space, ensuring that the entire inside of the box is illuminated, avoiding blind spots, and ensuring that the images captured by the first camera 16, the second camera 17, and the third camera 101 are clear and shadow-free, which is beneficial for analyzing the appearance characteristics of urine samples.

[0032] See Figures 4-7 The container tray 13 has multiple slots, and the electronic tongue module 18 and the urine sample cup 25 are both installed in the slots. The irradiation areas of the first camera 16 and the second camera 17 intersect each other and are set at an angle.

[0033] Specifically, the slots on the container tray 13 position the electronic tongue module 18 and the urine sample cup 25, ensuring their fixed positions during each test and improving test repeatability. Once the electronic tongue module 18 and the urine sample cup 25 are engaged in the slots, their positions are stable and they are not easily shaken, ensuring a smooth testing process. The illumination areas of the first camera 16 and the second camera 17 intersect and are set at an angle, allowing the two cameras to acquire image information of the urine sample from different angles. These images complement each other, providing a more comprehensive reflection of the urine sample's appearance characteristics, which is helpful for subsequent multi-angle and more detailed analysis of the urine sample, thus improving the reliability of the test results.

[0034] A temperature sensor 29 is located on one side of the bottom of the diagnostic instrument cover 12, at approximately half the length of the cover and near one of the LED support lights 28. Specifically, the temperature sensor 29 monitors the temperature within the testing space. Its placement at half the length of the bottom of the cover 12, near one of the LED support lights 28, allows for effective sensing of temperature changes throughout the testing space and timely detection of the impact of heat generated by the LED support lights 28 on the ambient temperature. Temperature monitoring ensures that the testing environment temperature remains within a suitable range, preventing excessively high or low temperatures from affecting the accuracy of test results. It also allows for the timely detection of problems such as abnormal heat dissipation, ensuring stable instrument operation.

[0035] The fixing component 19 and the LED spotlight 21 are located above the feed inlet, and the installation height of the fixing component 19 is greater than the height of the urine sample cup 25. A handle 31 is provided on one side of the outer wall of the diagnostic instrument body 11. Specifically, the fixing component 19 and the LED spotlight 21 are located above the feed inlet, so that after the urine sample cup 25 is placed in, the LED spotlight 21 can immediately illuminate the urine sample, facilitating image acquisition by the camera. The installation height of the fixing component 19 is greater than the height of the urine sample cup 25, ensuring that the LED spotlight 21 has sufficient space to illuminate all parts of the urine sample, avoiding shadows due to insufficient height. The handle 31 on one side of the outer wall of the diagnostic instrument body 11 facilitates the operator's handling and movement of the entire diagnostic instrument, improving the convenience of instrument use and making it easier to transfer the instrument between different testing locations.

[0036] The intelligent urine diagnostic instrument is used as follows: When using the device, the operator first moves the instrument to the testing position using handle 31, opens the diagnostic instrument cover 12, and places the urine sample cup 25, containing the urine sample, into the slot of the container tray 13 through the feed inlet. Simultaneously, the electronic tongue module 18 is also inserted into the slot to ensure a fixed position. At this time, the urine sample cup 25 serves as the carrier for testing, providing a sample for subsequent testing. Next, the relevant equipment is connected. The first camera 16, second camera 17, and other electrical devices are powered using a three-port socket 14, and data transmission and connection to external devices are achieved through a USB interface 15. After the device is turned on, the first camera 16 and second camera 17 acquire images of the urine sample from different angles, while the third camera 101 acquires a top-view image of the urine sample. These images complement each other, providing data for analyzing the appearance characteristics of the urine, such as color and transparency. Simultaneously, the electronic tongue module 18 analyzes the gustatory characteristics of the urine to obtain chemical composition information, and the electronic nose module 26 detects the odor components of the urine to assist in judging the health status of the sample. The LED spotlight 21 on the fixing component 19 illuminates the urine sample, and together with the LED bracket lights 28 symmetrically arranged on both sides of the inner wall, it ensures that there are no blind spots in the box, so that the camera can capture clear images.

