Tongue picture intelligent acquisition and auxiliary dialectical analysis method and tongue picture intelligent acquisition and auxiliary dialectical analysis system

Abstract

The invention is suitable for the field of traditional Chinese medicine intelligence, and provides a tongue picture intelligent acquisition and auxiliary dialectical analysis method, which comprises the following steps of: obtaining tongue picture data of each family member based on a constructed family group, and obtaining a multi-dimensional feature vector and corresponding auxiliary dialectical information after analysis; performing comparison and clustering analysis on the multi-dimensional feature vector and the auxiliary dialectical information to identify a common constitution type and a common feature vector of the family group; retrieving a corresponding dietary therapy substance set and a behavior set from a preset health knowledge map based on the common constitution type, and combining the dietary therapy substance set and the behavior set through a rule engine to generate a group conditioning scheme; and generating differentiated individual conditioning schemes by calculating the deviation degree of the multi-dimensional feature vector and the common feature vector of each family member. According to the invention, while the overall efficiency of family health management is ensured, the intervention effect of precise adaptation to each family member is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent traditional Chinese medicine, specifically to a method and system for intelligent tongue image acquisition and auxiliary syndrome differentiation analysis. Background Technology

[0002] Tongue diagnosis, a distinctive diagnostic method in Traditional Chinese Medicine (TCM), is currently a hot research topic in terms of its digitalization and intelligentization. Existing technologies include various standalone or cloud-based systems that capture tongue images using cameras and utilize image processing and artificial intelligence algorithms for feature extraction and constitution-assisted diagnosis. These systems can provide users with personalized tongue image analysis reports and health advice, effectively improving the objectivity and convenience of traditional tongue diagnosis.

[0003] However, existing intelligent tongue image analysis systems all focus on individual-level health assessments. Traditional Chinese medicine theory emphasizes the need to comprehensively consider various factors, neglecting the profound impact of the core environmental factor of "family" on individual health. Family members share similar genetic backgrounds, dietary structures, living environments, and emotional atmospheres, easily leading to common constitutional imbalances, such as a family-wide tendency towards phlegm-dampness or qi deficiency. Current technology cannot aggregate, compare, and correlate tongue image and constitution data among family members, making it difficult to identify common health risks from a holistic family perspective and provide collaborative management solutions.

[0004] Therefore, existing technologies have obvious shortcomings, which lead to isolated health management plans and even contradictions among family members' health adjustment plans. This greatly hinders the implementation of specific plans and reduces the overall efficiency and practicality of health intervention. Therefore, a method and system for intelligent tongue image acquisition and auxiliary diagnostic analysis is proposed to solve the above problems. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for intelligent acquisition and auxiliary diagnostic analysis of tongue images, so as to solve the problems existing in the above-mentioned background technology.

[0006] This invention is implemented as follows: a method for intelligent acquisition and assisted diagnostic analysis of tongue images, the method comprising the following steps:

[0007] Based on the established family group, tongue image data of each family member is obtained. After analysis, multidimensional feature vectors and corresponding auxiliary diagnostic information are obtained. The multidimensional feature vectors include tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number, and tongue body slenderness index.

[0008] By comparing and clustering multidimensional feature vectors and auxiliary diagnostic information, the common physical constitution type and common feature vector of the family group are identified, and a group health analysis report is generated.

[0009] Based on the common constitution type, the corresponding set of dietary therapy substances and set of behaviors are retrieved from the preset health knowledge graph. The set of dietary therapy substances and set of behaviors are combined by the rule engine to generate a group conditioning plan.

[0010] Based on the group conditioning plan, a differentiated individual conditioning plan is generated by calculating the deviation between the multidimensional feature vector of each family member and the common feature vector. The group health analysis report, the group conditioning plan and the individual conditioning plan are then integrated and pushed to the user's terminal.

[0011] As a further aspect of the present invention: the step of acquiring tongue image data of each family member, and obtaining multi-dimensional feature vectors and corresponding auxiliary diagnostic information after analysis, specifically includes:

[0012] Tongue image data of each family member is collected using a confocal white light scanning device. The tongue image data includes three-dimensional point cloud data of the tongue surface and its corresponding RGB color information.

[0013] The tongue region is segmented from the three-dimensional point cloud data, and the theoretical base plane characterizing the texture of the tongue is calculated using a plane fitting algorithm based on the point cloud data of the tongue region.

[0014] Based on the three-dimensional point cloud data and the theoretical base plane, a multi-dimensional feature vector containing the coating thickness index, tooth mark depth, tooth mark number, and tongue body slenderness index is calculated and generated.

[0015] The indicators in the multidimensional feature vector are compared with preset thresholds to generate preliminary auxiliary diagnostic information.

[0016] As a further aspect of the present invention, the method further includes:

[0017] Based on the three-dimensional point cloud data, areas with a tongue coating thickness index greater than a set threshold and areas marked with teeth marks are excluded, and the remaining areas are defined as tongue quality analysis areas.

