Auxiliary diagnosis and treatment system based on traditional Chinese medicine large model

Abstract

The invention belongs to the technical field of traditional Chinese medicine auxiliary diagnosis and treatment, and particularly discloses an auxiliary diagnosis and treatment system based on a traditional Chinese medicine large model. Therefore, a tongue picture change curve is generated by utilizing tongue picture characteristics under a time sequence, a syndrome type evolution path is output by utilizing a tongue picture-syndrome type model, clinical indexes of a patient in the time sequence at the same period are associated with the syndrome type evolution path to form a syndrome type-disease incidence matrix, and a prescription compatibility scheme is recommended according to the syndrome type-disease incidence matrix. According to the method, the treatment scheme is ensured to be based on the syndrome type, the specific pathological mechanism and clinical manifestation are fully considered, the treatment scheme is prevented from being disjointed with the actual illness state, more accurate and personalized traditional Chinese medicine auxiliary diagnosis and treatment are achieved, in addition, automatic data processing and analysis reduce manual intervention, the diagnosis and treatment efficiency is improved, and medical resources are saved.

Description

Technical Field

[0001] The present invention belongs to the technical field of auxiliary diagnosis and treatment of traditional Chinese medicine, and specifically discloses an auxiliary diagnosis and treatment system based on a large model of traditional Chinese medicine. Background Art

[0002] In modern society, with the significant changes in lifestyle, factors such as long-term sitting, unreasonable diet structure and high stress have led to the increasing prevalence of chronic diseases and sub-health conditions, which has prompted the growth of demand for TCM-assisted diagnosis and treatment, because TCM, with its holistic concept and dialectical treatment method, can provide personalized comprehensive assessment and treatment plans, especially in chronic disease management and clinical diagnosis.

[0003] In the TCM diagnosis and treatment system, comprehensive assessment of the patient's health status through observation, auscultation, inquiry, and palpation is a key step in syndrome differentiation and treatment. The correlation analysis between the results of the inquiry and the syndrome type provides an important basis for the subsequent prescription compatibility. With the development of information technology, modern technologies such as artificial intelligence, big data, and cloud computing have gradually been integrated into the field of TCM. By building a syndrome type model based on big data, the efficiency of diagnosis and treatment has been significantly improved.

[0004] There are existing technical solutions for TCM-assisted diagnosis and treatment under the syndrome model of big data, such as the system proposed by the Chinese invention patent with publication number CN119170207A, which uses four diagnostic instruments to collect pulse and tongue images of patients, and inputs the preset TCM syndrome differentiation model to obtain the patient's syndrome type, thereby generating corresponding prescription information. The system assists doctors in collecting four diagnostic information and prescribing syndromes through the interaction between smart terminals and four diagnostic instruments, thereby improving the efficiency of diagnosis and treatment.

[0005] Another example is that the invention patent with publication number CN112992344A proposes an intelligent auxiliary diagnosis and treatment system for traditional Chinese medicine. It collects the patient's symptom information (such as tongue coating and pulse) through medical questionnaires, converts it into corresponding syndrome types, selects the corresponding prescription from the prescription library, and then determines the final prescription based on the drug addition and subtraction rules, effectively improving the efficiency of traditional Chinese medicine diagnosis and treatment.

[0006] Although the above two schemes have significantly improved the efficiency of diagnosis and treatment by utilizing the TCM model of syndrome and symptoms, they have failed to fully consider the complex relationship between syndrome and pathological mechanism. TCM syndrome is not only a high-level summary of disease symptoms, but also a profound revelation of the internal causes and development trends of the disease. Each syndrome reflects a specific pathological mechanism, which determines the complex synergistic relationship between different syndromes and pathological states. Under this complex synergistic relationship, relying solely on syndromes for prescription compatibility may ignore the patient's detailed pathological information, resulting in a lack of necessary clinical pathological support for the treatment plan, which can easily cause the treatment plan to be out of touch with the actual condition and affect the final treatment effect. Summary of the invention

[0007] In view of this, the present invention aims to propose an auxiliary diagnosis and treatment system based on the TCM big model. In the TCM diagnosis and treatment system, tongue image analysis, as the core link of the four diagnoses, is a key indicator reflecting the syndrome. By focusing on the tongue image-syndrome model and taking the patient's clinical indicators into consideration, a more comprehensive syndrome-clinical indicator association model is established, thereby achieving more accurate and personalized TCM auxiliary diagnosis.

[0008] The purpose of the present invention can be achieved through the following technical solutions: an auxiliary diagnosis and treatment system based on a large model of traditional Chinese medicine, comprising: a tongue image data acquisition module: receiving tongue images uploaded by patients regularly at the same time period, and triggering tongue image enhancement processing based on an image quality detection feedback mechanism of tongue clarity and color uniformity.

[0009] Tongue feature extraction module: The enhanced tongue image is divided into three tongue regions: tongue tip, tongue middle, and tongue root, and the tongue features of each tongue region are extracted, specifically the tongue color HSV value, the percentage of pixels with thick fur, and the crack density.

[0010] Tongue image change analysis module: The tongue image features of regularly uploaded tongue images are stored in time series to generate a tongue image change curve containing the tongue color, coating thickness, and cracks corresponding to each tongue area.

[0011] TCM big model prediction module: import the tongue image change curve into the constructed tongue image-syndrome model to analyze the syndrome type under the time series, and output the syndrome type evolution path.