[0037] The cooling fan 22, protected and guided by the fan cover 27, reduces the heat generated inside the instrument due to equipment operation, ensuring stable operation of all components. The temperature sensor 29 monitors the temperature of the detection space in real time, preventing temperature from affecting the detection results. The non-transparent diagnostic instrument housing 11 and background plate 24 reduce external light interference and enhance image acquisition contrast.

[0038] The above is merely a further description of the present invention and is not intended to limit the scope of this patent. Any equivalent implementation of the present invention should be included within the scope of the claims of this patent.

Claims

1. A smart diagnostic method for urine based on Tibetan medicine, characterized in that, The method specifically includes the following steps: Step 1: Data Acquisition Information data of images and video sequences of urine samples or vapor from a specified angle are collected from the patient. The urine sample includes nine characteristics: color, foam, odor, vapor, floating matter, suspended matter, time of change, mode of change, and swirling behavior. Step 2: Intelligent Recognition of Urine Characteristics The data collected in step 1 is input into the trained Tibetan medicine urine sample intelligent diagnostic model, and its output is the quantitative phenotypic metadata of the case urine sample. The Tibetan medicine urine sample intelligent diagnostic model is trained using urine sample phenotypic metadata, so that the model can learn to identify and distinguish complex Tibetan medicine urine phenotypic traits from limited labeled samples. Step 3: Tibetan Medicine Syndrome Differentiation Based on the quantitative trait metadata, and using the preset Tibetan medicine syndrome classification rules, at least one Tibetan medicine urine diagnosis syndrome corresponding to the case is determined; Step 4: Generating Auxiliary Diagnostic Suggestions The identified Tibetan medicine urine diagnosis syndrome is converted into a query vector by the RAG module, and a similarity search is performed in its vector knowledge base. The obtained knowledge fragments from Tibetan medicine monographs are combined into an enhanced prompt context, which is then input into the generative model to generate auxiliary diagnostic suggestion text based on Tibetan medicine theory and current case examples. Step 5: Output of diagnostic results Through the human-computer interaction interface of the management module, the quantitative trait metadata, Tibetan medicine urine diagnosis syndrome, and auxiliary diagnostic suggestions are integrated and output to form a complete diagnostic report.

2. The intelligent diagnostic method for urine based on Tibetan medicine according to claim 1, characterized in that, The Tibetan medicine urine sample intelligent diagnostic model is constructed from a multimodal sample data acquisition module, a RAG module, a generative module, a standardized data management module, and a model training module. It is trained based on a few-shot learning algorithm. This algorithm employs a model framework based on meta-learning or metric learning, utilizing keyframe images and their corresponding metadata in the training dataset to train the model. This enables the model to learn to identify and distinguish complex Tibetan medicine urine characteristics from a limited set of labeled samples. The metadata in the training dataset undergoes multi-dimensional statistical analysis to uncover the inherent correlation between urine characteristics and parameters such as time and temperature. This correlation is then embedded as prior knowledge or used to optimize the model's decision-making logic. The multi-dimensional statistical analysis includes analyzing the probability of occurrence, duration, or evolutionary pattern of different urine characteristics with changes in time and temperature, and establishing a statistical probability model.

3. The intelligent diagnostic method for urine based on Tibetan medicine according to claim 1, characterized in that, The data acquisition step includes images of steam at specified angles, including a top frontal view of the urine sample, at least one side view, and a side view of the steam; the quantitative morphological metadata in the urine morphology intelligent identification step includes numerical or categorical descriptions of one or more of the following: color, foam, odor, steam, floating matter, suspended matter, change time, change mode, and swirling condition; the preset Tibetan medicine morphology classification rules in the Tibetan medicine morphology determination step are rule-based reasoning models or a lightweight classifier.