[0018] Extract the RGB color information from the tongue analysis area, convert it to the HSV color space, and calculate its average value to obtain the tongue color HSV space value;

[0019] The hue values ​​in the calculated tongue color HSV space are compared with the preset range, and the comparison results are added to the preliminary auxiliary diagnostic information.

[0020] The tongue color HSV spatial values ​​are combined with the coating thickness index, teeth mark depth, teeth mark number, and tongue body thickness index to form a complete multidimensional feature vector.

[0021] As a further aspect of the present invention: the step of comparing and clustering multidimensional feature vectors and auxiliary diagnostic information specifically includes:

[0022] The multidimensional feature vectors of each family member are concatenated with the auxiliary diagnostic information encoded by One-Hot to obtain the family multimodal data matrix;

[0023] A consensus clustering algorithm was applied to the family multimodal data matrix. The data was resampled and clustered multiple times using K-Means and hierarchical clustering methods, and the confidence score of each member in belonging to their respective cluster was calculated.

[0024] Members with a confidence level higher than a set value are selected to form a core member set, and the constitution type with the highest proportion in this set is taken as the common constitution type of the family group;

[0025] The arithmetic mean of the multidimensional feature vectors of all members in the core member set is calculated, and this average vector is defined as the common feature vector of the family.

[0026] As a further aspect of the present invention: the set of dietary therapeutic substances is a set of nodes in the health knowledge graph that are connected to the common constitution type nodes through a constitution conditioning relationship, and whose entity type is edible entity or medicinal edible entity, wherein each entity node contains three data fields: substance name, flavor attribute, and meridian attribute.

[0027] The behavior set is a set of nodes in the health knowledge graph that are connected to the common physical constitution type nodes through behavioral intervention relationships, and whose entity type is a daily routine entity or a physical activity entity. Each entity node contains three data fields: behavior description, recommendation intensity, and prohibited time period.

[0028] As a further aspect of the present invention: the step of calculating the deviation between the multidimensional feature vector of each family member and the common feature vector specifically includes:

[0029] The dataset consisting of the common feature vector and the multidimensional feature vectors of all family members is standardized, and the sample covariance matrix of the standardized dataset is calculated.

[0030] For each family member, calculate the Mahalanobis distance between their standardized multidimensional feature vector and their standardized common feature vector. The value of the Mahalanobis distance is the overall deviation.

[0031] The overall deviation is decomposed into the tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number, and tongue body thickness index of the multidimensional feature vector, and the contribution percentage of each dimension relative to the overall deviation is calculated.

[0032] The overall deviation is combined with all contribution percentages to generate a structured individualized deviation profile, which is used to generate an individualized treatment plan.

[0033] Another objective of this invention is to provide a tongue image intelligent acquisition and auxiliary diagnostic analysis system, the system comprising:

[0034] The acquisition and analysis module is used to acquire tongue image data of each family member based on the constructed family group. After analysis, multidimensional feature vectors and corresponding auxiliary diagnostic information are obtained. The multidimensional feature vectors include tongue color HSV space value, coating thickness index, tooth mark depth, number of tooth marks and tongue body slenderness index.

[0035] The commonality analysis module is used to identify the common physical constitution type and common feature vector of the family group by comparing and clustering multidimensional feature vectors and auxiliary diagnostic information, and to generate a group health analysis report.

[0036] The group treatment plan generation module is used to retrieve the corresponding set of dietary therapy substances and set of behaviors from the preset health knowledge graph based on the common physical constitution type, and to combine the set of dietary therapy substances and set of behaviors through the rule engine to generate a group conditioning plan.

[0037] The individualized module is used to generate differentiated individualized treatment plans based on the group treatment plan by calculating the deviation between the multidimensional feature vectors of each family member and the common feature vectors, and to integrate and push the group health analysis report, the group treatment plan and the individualized treatment plan to the user's terminal.

[0038] As a further aspect of the present invention: the acquisition and analysis module includes:

[0039] The data acquisition unit is used to acquire tongue image data of each family member through a confocal white light scanning device. The tongue image data includes three-dimensional point cloud data of the tongue surface and its corresponding RGB color information.

[0040] The base determination unit is used to segment the tongue region from the three-dimensional point cloud data and calculate the theoretical base plane characterizing the texture of the tongue based on the point cloud data of the tongue region using a plane fitting algorithm.

[0041] The sub-component calculation unit is used to calculate and generate a multi-dimensional feature vector containing the coating thickness index, tooth mark depth, tooth mark number, and tongue body slenderness index based on the three-dimensional point cloud data and the theoretical base plane.

[0042] The indicator analysis unit is used to compare the various indicators in the multidimensional feature vector with preset thresholds to generate preliminary auxiliary diagnostic information.