[0012] Diagnosis and treatment association decision module: The clinical indicators of the patient in the same time series are associated with the syndrome evolution path to form a syndrome-disease association matrix, and the prescription combination scheme is recommended based on it.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. The present invention imports the tongue image change curve generated by the tongue image characteristics corresponding to the patient's tongue image in the time series into the tongue image-syndrome model to output the syndrome evolution path, and combines it with the patient's clinical indicators to construct a syndrome-disease association matrix, and recommends personalized prescription combinations based on this, ensuring that the treatment plan is not only based on the syndrome, but also fully considers the specific pathological mechanism and clinical manifestations, avoiding the disconnection between the treatment plan and the actual condition, and achieving more accurate and personalized Chinese medicine-assisted diagnosis and treatment.

[0014] 2. Before using the tongue image uploaded by the patient for feature extraction, the present invention triggers image enhancement processing through an image quality detection feedback mechanism based on tongue clarity and color uniformity, which can effectively improve low-quality images and help improve the accuracy of subsequent tongue feature extraction. In addition, this mechanism ensures that all tongue images for analysis have consistent quality standards, improving the reliability and comparability of time series analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0016] Figure 1 It is a schematic diagram of the module composition and data flow in the system of the present invention.

[0017] Figure 2 This is a flowchart for implementing the process of triggering tongue image enhancement processing in the present invention.

[0018] Figure 3 Schematic diagram of the evolution path of the syndrome type in the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] The present invention proposes an auxiliary diagnosis and treatment system based on a TCM big model, comprising a tongue image data acquisition module, a tongue image feature extraction module, a tongue image change analysis module, a TCM big model prediction module, and a diagnosis and treatment association decision module.

[0021] See also Figure 1 As shown, the modules are connected end to end, and the data flow relationship between the modules is as follows: the tongue image data acquisition module serves as the starting point of the system and provides raw data for subsequent modules.

[0022] The tongue image feature extraction module processes the tongue image provided by the tongue image data acquisition module and transmits the extracted features to the tongue image change analysis module.

[0023] The tongue image change analysis module analyzes the changes in tongue image over time and passes the results to the TCM large model prediction module.

[0024] The TCM big model prediction module receives and processes the data provided by the tongue pattern change analysis module, determines the patient's syndrome evolution path, and passes this information to the diagnosis and treatment-related decision-making module.

[0025] The diagnosis and treatment-related decision-making module receives the results of the TCM big model prediction module, integrates the patient's clinical indicators, and generates treatment recommendations after comprehensive analysis.

[0026] Specifically, the tongue image data acquisition module receives tongue images uploaded by patients at regular intervals and triggers tongue image enhancement processing based on an image quality detection feedback mechanism of tongue clarity and color uniformity.

[0027] It is important to know that the regular data collection mentioned above refers to the acquisition of tongue images at fixed intervals and cycles. This interval can be daily or other time frequencies set according to clinical needs, and the cycle can be one month, two months or longer. Through this regular data collection, it is possible to effectively monitor the changes in the patient's tongue image over time, providing continuous and reliable data support for the subsequent generation of tongue image change curves.

[0028] The same time period means that the patient selects the same time period when uploading tongue images every day. For example, upload tongue images between 8 and 10 a.m. every day. Since the physiological state of the human body may change at different times of the day, in order to reduce the impact of such daytime fluctuations on the data, selecting to collect tongue images at the same time period can ensure the consistency and stability of the data, thereby improving the reliability of the analysis results.

[0029] The combined effect of "regularly" and "same period" is to ensure the regularity and consistency of the collection of tongue image data, thereby providing reliable basic data for the TCM-assisted diagnosis and treatment system.

[0030] In the preferred implementation of the above scheme, see Figure 2 As shown, the image quality detection feedback mechanism based on the tongue clarity and color uniformity triggers the tongue image enhancement processing. See the following process: a noise preprocessing operation is performed on the received tongue image.

[0031] The above noise preprocessing operation is specifically as follows in one example: smoothing the image to remove high-frequency noise to prevent noise from being mistaken for edges. Noise preprocessing can reduce noise interference and improve the accuracy of edge detection.

[0032] The edge detection algorithm is used to extract the tongue edge pixels from the preprocessed image.

[0033] The purpose of extracting the edge pixels of the tongue through edge detection is that the edge pixels of the tongue are the dividing line between the tongue and the background. This operation aims to accurately identify the contour of the tongue and can reflect the edge sharpness and edge continuity. The edge sharpness and edge continuity are the core indicators for measuring image clarity, that is, the clarity of the tongue. Edge sharpness reflects the clarity of image details, while edge continuity reflects the integrity of the tongue contour in the image. The combination of the two can comprehensively evaluate the clarity of the tongue.

[0034] The edge sharpness and edge continuity were evaluated based on the tongue edge pixels.

[0035] As an implementation method of the above scheme, edge sharpness is evaluated by analyzing the rate of change of pixels on the edge of the tongue. High sharpness means clear edges and distinct boundaries, which helps to accurately extract tongue features.

[0036] In a specific example, the edge sharpness evaluation may be performed by calculating the average value or the maximum value of the gradient amplitude of the edge pixels as a quantitative index of the edge sharpness.