4. A diagnostic system constructed using the intelligent urine diagnosis method based on Tibetan medicine as described in claim 1, characterized in that, This intelligent urine diagnosis system comprises a management subsystem, a data acquisition subsystem, a training subsystem, a diagnostic subsystem, and a RAG subsystem. These subsystems are interconnected and work collaboratively to form an intelligent diagnostic model for Tibetan medicine urine samples, generating preliminary diagnostic results based on phenotypic metadata. The management subsystem includes at least: patient identity information, medical history, and diagnostic records, as well as case management of urine sample images, videos, and related metadata. Case management includes: user registration, login, and access control, and coordinating data flow and operations between the subsystems. The data acquisition subsystem consists of an image acquisition module and an odor acquisition module, collecting multimodal urine sample data and transmitting it to the training, diagnostic, and management subsystems for unified management. The training subsystem tags the images or videos of urine samples collected by the image acquisition module with metadata and automatically incorporates the odor data collected by the odor acquisition module into the metadata. Based on the images or videos of the urine samples and the metadata, it performs intelligent diagnosis of Tibetan medicine urine samples. The intelligent diagnostic model is trained, and the trained model is tested and evaluated to assess its diagnostic effectiveness. The RAG subsystem includes a knowledge base construction module, a document processing module, and a diagnostic enhancement module. The knowledge base construction module constructs a knowledge base from accessed multi-source Tibetan medicine monographs and electronic documents. The document processing module parses and preprocesses the electronic documents, including text extraction, word segmentation, noise reduction, and standardization, to form structured text data. Based on natural language processing technology, the structured text data is transformed into high-dimensional vectors to construct a searchable vector knowledge base. The diagnostic enhancement module transforms the case's urine morphology metadata and preliminary diagnostic results into query vectors, performs similarity searches in the vector knowledge base, and combines the retrieved Tibetan medicine knowledge fragments with the current case information to form an enhanced context. This enhanced context is then input into the trained Tibetan medicine urine sample intelligent diagnostic model for reasoning, generating auxiliary diagnostic suggestions based on Tibetan medicine theory. These suggestions are then returned to the user interface of the diagnostic subsystem or management subsystem for final decision-making reference.

5. A diagnostic instrument constructed using the intelligent urine diagnosis method based on Tibetan medicine as described in claim 1, comprising a diagnostic instrument body (11) and a detection space formed by an openable and closable diagnostic instrument cover (12), characterized in that, The main body (11) of the diagnostic instrument box is equipped with a container tray (13), a fastener (19), a cooling fan (22), a fan cover (27), and an LED bracket light (28). One side of the inner wall is equipped with a three-port socket (14), a USB terminal (15), and a first camera (16) and a second camera (17) for detection and imaging. The other side of the inner wall is equipped with a fixed bracket (23). The container tray (13) is equipped with multiple slots for placing a urine sample cup (25) and an electronic tongue module (18). An electronic nose is fixedly connected to the fixed bracket (23). Module (10); The fan cover (27) is set on the top of the diagnostic instrument box cover (12); The LED bracket lights (28) are set on both sides of the inner wall of the diagnostic instrument box body (11), and the two sets of LED bracket lights (28) are symmetrically arranged; The diagnostic instrument box cover (12) is provided with a cooling fan (22) and a temperature sensor (29), and the temperature sensor (29) is located at half the length of the diagnostic instrument box cover (12) and close to one of the sets of LED bracket lights (28); The fixing part (19) is provided with an LED spotlight (21).

6. The intelligent urine diagnostic instrument based on Tibetan medicine according to claim 5, characterized in that, The illumination areas of the first camera (16) and the second camera (17) are intersecting and set at an angle.

7. The intelligent urine diagnostic instrument based on Tibetan medicine according to claim 5, characterized in that, The main body (11) of the diagnostic instrument box is made of non-transparent material. A handle (31) is provided on one side of the outer wall, and a background plate (24) for enhancing the contrast of image acquisition is provided on one side of the inner wall and below the fixed bracket (23). The background plate (24) has at least two different background areas to enhance the contrast of image acquisition.

8. The intelligent urine diagnostic instrument based on Tibetan medicine according to claim 5, characterized in that, The installation height of the fastener (19) is greater than the height of the urine sample cup (25).

Images