[0043] As a further aspect of the present invention: the commonality analysis module includes:

[0044] The data matrix construction unit is used to concatenate the multidimensional feature vectors of each family member with the auxiliary diagnostic information encoded by One-Hot to obtain a family multimodal data matrix;

[0045] The algorithm processing unit is used to perform consensus clustering algorithm on the family multimodal data matrix. It repeatedly uses K-Means and hierarchical clustering to resample and cluster the data, and calculates the confidence of each member to belong to its cluster.

[0046] The filtering unit is used to filter members whose attribution confidence is higher than a set value to form a core member set, and the constitution type with the highest proportion in this set is taken as the common constitution type of the family group;

[0047] The commonality calculation unit is used to calculate the arithmetic mean of the multidimensional feature vectors of all members in the core member set, and define the average value vector as the common feature vector of the family.

[0048] As a further aspect of the present invention: the individual differentiation module includes:

[0049] The dataset construction unit is used to standardize the dataset composed of the common feature vector and the multidimensional feature vectors of all family members, and to calculate the sample covariance matrix of the standardized dataset.

[0050] The deviation calculation unit is used to calculate the Mahalanobis distance between the standardized multidimensional feature vector and the standardized common feature vector for each family member. The value of the Mahalanobis distance is the comprehensive deviation.

[0051] The deviation decomposition unit is used to decompose the comprehensive deviation into the tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number, and tongue body thickness index of the multidimensional feature vector, and calculate the percentage contribution of each dimension to the comprehensive deviation.

[0052] The data combination unit is used to combine the overall deviation with all contribution percentages to generate a structured individualized deviation profile, which is used to generate an individualized treatment plan.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] This invention constructs family groups and collects tongue image data from members. Analysis yields multi-dimensional feature vectors and diagnostic information containing quantitative indicators such as tongue color, coating thickness, teeth marks, and tongue shape. Cluster analysis is then used to identify common family characteristics. Based on this, a group-based conditioning plan is generated using a health knowledge graph. Personalized plans are derived by calculating the quantitative deviation between individual characteristics and common family characteristics, and finally, these plans are integrated and pushed to users. Firstly, by establishing a family health database and a group analysis model, this invention fundamentally breaks through the bottleneck of existing technologies being limited to individual assessment, achieving a paradigm shift from individual health monitoring to family health system management. Secondly, through a two-tiered plan architecture of group and individualized approaches, it effectively solves the prominent problem of traditional health management plans being isolated or even contradictory in family settings, significantly improving the synergy and feasibility of intervention measures. Finally, in terms of personalization, it is based on precise data difference analysis, ensuring the overall efficiency of family health management while achieving precise intervention effects tailored to each family member, thereby comprehensively improving individual compliance and the overall quality of health intervention. Attached Figure Description

[0055] Figure 1 This is a flowchart of a method for intelligent tongue image acquisition and auxiliary diagnostic analysis.

[0056] Figure 2 This is a flowchart illustrating the process of obtaining multidimensional feature vectors and corresponding auxiliary diagnostic information in a method for intelligent tongue image acquisition and auxiliary diagnostic analysis.

[0057] Figure 3 This is a flowchart illustrating the comparison and clustering analysis of multidimensional feature vectors and auxiliary diagnostic information in a method for intelligent tongue image acquisition and auxiliary diagnostic analysis.

[0058] Figure 4 This is a flowchart illustrating the deviation between the multidimensional feature vector and the common feature vector of each family member in a method for intelligent tongue image acquisition and assisted diagnostic analysis.

[0059] Figure 5 This is a schematic diagram of a tongue image intelligent acquisition and auxiliary diagnostic analysis system.

[0060] Figure 6 This is a schematic diagram of the acquisition and analysis module in a tongue image intelligent acquisition and auxiliary diagnostic analysis system.

[0061] Figure 7 This is a schematic diagram of the commonality analysis module in a tongue image intelligent acquisition and auxiliary diagnostic analysis system.

[0062] Figure 8 This is a schematic diagram of the individual difference module in an intelligent tongue image acquisition and auxiliary diagnostic analysis system. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0064] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0065] like Figure 1 As shown in the figure, this invention provides a method for intelligent tongue image acquisition and assisted diagnostic analysis, the method comprising the following steps:

[0066] S100: Based on the established family group, obtain the tongue image data of each family member, and after analysis, obtain multi-dimensional feature vectors and corresponding auxiliary diagnostic information. The multi-dimensional feature vectors include tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number and tongue body fat index.

[0067] S200, through comparison and cluster analysis of multidimensional feature vectors and auxiliary diagnostic information, is used to identify the common physical constitution type and common feature vector of the family group and generate a group health analysis report.