[0037] As another implementation of the above scheme, edge continuity is to evaluate the coherence of the edge pixels of the tongue, that is, whether the edge is complete and unbroken. Good edge continuity indicates that the tongue structure in the image is complete and can more truly reflect the actual state of the tongue.

[0038] In a specific example, the edge continuity assessment can be performed by calculating the ratio of the complete edge length to the total circumference of the tongue as a quantitative indicator of edge continuity.

[0039] It should be pointed out that the edge sharpness and edge continuity evaluated above are dimensionless data.

[0040] The tongue clarity is defined as the product of edge sharpness and edge continuity. Under this definition, the tongue clarity of the tongue image is obtained based on the evaluated edge sharpness and edge continuity.

[0041] It should be understood that the above definition of tongue clarity is based on a comprehensive consideration of edge sharpness and edge continuity. Specifically, edge sharpness focuses on the clarity of the edge, reflecting the sharpness of the tongue contour boundary details in the image; while edge continuity focuses on the integrity of the edge line, that is, whether the edge is coherent and unbroken. A single indicator is difficult to fully describe the tongue clarity, because the usability of an image with high sharpness but low continuity will be greatly reduced in practical applications. Therefore, combining these two indicators can more comprehensively reflect the overall quality of the tongue image.

[0042] The reason why edge sharpness and edge continuity are multiplied in the process of combining the two is that multiplication can effectively amplify the joint influence of the two indicators. If one of the indicators is low, such as edge continuity, the final tongue clarity will be significantly reduced even if the other indicator, such as edge sharpness, is high. This mechanism ensures that the tongue clarity will be high only when the edge is both clear and continuous.

[0043] Image segmentation technology is used to separate the tongue area from the background and generate a corresponding binary mask to ensure that the pixel value in the tongue area is marked as 1, while the background area is marked as 0.

[0044] It should be pointed out that the tongue area can be accurately identified and separated through image segmentation technology, eliminating interference from the background and other irrelevant parts.

[0045] It should be further pointed out that the binary mask of the tongue area and the background ensures that the color values ​​of all pixels in the tongue area are accurately extracted in the HSV color space without being interfered by the background, which is conducive to obtaining more reliable color uniformity evaluation results.

[0046] The tongue image is converted from RGB color space to HSV color space, and the mask generated above is applied to accurately extract the color values ​​of all pixels in the tongue area.

[0047] It is important to know that in the HSV color space, the H channel represents the hue, which describes the type of color, such as red, green or blue, and the S channel represents the saturation, which reflects the purity of the color, that is, the degree of gray in the color. The higher the saturation, the brighter the color; the lower the saturation, the closer the color is to gray. The V channel represents the brightness, which indicates the brightness of the color. These three channels describe different properties of color.

[0048] The above color space conversion is to better analyze color distribution. This is because the HSV color space is superior to the RGB color space in representing the differences in human visual perception of color.

[0049] The standard deviation of the pixel values ​​of each color channel is calculated independently, and the color uniformity is obtained by weighted summing of the standard deviations of each color channel.

[0050] The weight value of each color channel in the above can be set according to needs.

[0051] The tongue clarity and color uniformity of the tongue image are compared with the set standard thresholds respectively. If both meet the standard thresholds, the tongue image enhancement processing will not be triggered. Conversely, if any indicator fails to reach the standard threshold, the tongue image enhancement processing will be automatically triggered.

[0052] The thresholds for the above-mentioned tongue clarity and color uniformity can be set based on experimental data or experience.

[0053] It should be emphasized that the tongue clarity meets the standard threshold means that the tongue clarity is greater than or equal to the standard threshold, and the color uniformity is calculated based on the standard deviation of the color channel. The larger the standard deviation, the more uneven the color distribution and the existence of color inconsistency. Therefore, the color uniformity meets the standard threshold means that the color uniformity is less than or equal to the standard threshold.

[0054] It should be explained that tongue clarity and color uniformity are selected as quality indicators in the quality detection of tongue images because tongue clarity and color uniformity reflect the key aspects of image quality from different perspectives. Tongue clarity focuses on the structural information of the image, while color uniformity focuses on the color information. The combination of the two can provide a comprehensive and detailed image quality assessment.

[0055] In a further preferred implementation of the above scheme, the tongue image enhancement processing operation is as follows: the image enhancement processing rules are defined as follows: a) when the tongue body clarity of the tongue image does not meet the standard threshold, the tongue image is subjected to edge degeneration processing.

[0056] b) When the color uniformity of the tongue image does not meet the standard threshold, the tongue image is subjected to color correction processing.

[0057] When the enhancement processing of the tongue image is triggered, the image enhancement processing is performed using the above rules within a specified time window.

[0058] It should be understood that the setting of a specified time window when triggering the enhancement processing of tongue images is based on the following considerations: patients upload tongue images within a fixed time period. If there is no time window limit, the image enhancement processing is allowed to run indefinitely, and long-term image processing may cause a backlog in the processing queue, thereby delaying the response speed of the entire system; in addition, even after image enhancement processing, some images may still fail to meet the standard threshold. In this case, the patient needs to re-upload a new tongue image. In order to ensure the consistency and comparability of the data, the patient's tongue image should be uploaded within the same time period. If there is no time window limit, the time point of re-uploading may become uncertain, thereby affecting the consistency and reliability of the time series of subsequent analysis.

[0059] Applied to the above understanding, the specified time window should be shorter than the fixed time period for uploading tongue images.

[0060] After each image enhancement process, the tongue clarity and color uniformity of the enhanced image are recalculated and compared with the standard threshold again. If it still does not meet the standard, multiple iterative enhancements are performed according to the above rules until the standard threshold is met or the available processing time in the time window is exhausted.

[0061] If the tongue clarity and color uniformity of the tongue image still do not meet the standard threshold at the end of the specified time window, a retake instruction is automatically generated, prompting the user to retake the tongue image.

[0062] The present invention triggers image enhancement processing by using an image quality detection feedback mechanism based on tongue clarity and color uniformity before using the tongue image uploaded by the patient for feature extraction, which can effectively improve low-quality images and help improve the accuracy of subsequent tongue feature extraction. In addition, this mechanism ensures that all tongue images for analysis have consistent quality standards, improving the reliability and comparability of time series analysis.

[0063] The tongue feature extraction module is used to divide the enhanced tongue image into three tongue regions: tongue tip, tongue middle, and tongue root, and extract the tongue features of each tongue region, specifically the tongue color HSV value, the percentage of tongue coating thickness pixels, and the crack density.

[0064] It should be pointed out that the above-mentioned extraction of tongue features by dividing the tongue body area is because the three areas of the tongue tip, tongue middle and tongue root have different physiological functions and pathological manifestations in traditional Chinese medicine theory. For example, the tongue tip is often related to the heart and lungs, the tongue middle is related to the spleen and stomach, and the tongue root is closely related to the kidneys. Therefore, extracting features by partition can more accurately reflect the specific health status of each part.

[0065] It should also be pointed out that the tongue color HSV value, the percentage of pixels of thick fur and the crack density were selected as tongue image features because tongue color is one of the important bases for diagnosis in traditional Chinese medicine and reflects the state of qi, blood and body fluids in the body. Quantitative analysis of the HSV color space can more accurately describe the tongue color state.

[0066] The thickness of the tongue coating is an important indicator for judging pathological conditions such as dampness, heat, phlegm and fluid in the body. By quantitatively analyzing the pixel ratio of the tongue coating covered area, the thickness of the tongue coating and its distribution can be objectively evaluated.

[0067] The presence and distribution of cracks on the tongue are also important reference factors in TCM diagnosis and treatment. The density of cracks reflects the microstructure of the tongue surface and can reflect the functional status and pathological changes of internal organs in the body.

[0068] Through the three characteristics of tongue color HSV value, coating thickness pixel ratio and crack density, the tongue image can be comprehensively evaluated from multiple dimensions such as color, texture and structure, providing richer diagnostic information.

[0069] In an optional implementation of the above scheme, the division of the tongue area can follow the following geometric rules: Tongue tip area: approximately the front third of the front end of the tongue.

[0070] Mid-tongue area: approximately the middle third of the tongue.

[0071] Tongue root area: approximately the posterior third of the tongue.

[0072] In another optional implementation of the above scheme, extracting the tongue color HSV value, the percentage of tongue coating thickness pixels and the crack density includes the following: performing a conversion from the RGB color space to the HSV color space for the three segmented areas of the tongue tip, the tongue middle and the tongue root.

[0073] In the HSV color space, the average values ​​of the H, S, and V channels of the pixels in each area are calculated to obtain the tongue color HSV value of each area.

[0074] A specific threshold range is defined in the HSV color space to identify white pixels as the tongue coating mark. The number of white pixels that meet the threshold condition is counted for the tip, middle and root of the tongue, and their proportion to the total number of pixels in each region is calculated as the proportion of thick coating pixels in the region.

[0075] It is important to know that thick moss usually appears as white or nearly white pixels in the HSV color space, which can be adjusted by setting a specific threshold range, such as , , to identify white pixels.

[0076] Cracks are identified in each region, and the ratio of the cumulative length of all identified cracks to the region area is calculated as the crack density.

[0077] It is important to know that cracks usually appear as dark linear structures on the surface of the tongue, and cracks can be identified by edge detection algorithms or morphological gradient operations.

[0078] The tongue image change analysis module is used to store the tongue image features of regularly uploaded tongue image images in time series, and generate a tongue image change curve including the tongue color, coating thickness and density corresponding to each tongue area.

[0079] The above-mentioned tongue image change curve is generated as follows: for each tongue area, based on its corresponding time series tongue image characteristics, the tongue color change curve, the tongue coating thickness change curve and the crack change curve are respectively drawn in a two-dimensional coordinate system with time as the horizontal axis and the tongue color HSV value, the percentage of pixels of the coating thickness and the crack density as the vertical axis.

[0080] The TCM big model prediction module is used to import the tongue image change curve into the constructed tongue image-syndrome model to analyze the syndrome type under the time series and output the syndrome type evolution path.

[0081] It should be noted that the tongue image-syndrome model mentioned above takes the tongue image characteristics corresponding to each tongue area as input and the syndrome type as output.

[0082] The basis for constructing the above-mentioned tongue image-syndrome model is that traditional Chinese medicine emphasizes "differentiation and treatment". Tongue image is one of the important bases for syndrome differentiation. Different tongue areas correspond to different parts. Tongue color, tongue coating thickness and crack characteristics are closely related to specific syndromes.