[0068] S300, based on the common constitution type, retrieve the corresponding set of dietary therapy substances and set of behaviors from the preset health knowledge graph, and combine the set of dietary therapy substances and set of behaviors through the rule engine to generate a group conditioning plan;

[0069] S400, based on the group conditioning plan, generates a differentiated individual conditioning plan by calculating the deviation between the multidimensional feature vector of each family member and the common feature vector, and integrates and pushes the group health analysis report, group conditioning plan and individual conditioning plan to the user's terminal.

[0070] It should be noted that the aforementioned family group refers to a digital family unit established through account association, used to aggregate tongue image data of multiple members. The multidimensional feature vector is a standardized set of quantitative indicators composed of tongue color HSV space values ​​(representing the hue, saturation, and brightness of tongue color), coating thickness index (quantifying the thickness of tongue coating accumulation), teeth mark depth (measuring the degree of tongue edge depression), teeth mark number (statistically counting serrated marks on the tongue edge), and tongue body plumpness index (reflecting tongue morphological characteristics). The health knowledge graph is a structured knowledge base containing entities and their relationships related to TCM constitution, dietary therapy substances, and behavioral interventions. Furthermore, the dietary therapy substance set is a set of nodes in the health knowledge graph that are connected to the common constitution type nodes through constitution conditioning relationships, and whose entity type is edible or medicinal edible entities. Each entity node contains three data fields: substance name, flavor attribute, and meridian attribute. The behavior set is a set of nodes in the health knowledge graph that are connected to the common constitution type nodes through behavioral intervention relationships, and whose entity type is either a lifestyle or physical activity entity. Each entity node contains three data fields: behavior description, recommendation strength, and contraindication period. It should be noted that the constitution conditioning relationship is a directed edge in the graph connecting a specific constitution type node to an entity node that has a conditioning effect on it. Edible entities refer to substance nodes that can be eaten as ordinary food. Medicinal edible entities refer to substance nodes that have both edible value and clear medicinal effects. The nature and flavor attributes are data fields characterizing the cold, hot, warm, and cool medicinal properties and the five flavors of a substance. The meridian tropism attribute is a data field characterizing the action of a substance on the meridians and internal organs of the human body. The behavioral intervention relationship is a directed edge in the graph connecting constitution type nodes to behavioral nodes with regulatory effects. Lifestyle entities refer to behavioral nodes related to daily rhythms and habits. Physical activity entities refer to behavioral nodes related to physical activities such as exercise. The recommendation strength is a field that quantitatively characterizes the strength of the behavioral intervention relationship. The forbidden period is a field that identifies the time interval in which the action is not appropriate.

[0071] In this embodiment of the invention, a family group is constructed and tongue image data of its members is collected. Analysis yields a multi-dimensional feature vector and diagnostic information containing quantitative indicators such as tongue color, coating thickness, teeth marks, and tongue shape. Cluster analysis is then used to identify commonalities within the family. Based on this, a group-based conditioning plan is generated using a health knowledge graph. Personalized plans are derived by calculating the quantitative deviation between individual characteristics and common family characteristics, and finally, these plans are integrated and pushed to the user. Firstly, by establishing a family health database and a group analysis model, the invention fundamentally breaks through the bottleneck of existing technologies being limited to individual assessment, achieving a paradigm shift from individual health monitoring to family health system management. Secondly, through a two-tiered plan architecture of group and individualized approaches, the invention effectively solves the prominent problem of traditional health management plans being isolated or even contradictory in family settings, significantly improving the synergy and feasibility of intervention measures. Finally, in terms of personalization, it is based on precise data difference analysis, ensuring the overall efficiency of family health management while achieving precise intervention effects tailored to each family member, thereby comprehensively improving individual compliance and the overall quality of health intervention.

[0072] like Figure 2 As shown in the preferred embodiment of the present invention, the step of acquiring tongue image data of each family member and obtaining multi-dimensional feature vectors and corresponding auxiliary diagnostic information after analysis specifically includes:

[0073] S101, Collect tongue image data of each family member through a confocal white light scanning device. The tongue image data includes three-dimensional point cloud data of the tongue surface and its corresponding RGB color information.

[0074] S102, the tongue region is segmented from the three-dimensional point cloud data, and the theoretical base plane characterizing the texture of the tongue is calculated using a plane fitting algorithm based on the point cloud data of the tongue region.

[0075] S103, based on the three-dimensional point cloud data and the theoretical base plane, calculate and generate a multi-dimensional feature vector including the coating thickness index, tooth mark depth, tooth mark number, and tongue body thickness index;

[0076] S104, compare each indicator in the multidimensional feature vector with a preset threshold to generate preliminary auxiliary diagnostic information;

[0077] S105, Based on the three-dimensional point cloud data, exclude areas where the tongue coating thickness index is greater than a set threshold and areas marked with teeth marks, and define the remaining areas as tongue quality analysis areas.

[0078] S106. Extract the RGB color information in the tongue analysis area, convert it to the HSV color space, and calculate its average value to obtain the tongue color HSV space value.