[0083] In one example, a red tip of the tongue, a thick and greasy coating in the middle of the tongue, and a pale root of the tongue correspond to the syndrome of hyperactivity of heart fire, damp heat in the spleen and stomach, and kidney yang deficiency, with the specific explanation being as follows: Red tip of the tongue, thin yellow coating: indicates hyperactivity of heart fire, which is commonly seen in symptoms such as insomnia, palpitations, and dry mouth.

[0084] Pale red tongue with thick and greasy coating: indicates damp-heat in the spleen and stomach, commonly seen in symptoms such as indigestion, abdominal distension, nausea and vomiting.

[0085] The root of the tongue is pale and the tongue coating is white and slippery: it indicates kidney yang deficiency, which is commonly seen in symptoms such as chills, cold limbs, and frequent urination at night.

[0086] When constructing the tongue image-syndrome model, a machine learning approach can be used: first, a large number of tongue images are collected, and each tongue image is labeled with the syndrome type by a professional Chinese medicine practitioner to ensure the authenticity and accuracy of the data. At the same time, the collected tongue images are preprocessed and the tongue features are extracted to form a feature vector. Subsequently, the feature vector and syndrome type annotation of each tongue image are used to train the model. The model performance and model adjustment are evaluated during the training process to achieve accurate mapping between tongue image features and syndrome types.

[0087] According to a preferred embodiment, the analysis of syndrome types under time series refers to the following process: the tongue image change curves corresponding to the tongue color, coating thickness, and cracks of each tongue body area are time-segmented. The specific segmentation process is as follows: the tongue image change curves corresponding to the tongue color, coating thickness, and density of each tongue body area are gradually identified from the starting time for the tongue color HSV value, coating thickness pixel ratio, and crack density changes at adjacent times. If all tongue image features have not changed within a certain adjacent time, the change identification of the tongue image features at the next adjacent time is performed until a certain tongue image parameter changes within the adjacent time, and the subsequent time corresponding to the adjacent time is used as the change time.

[0088] The above change recognition process is as follows: the differences in tongue color HSV value, fur thickness pixel ratio and crack density are calculated for adjacent times, and compared with the set critical difference. If the difference in a tongue parameter is greater than the critical difference, the recognition is changed.

[0089] It is important to understand that at each change time, at least one tongue characteristic changes.

[0090] Mark each change time as a divided time point on the tongue image change curve.

[0091] The above steps eventually generate a series of segmentation points, which divide the tongue image change curve into multiple time periods. The tongue image features in each time period remain relatively stable, while at least one feature changes significantly when crossing the segmentation point.

[0092] For each time point, the tongue image features of each tongue area are extracted from the above change curve.

[0093] The tongue image features corresponding to each tongue area at each time point are input into the tongue image-syndrome type model as the input set to obtain the syndrome type at each time point.

[0094] It should be pointed out that the above extraction of tongue features according to the time points of segmentation to determine the syndrome type is because when crossing a segmentation point, at least one tongue feature has changed significantly. This significant change usually marks an important change in the patient's tongue state. By extracting tongue features at each segmentation point, these important tongue state changes can be captured, and changes in tongue state are often closely related to changes in syndrome types. For example, a change in tongue color from pale white to red may indicate a change from qi deficiency to heat syndrome. By extracting features at each segmentation point and inputting them into the model, these changes in syndrome types can be more accurately identified, thereby improving the accuracy of diagnosis. In addition, if the tongue features do not change significantly within a period of time, there is no need to frequently re-evaluate the syndrome type. By setting segmentation points, new evaluations can be performed only when the features change, reducing unnecessary calculations and optimizing the use of system resources.

[0095] According to another preferred embodiment, see Figure 3 As shown, the output syndrome type evolution path includes the following contents: using lines to connect the syndromes of adjacent time points in the time series in sequence to form a continuous syndrome type evolution path.

[0096] For each time point, compare its syndrome type with the syndrome type at the previous time point to see if it is the same. If different, identify the time point as a syndrome type change time point and mark the syndrome type change time point in the syndrome type evolution path.

[0097] As an example of the above scheme, the time point of marking syndrome type change can be highlighted by using a specific symbol, such as an asterisk or a triangle mark.

[0098] Another example is that color marking can be used to mark the time points of syndrome type changes, such as using different colors or line types to distinguish the time periods before and after the change to enhance visual contrast.

[0099] The diagnosis and treatment association decision module is used to associate the patient's clinical indicators in the time series with the syndrome evolution path during the same period to form a syndrome-pathology association matrix, and recommend prescription compatibility plans based on this.

[0100] In the manner that the above scheme can be implemented, the clinical indicators of the patient in the same time series are associated with the syndrome evolution path to form a syndrome-pathology association matrix. The specific implementation is as follows: based on the syndrome change time points marked on the syndrome evolution path, the syndrome corresponding to each syndrome change time point is extracted.

[0101] For each syndrome type change time point, a time interval is determined under the set time delay condition, and the corresponding clinical indicator data are extracted within this time interval to form a syndrome type and clinical indicator mapping set corresponding to each syndrome type change time point.

[0102] In the above operation example, the time point of syndrome type change The time delay condition is The time interval determined below is .