[0079] S107, compare the hue value in the calculated tongue color HSV space with the preset range, and supplement the comparison result into the preliminary auxiliary diagnostic information;

[0080] S108, the tongue color HSV spatial values ​​are combined with the coating thickness index, teeth mark depth, teeth mark number, and tongue body plumpness index to form a complete multidimensional feature vector.

[0081] In this embodiment of the invention, a high-precision three-dimensional point cloud and true-color RGB information of the tongue are simultaneously acquired using a confocal white light scanning device to establish a digital tongue model, laying the data foundation for subsequent quantitative analysis. In specific implementation, the tongue region is first segmented in the three-dimensional point cloud, and a theoretical base plane is constructed using a plane fitting algorithm. This serves as the geometric benchmark for accurate calculation of the tongue coating thickness index (the objective coating thickness value is obtained by calculating the average distance from each point cloud to the theoretical base plane). Simultaneously, tooth mark features are automatically identified and their depth and quantity are quantified. Combined with the tongue contour, a plumpness / thinness index is calculated, forming a subset of feature vectors representing morphological features. In the color analysis stage, the system accurately defines the tongue body region by excluding thick coating areas and tooth mark areas, extracts the RGB values ​​of this region, and converts them to the HSV color space. A stable and reliable tongue color HSV value is obtained by calculating the average value. Finally, the tongue color HSV value is integrated with the aforementioned morphological indicators into a complete multi-dimensional feature vector. This vector is then automatically generated by comparing it with a system within a preset threshold range, integrating quantitative features with auxiliary diagnostic information based on traditional Chinese medicine differentiation. In summary, by employing a three-dimensional scanning and layered analysis approach, the measurement errors caused by perspective, deformation, and light interference in traditional two-dimensional tongue image analysis are effectively resolved. This approach enables millimeter-level precise quantification of key features such as tongue coating thickness and teeth marks, as well as stable and interference-resistant extraction of tongue color, significantly improving the objectivity, repeatability, and diagnostic value of tongue image analysis.

[0082] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of comparing and clustering multidimensional feature vectors and auxiliary diagnostic information specifically includes:

[0083] S201, the multidimensional feature vector of each family member is concatenated with the auxiliary identification information after One-Hot encoding to obtain the family multimodal data matrix;

[0084] S202, executes consensus clustering algorithm on family multimodal data matrix, uses K-Means and hierarchical clustering method to resample and cluster data multiple times, and calculates the confidence of each member to belong to its cluster;

[0085] S203, select members with a confidence level higher than the set value to form a core member set, and use the constitution type with the highest proportion in this set as the common constitution type of the family group;

[0086] S204, calculate the arithmetic mean of the multidimensional feature vectors of all members in the core member set, and define the average vector as the common feature vector of the family.

[0087] In this embodiment of the invention, a family multimodal data matrix is ​​constructed by concatenating the multidimensional feature vectors of each family member with auxiliary diagnostic information encoded by One-Hot, thus achieving the organic integration of tongue appearance quantitative features and TCM diagnostic information. Based on this, a consensus clustering algorithm is employed, repeatedly using K-Means and hierarchical clustering to resample and cluster the data, and calculating the confidence level of each member's affiliation to their respective cluster. This effectively improves the stability and reliability of constitution classification (for example, when a member is consistently classified into the phlegm-dampness constitution group in multiple clustering results, their affiliation confidence level will be significantly higher than that of members whose affiliation fluctuates between different groups). Furthermore, members with a affiliation confidence level higher than a set value (e.g., 90%) are selected to form a core member set, and the constitution type with the highest proportion in this set is used as the common constitution type of the family group, ensuring the representativeness of the commonality judgment. Finally, by calculating the arithmetic mean of the multidimensional feature vectors of all members in the core member set, a common feature vector representing the overall tongue appearance characteristics of the family is generated. This technical approach, through multimodal data fusion and consensus clustering mechanisms, effectively solves the problem of difficulty in extracting common features due to individual differences in traditional family health analysis, significantly improves the accuracy and stability of family physical type determination, and provides reliable data support and scientific basis for the subsequent development of targeted family health management plans.

[0088] like Figure 4 As shown in the preferred embodiment of the present invention, the step of calculating the deviation between the multidimensional feature vector of each family member and the common feature vector specifically includes:

[0089] S401, standardize the dataset composed of the common feature vector and the multidimensional feature vectors of all family members, and calculate the sample covariance matrix of the standardized dataset.

[0090] S402, For each family member, calculate the Mahalanobis distance between its standardized multidimensional feature vector and its standardized common feature vector. The value of the Mahalanobis distance is the comprehensive deviation.

[0091] S403, decompose the comprehensive deviation into the tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number, and tongue body thickness index of the multidimensional feature vector, and calculate the percentage contribution of each dimension to the comprehensive deviation.