[0103] It should be understood that in the process of extracting clinical indicator data based on the syndrome evolution path, setting a time delay condition to determine the time interval for clinical indicator extraction is based on the fact that TCM theory believes that changes in syndromes are often a comprehensive reflection of the body's internal pathological state, and this change usually lags behind changes in certain clinical indicators. For example, blood biochemical indicators such as blood sugar and blood lipids may change before tongue characteristics. By setting a time delay condition, these clinical indicators that change first can be captured, so as to better understand the potential mechanism of syndrome changes. In addition, the time delay condition allows the system to collect clinical indicator data within a reasonable time interval, thereby obtaining more comprehensive and representative information. This method can reduce the errors caused by sampling at a single time point.

[0104] It should be pointed out that the clinical indicators mentioned above include but are not limited to physical sign indicators, blood biochemical indicators, endocrine hormone indicators, etc., among which physical sign indicators include body temperature, heart rate, respiratory rate, blood pressure, etc., blood biochemical indicators include blood sugar, blood lipids, blood routine, electrolytes, etc., and endocrine hormone indicators include thyroid, sex hormones, etc.

[0105] Change identification is performed on the clinical indicator mapping set at the adjacent syndrome change time points under the time delay condition to determine the corresponding changed clinical indicators and their change amounts when the syndrome changes.

[0106] The change recognition mentioned above can be similarly referred to the change recognition method of tongue image parameters at adjacent times. The specific recognition steps are as follows: (1) Generate a clinical indicator change curve in a coordinate system with time as the horizontal axis and clinical indicators as the vertical axis for each clinical indicator number extracted within a time interval determined under time delay conditions.

[0107] (2) Extracting the stable period data from the generated clinical indicator change curve as the representative data of the clinical indicator. Specifically, the representative data of the clinical indicator can be the mean value or the median.

[0108] It is important to know that the above-mentioned stable period data refers to the small changes in clinical indicators within a certain period of time, which can reflect the typical values ​​of clinical indicators within the time range.

[0109] (3) Subtract the representative data of each clinical indicator corresponding to the adjacent syndrome change time points to obtain the difference of each clinical indicator and compare it with the critical difference. If the difference of a clinical indicator is greater than the critical difference, it is recognized that the clinical indicator has changed.

[0110] Furthermore, the change amount of the changed clinical indicators corresponding to adjacent syndrome type change time points is the difference of the changed clinical indicators.

[0111] Compare the syndrome types before and after each syndrome type change time point, and classify the same syndrome types before and after into one category.

[0112] The changing frequency and average changing amount of the changing clinical indicators in each type of before and after syndrome type combination are counted, and the correlation strength is assigned to the changing clinical indicators. The specific implementation is as follows: the changing frequency and average changing amount of the changing clinical indicators in each type of before and after syndrome type combination are normalized respectively.

[0113] It should be understood that the frequency of change refers to the number of times the changing clinical indicator changes repeatedly when the syndrome type changes before and after. The higher the frequency of change, the greater the frequency of change of the clinical indicator during the specific syndrome type conversion process, indicating that it has a stronger correlation with the syndrome type conversion.

[0114] The average change refers to the average value of each change of the changing clinical indicator during the conversion of the previous and subsequent syndrome types. The larger the average change, the greater the amplitude of the change of the clinical indicator during the conversion of a specific syndrome type, indicating that its influence in the conversion of syndrome types is more significant.

[0115] The above normalization of the change frequency and the average change is to map these values ​​into a standardized interval, such as the interval [0, 1]. This processing method ensures that data of different scales and magnitudes can be compared and analyzed on the same basis.

[0116] The change frequency and average change amount of the normalized changing clinical indicators are fused and calculated to obtain the correlation strength.

[0117] In one example, the fusion calculation can adopt the weighted summation method, and the weights of the change frequency and the average change amount can be set according to experience, and the sum of the weights of the change frequency and the average change amount is 1. This method can adjust the weights according to actual needs and is suitable for situations where the change frequency and the average change amount need to be comprehensively considered.

[0118] In another example, the fusion calculation can use multiplication. This method emphasizes that the correlation strength will only increase significantly when both the frequency of change and the average change amount are large. It is suitable for strictly screening changing clinical indicators.

[0119] It is important to understand that if a clinical indicator changes synchronously during the transformation of a patient's syndrome, it indicates that there may be an intrinsic relationship between the clinical indicator and the syndrome. By using the method of quantifying the frequency and amount of change, this association can be systematically measured and evaluated, thereby clarifying the strength and significance of the relationship between the two.

[0120] A syndrome-pathology association matrix was constructed based on the correlation strength between the previous and next syndrome combinations and the changing clinical indicators, in which the rows of the matrix represented the previous and next syndrome combinations, the columns represented the clinical indicators, and the matrix elements represented the correlation strength between the previous and next syndrome combinations and the changing clinical indicators.

[0121] In the above example, suppose we have three different combinations of before and after syndromes: Qi deficiency syndrome -> damp-heat syndrome, Qi deficiency syndrome -> Yin deficiency and hyperactivity of fire syndrome, damp-heat syndrome -> Yin deficiency and hyperactivity of fire syndrome, recorded as A, B, and C respectively, and the corresponding clinical indicators of these three syndrome combinations are: blood sugar, blood lipids, and white blood cell count. Based on this information, we will construct a syndrome-pathology association matrix as follows: .

[0122] For the syndrome conversion from "Qi deficiency syndrome -> damp-heat syndrome", the correlation strength of blood glucose is 0.75, indicating that the change of blood glucose in this process is more significant, which may be one of the key factors in the syndrome conversion.

[0123] For the syndrome conversion from Qi deficiency syndrome to Yin deficiency and hyperactivity of fire syndrome, the correlation strength of white blood cell count was 0.70, indicating that the change of white blood cell count was the most significant in this process and may have an important impact on the syndrome conversion.

[0124] For the syndrome conversion from damp-heat syndrome to Yin deficiency and hyperactivity of fire syndrome, the correlation strength of blood lipids is 0.85, indicating that the changes in blood lipids in this process are more significant, which may be one of the important signs of syndrome conversion.

[0125] In another implementation of the above scheme, the recommended prescription combination scheme includes the following content: a threshold is set based on the range of association strength values ​​in the association matrix to define a strong association.

[0126] For example, since the strength of association is calculated by fusing the normalized frequency of change and the average change, its value range is between [0, 1]. In order to emphasize the strong association, a threshold value, such as 0.5, can be set to define a strong association. This means that when the strength of association exceeds 0.5, it is considered that the clinical indicator has a strong association with the combination of the previous and next syndromes, thereby highlighting those indicators with significant effects.

[0127] The syndrome evolution path and syndrome-pathology association matrix are used to identify the syndrome evolution of the current patient and the clinical indicators that are strongly associated with it.

[0128] From the basic prescription library consisting of the mapping between syndrome types and commonly used prescriptions, basic prescriptions suitable for the current syndrome type evolution are screened out.

[0129] The clinical indicators in the correlation matrix that are strongly correlated with the current syndrome evolution are prioritized in the prescription compatibility according to the strength of correlation, and the basic prescription is optimized accordingly.

[0130] In the example of syndrome-pathology association matrix, suppose the current syndrome evolves from Qi deficiency syndrome to Yin deficiency and fire excess syndrome. According to the matrix, the clinical indicators strongly associated with this syndrome transition include blood lipids and white blood cell count, with association strengths of 0.6 and 0.7, respectively. Given the higher association strength of white blood cell count, it indicates that it has a greater impact in this syndrome conversion process. Therefore, in the basic prescription, it should be given priority to adjust the drug selection according to the changes in white blood cell count to optimize the treatment plan.

[0131] Specific drug selection is as follows: If the white blood cell count is elevated, indicating possible inflammation or infection, Chinese medicine with the effects of clearing heat and detoxifying, cooling blood and reducing swelling, such as honeysuckle, forsythia, etc., can be added to the basic prescription.

[0132] If the white blood cell count is low, it may indicate low immune function. Consider adding Chinese medicine that replenishes qi and blood and enhances immunity, such as astragalus and codonopsis.

[0133] Under the premise of ensuring the balance of the overall prescription, the drug combination is adjusted according to the specific changes in the white blood cell count to ensure that the prescription can effectively respond to the current pathological state and promote the transformation of the syndrome in a healthy direction.

[0134] It is important to note that prescription combinations are not static and should be adjusted regularly based on changes in the patient's condition and treatment outcomes.

[0135] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0136] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0137] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0138] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0139] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An auxiliary diagnosis and treatment system based on a large model of traditional Chinese medicine, characterized in that: include: Tongue image data collection module: receives tongue images uploaded by patients at regular intervals and triggers tongue image enhancement processing based on the image quality detection feedback mechanism of tongue clarity and color uniformity; Tongue feature extraction module: The enhanced tongue image is divided into three tongue regions: tongue tip, tongue middle, and tongue root, and the tongue features of each tongue region are extracted, specifically the tongue color HSV value, the percentage of fur thickness pixels, and the crack density; Tongue image change analysis module: the tongue image features of regularly uploaded tongue images are stored in time series, and a tongue image change curve including tongue color, coating thickness, and cracks corresponding to each tongue area is generated; TCM big model prediction module: import the tongue image change curve into the constructed tongue image-syndrome model to analyze the syndrome type under the time series, and output the syndrome type evolution path; Diagnosis and treatment association decision module: The clinical indicators of the patient in the same time series are associated with the syndrome evolution path to form a syndrome-pathology association matrix, and the prescription combination scheme is recommended based on it.

2. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 1, characterized in that: The image quality detection feedback mechanism based on tongue clarity and color uniformity triggers tongue image enhancement processing, see the following process: Performing a noise preprocessing operation on the received tongue image; The edge detection algorithm is used to extract the edge pixels of the tongue from the preprocessed image; The edge sharpness and edge continuity were evaluated based on the tongue edge pixels. The tongue clarity is defined as the product of edge sharpness and edge continuity. Under this definition, the tongue clarity of the tongue image is obtained based on the evaluated edge sharpness and edge continuity. Image segmentation technology is used to separate the tongue area from the background and generate a corresponding binary mask to ensure that the pixel value in the tongue area is marked as 1, while the background area is marked as 0; Convert the tongue image from RGB color space to HSV color space, and apply the mask generated above to accurately extract the color values ​​of all pixels in the tongue area; The standard deviation of the pixel value is calculated independently for each color channel, and the color uniformity is obtained by weighted summing of the standard deviations of each color channel; The tongue clarity and color uniformity of the tongue image are compared with the set standard thresholds. If both meet the standard thresholds, the tongue image enhancement processing will not be triggered. Conversely, if any indicator fails to reach the standard threshold, the tongue image enhancement processing will be automatically triggered.

3. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 2, characterized in that: The image quality detection feedback mechanism based on tongue clarity and color uniformity triggers the tongue image enhancement processing and further includes the following contents: Define the image enhancement processing rules as: a) When the tongue body clarity of the tongue image does not meet the standard threshold, the tongue image is subjected to edge degeneration processing; b) When the color uniformity of the tongue image does not meet the standard threshold, the tongue image is subjected to color correction processing; When the enhancement process of the tongue image is triggered, the image enhancement process is performed using the above rules within the specified time window; After each image enhancement process, the tongue clarity and color uniformity of the enhanced image are recalculated and compared with the standard threshold again. If it still does not meet the standard, multiple iterations of enhancement are performed according to the above rules until the standard threshold is met or the available processing time in the time window is exhausted; If the tongue clarity and color uniformity of the tongue image still do not meet the standard threshold at the end of the specified time window, a retake instruction is automatically generated, prompting the user to retake the tongue image.

4. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 1, characterized in that: The extraction of tongue color HSV value, moss thickness pixel ratio and crack density includes the following contents: For the segmented tongue tip, tongue middle and tongue root regions, perform conversion from RGB color space to HSV color space; In the HSV color space, the average values ​​of the H, S, and V channels of the pixels in each area are calculated to obtain the tongue color HSV value of each area; A specific threshold range is defined in the HSV color space to identify white pixels as tongue coating marks. The number of white pixels that meet the threshold condition is counted for the three regions of the tongue tip, middle tongue, and root, and the proportion of white pixels to the total number of pixels in each region is calculated as the proportion of thick coating pixels in the region. Cracks are identified in each region, and the ratio of the cumulative length of all identified cracks to the region area is calculated as the crack density.

5. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 1, characterized in that: The analysis of the syndrome type under the time series refers to the following process: The tongue image change curves corresponding to tongue color, tongue coating thickness, and cracks in each tongue area are divided into time segments; For each time point, the tongue image features of each tongue area are extracted from the above change curves; The tongue image features corresponding to each tongue area at each time point are input into the tongue image-syndrome type association model as the input set to obtain the syndrome type at each time point.

6. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 5, characterized in that: The tongue image change curves corresponding to tongue color, tongue coating thickness, and cracks in each tongue area are divided into time segments and analyzed as follows: The tongue image change curves corresponding to the tongue color, fur thickness, and density of each tongue area are used to gradually identify the changes in the tongue color HSV value, fur thickness pixel ratio, and crack density at adjacent times starting from the starting time. If all tongue image features have not changed within a certain adjacent time, the change identification of the tongue image features at the next adjacent time is performed until a certain tongue image parameter changes within the adjacent time, and the subsequent time corresponding to the adjacent time is regarded as the change time; Mark each change time as a divided time point on the tongue image change curve.

7. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 1, characterized in that: The output certificate type evolution path includes the following contents: Use lines to connect the syndromes at adjacent time points in the time series in sequence to form a continuous syndrome evolution path; For each time point, compare its syndrome type with the syndrome type at the previous time point to see if it is the same. If different, identify the time point as a syndrome type change time point and mark the syndrome type change time point in the syndrome type evolution path.

8. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 7, characterized in that: The specific implementation of the syndrome-pathology association matrix is ​​as follows: Extract the syndrome type corresponding to each syndrome type change time point based on the syndrome type change time point marked on the syndrome type evolution path; For each syndrome type change time point, a time interval is determined under the set time delay condition, and the corresponding clinical indicator data is extracted within this time interval to form a syndrome type and clinical indicator mapping set corresponding to each syndrome type change time point; Perform change recognition on the clinical indicator mapping set at adjacent syndrome type change time points to determine the corresponding changed clinical indicators and their change amounts when the syndrome type changes; Compare the syndrome types before and after each syndrome type change time point, and classify the same syndrome types before and after into one category; The changing frequency and average changing amount of the changing clinical indicators in each combination of the previous and next syndromes are counted, and the correlation strength is assigned to the changing clinical indicators; A syndrome-pathology association matrix was constructed based on the correlation strength between the previous and next syndrome combinations and the changing clinical indicators, in which the rows of the matrix represented the previous and next syndrome combinations, the columns represented the clinical indicators, and the matrix elements represented the correlation strength between the previous and next syndrome combinations and the changing clinical indicators.

9. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 8, characterized in that: The change clinical indicator allocation association strength includes the following: Normalize the frequency and average change of clinical indicators in each combination of previous and next syndrome types; The change frequency and average change amount of the normalized clinical indicators were fused and calculated to obtain the correlation strength.

10. The auxiliary diagnosis and treatment system based on the TCM big model as claimed in claim 8, characterized in that: The recommended prescription compatibility scheme is as follows: A threshold is set based on the range of association strength values ​​in the association matrix to define a strong association; Use the syndrome evolution path and syndrome-pathology association matrix to identify the syndrome evolution of the current patient and the clinical indicators that are strongly associated with it; Screen out basic prescriptions suitable for the current syndrome evolution from the basic prescription library consisting of the mapping between syndromes and commonly used prescriptions; The clinical indicators in the correlation matrix that are strongly correlated with the current syndrome evolution are prioritized in the prescription compatibility according to the strength of correlation, and the basic prescription is optimized accordingly.

Images