[0092] S404, The overall deviation is combined with all contribution percentages to generate a structured individualized deviation profile, which is used to generate an individualized treatment plan.

[0093] In this embodiment of the invention, the common feature vector and the feature vectors of all family members are unified to the same dimension through standardization processing, and their sample covariance matrix is ​​calculated to characterize the intrinsic correlation between each feature dimension. Based on this, the Mahalanobis distance between each family member's standardized feature vector and the family's common vector is calculated. This distance value serves as the comprehensive deviation, considering not only the degree of variation and interrelationship of each feature dimension, but also more accurately reflecting the degree of difference between the individual and the overall family status. For example, when a member's tongue color HSV value and tongue coating thickness index both significantly deviate from the common features, and these two dimensions are strongly correlated, the Mahalanobis distance can provide a more reasonable deviation assessment. Furthermore, by decomposing the comprehensive deviation into five specific dimensions—tongue color HSV spatial value, tongue coating thickness index, teeth mark depth, teeth mark number, and tongue body thickness index—the contribution percentage of each dimension can be calculated, thereby accurately identifying the main factors causing the deviation. For example, if a member's comprehensive deviation is 2.5, with the tongue coating thickness index contributing 40% and the tongue color H value contributing 35%, it indicates that the deviation mainly stems from abnormal tongue coating and tongue color changes, making individual analysis more accurate. Finally, the overall deviation degree is combined with the percentage contribution of each dimension to generate a structured, individualized deviation profile. This method, by introducing a dual mechanism of Mahalanobis distance and contribution analysis, not only achieves the scientific quantification of the degree of individual deviation, but also reveals the specific causes of the deviation, providing a key basis for the subsequent development of precise individualized treatment plans.

[0094] like Figure 5 As shown in the figure, this embodiment of the invention also provides a tongue image intelligent acquisition and auxiliary diagnostic analysis system, the system comprising:

[0095] The acquisition and analysis module 100 is used to acquire tongue image data of each family member based on the constructed family group, and after analysis, obtain multidimensional feature vectors and corresponding auxiliary diagnostic information. The multidimensional feature vectors include tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number and tongue body slenderness index.

[0096] The commonality analysis module 200 is used to identify the common physical constitution type and common feature vector of the family group by comparing and clustering multidimensional feature vectors and auxiliary diagnostic information, and to generate a group health analysis report.

[0097] The group treatment plan generation module 300 is used to retrieve the corresponding set of dietary therapy substances and set of behaviors from the preset health knowledge graph based on the common physical constitution type, and to combine the set of dietary therapy substances and set of behaviors through the rule engine to generate a group conditioning plan.

[0098] The individual differentiation module 400 is used to generate differentiated individual conditioning plans based on the group conditioning plan by calculating the deviation between the multidimensional feature vector of each family member and the common feature vector, and to integrate and push the group health analysis report, the group conditioning plan and the individual conditioning plan to the user's terminal.

[0099] In this embodiment of the invention, the system achieves standardized collection and feature extraction of tongue image data through the data acquisition and analysis module 100, generating multi-dimensional feature vectors and auxiliary diagnostic information containing quantitative indicators such as tongue color, coating thickness, teeth marks, and tongue shape; the commonality analysis module 200 identifies common constitution types and feature vectors through cluster analysis of family group data, and generates a group health analysis report; the group plan generation module 300 generates a conditioning plan suitable for the whole family based on knowledge graph retrieval and rule engine; the individual differentiation module 400 generates a personalized conditioning plan based on the group plan by calculating the quantitative deviation between individual characteristics and common family characteristics, and finally integrates and pushes the group report, group plan, and individual plan to the user terminal, realizing the systematization, precision, and personalization of family health management.

[0100] like Figure 6 As shown, in a preferred embodiment of the present invention, the acquisition and analysis module 100 includes:

[0101] The data acquisition unit 101 is used to acquire tongue image data of each family member through a confocal white light scanning device. The tongue image data includes three-dimensional point cloud data of the tongue surface and its corresponding RGB color information.

[0102] The base determination unit 102 is used to segment the tongue region in the three-dimensional point cloud data and calculate the theoretical base plane characterizing the texture of the tongue based on the point cloud data of the tongue region using a plane fitting algorithm.

[0103] The sub-item calculation unit 103 is used to calculate and generate a multi-dimensional feature vector containing the coating thickness index, tooth mark depth, tooth mark number, and tongue body slenderness index based on the three-dimensional point cloud data and the theoretical base plane.

[0104] The indicator analysis unit 104 is used to compare each indicator in the multidimensional feature vector with a preset threshold to generate preliminary auxiliary diagnostic information.

[0105] like Figure 7 As shown, in a preferred embodiment of the present invention, the commonality analysis module 200 includes:

[0106] The data matrix construction unit 201 is used to concatenate the multidimensional feature vectors of each family member with the auxiliary identification information after One-Hot encoding to obtain a family multimodal data matrix;

[0107] The algorithm processing unit 202 is used to perform consensus clustering algorithm on the family multimodal data matrix, and to resample and cluster the data multiple times using K-Means and hierarchical clustering methods, and to calculate the confidence of each member in belonging to its cluster.

[0108] The filtering unit 203 is used to filter members whose attribution confidence is higher than a set value to form a core member set, and the constitution type with the highest proportion in this set is taken as the common constitution type of the family group.

[0109] The commonality calculation unit 204 is used to calculate the arithmetic mean of the multidimensional feature vectors of all members in the core member set, and define the average value vector as the common feature vector of the family.

[0110] like Figure 8 As shown, in a preferred embodiment of the present invention, the individual differentiation module 400 includes:

[0111] Data set construction unit 401 is used to standardize the dataset composed of the common feature vector and the multidimensional feature vectors of all family members, and to calculate the sample covariance matrix of the standardized dataset.

[0112] The deviation calculation unit 402 is used to calculate the Mahalanobis distance between the standardized multidimensional feature vector and the standardized common feature vector for each family member. The value of the Mahalanobis distance is the comprehensive deviation.

[0113] The deviation decomposition unit 403 is used to decompose the comprehensive deviation into the tongue color HSV space value, coating thickness index, tooth mark depth, tooth mark number, and tongue body thickness index of the multidimensional feature vector, and calculate the percentage contribution of each dimension to the comprehensive deviation.

[0114] The data combination unit 404 is used to combine the comprehensive deviation degree with all contribution percentages to generate a structured individualized deviation profile, which is used to generate an individualized treatment plan.

[0115] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0116] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0117] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0118] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method, characterized in that, The method comprises the following steps: Based on the constructed family group, the tongue image data of each family member is obtained, and after analysis, a multi-dimensional feature vector and corresponding auxiliary syndrome information are obtained, the multi-dimensional feature vector comprising tongue color HSV space numerical value, moss thickness index, tooth mark depth, tooth mark number and tongue body fat index; By comparing and clustering the multi-dimensional feature vector and the auxiliary syndrome information, the common constitution type and the common feature vector of the family group are identified, and a group health analysis report is generated; Based on the common constitution type, a corresponding set of diet therapy materials and behavior sets are retrieved from the preset health knowledge graph, and the set of diet therapy materials and the set of behavior sets are combined by a rule engine to generate a group conditioning scheme; Based on the group conditioning scheme, individual conditioning schemes with differences are generated by calculating the deviation degree of the multi-dimensional feature vector of each family member and the common feature vector, and the group health analysis report, the group conditioning scheme and the individual conditioning scheme are integrated and pushed to the terminal of the user.

2. The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method according to claim 1, characterized in that, The step of obtaining the tongue image data of each family member and obtaining the multi-dimensional feature vector and the corresponding auxiliary syndrome information after analysis comprises: Collecting the tongue image data of each family member by a confocal white light scanning device, the tongue image data comprising three-dimensional point cloud data and corresponding RGB color information of the tongue surface; Segmenting the middle tongue area in the three-dimensional point cloud data, and calculating the theoretical base plane representing the tongue texture by a plane fitting algorithm according to the point cloud data of the middle tongue area; Based on the three-dimensional point cloud data and the theoretical base plane, a multi-dimensional feature vector comprising the moss thickness index, the tooth mark depth, the tooth mark number and the tongue body fat index is calculated and generated; Comparing each index in the multi-dimensional feature vector with a preset threshold to generate preliminary auxiliary syndrome information.

3. The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method according to claim 2, characterized in that, The method further comprises: Based on the three-dimensional point cloud data, excluding the area with moss thickness index greater than a set threshold and the area marked with tooth marks, and defining the remaining area as a tongue quality analysis area; Extracting the RGB color information in the tongue quality analysis area and converting it to HSV color space to obtain the tongue color HSV space numerical value; Comparing the hue value in the calculated tongue color HSV space numerical value with a preset range, and supplementing the comparison result to the preliminary auxiliary syndrome information; Combining the tongue color HSV space numerical value with the moss thickness index, the tooth mark depth, the tooth mark number and the tongue body fat index to form a complete multi-dimensional feature vector.

4. The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method according to claim 1, characterized in that, The step of comparing and clustering the multi-dimensional feature vector and the auxiliary syndrome information comprises: Splicing the multi-dimensional feature vector of each family member with the auxiliary syndrome information after One-Hot encoding to obtain a family multi-modal data matrix; Performing consensus clustering algorithm on the family multi-modal data matrix, repeatedly resampling and clustering the data by K-Means and hierarchical clustering method, and calculating the membership confidence of each member to the class cluster; Screening members with membership confidence higher than a set value to form a core member set, and taking the constitution type with the highest proportion in the set as the common constitution type of the family group; An arithmetic mean of the multi-dimensional feature vectors of all members in the core member set is calculated, and the mean vector is defined as the common feature vector of the family.

5. The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method according to claim 1, characterized in that, The diet therapy material set is a node set connected to the common constitution type node through a constitution conditioning relationship in the health knowledge graph, and the entity type is a dietary entity or a medicinal dietary entity, wherein each entity node includes three data fields of material name, property and channel attribute; The behavior set is a node set connected to the common constitution type node through a behavior intervention relationship in the health knowledge graph, and the entity type is a life schedule entity or a physical activity entity, wherein each entity node includes three data fields of behavior description, recommended intensity and taboo period.

6. The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis method according to claim 1, characterized in that, The step of calculating the deviation degree of the multi-dimensional feature vector of each family member from the common feature vector specifically includes: The common feature vector and the data set composed of the multi-dimensional feature vectors of all family members are standardized, and the sample covariance matrix of the standardized data set is calculated; For each family member, the Mahalanobis distance between the standardized multi-dimensional feature vector and the standardized common feature vector is calculated, and the value of the Mahalanobis distance is the comprehensive deviation degree; The comprehensive deviation degree is decomposed into tongue color HSV space values, moss thickness index, tooth mark depth, tooth mark number and tongue body fat index, and the contribution percentage of each dimension to the comprehensive deviation degree is calculated; The comprehensive deviation degree and all contribution percentages are combined to generate a structured individual deviation archive, which is used to generate an individual conditioning scheme subsequently.

7. A tongue image intelligent acquisition and auxiliary syndrome differentiation analysis system, characterized in that, The system comprises: A collection and analysis module is configured to obtain tongue image data of each family member based on the constructed family group, and obtain multi-dimensional feature vectors and corresponding auxiliary syndrome information after analysis, wherein the multi-dimensional feature vectors include tongue color HSV space values, moss thickness index, tooth mark depth, tooth mark number and tongue body fat index. A common analysis module is configured to compare and cluster analyze the multi-dimensional feature vectors and the auxiliary syndrome information, identify the common constitution type and the common feature vector of the family group, and generate a group health analysis report. A group scheme generation module is configured to retrieve corresponding diet therapy material sets and behavior sets from a preset health knowledge graph based on the common constitution type, and combine the diet therapy material sets and the behavior sets through a rule engine to generate a group conditioning scheme. An individual difference module is configured to generate individual conditioning schemes with differences based on the group conditioning scheme by calculating the deviation degree of the multi-dimensional feature vector of each family member from the common feature vector, and integrate and push the group health analysis report, the group conditioning scheme and the individual conditioning schemes to a terminal of a user. 8.The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis system according to claim 7, characterized in that, The collection and analysis module comprises: A data collection unit is configured to collect tongue image data of each family member through a confocal white light scanning device, wherein the tongue image data includes three-dimensional point cloud data and corresponding RGB color information of a tongue surface. A base determining unit is configured to segment a tongue middle region from the three-dimensional point cloud data, and calculate a theoretical base plane representing tongue texture according to point cloud data of the tongue middle region by using a plane fitting algorithm; A sub-item calculating unit is configured to calculate a multi-dimensional feature vector including moss thickness index, tooth mark depth, tooth mark number and tongue body fat index based on the three-dimensional point cloud data and the theoretical base plane; An index analyzing unit is configured to compare each index in the multi-dimensional feature vector with a preset threshold to generate preliminary auxiliary syndrome differentiation information. 9.The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis system according to claim 7, characterized in that, The commonality analysis module includes: A data matrix constructing unit is configured to splice the multi-dimensional feature vector of each family member with the auxiliary syndrome differentiation information after One-Hot encoding to obtain a family multi-modal data matrix; An algorithm processing unit is configured to perform consensus clustering algorithm on the family multi-modal data matrix, and repeatedly perform resampling clustering on the data by using K-Means and hierarchical clustering method, and calculate the membership confidence of each member to the class cluster; A screening unit is configured to screen members with membership confidence higher than a set value to form a core member set, and take the body constitution type with the highest proportion in the set as the common body constitution type of the family group; A commonality calculating unit is configured to calculate the arithmetic mean of the multi-dimensional feature vectors of all members in the core member set, and define the mean vector as the common feature vector of the family. 10.The tongue image intelligent acquisition and auxiliary syndrome differentiation analysis system according to claim 7, characterized in that, The individual difference module includes: A data set constructing unit is configured to perform standardization processing on a data set composed of the common feature vector and the multi-dimensional feature vectors of all family members, and calculate the sample covariance matrix of the standardized data set; A deviation calculating unit is configured to calculate the Mahalanobis distance between the standardized multi-dimensional feature vector and the standardized common feature vector for each family member, and the value of the Mahalanobis distance is the comprehensive deviation; A deviation decomposing unit is configured to decompose the comprehensive deviation to the tongue color HSV space value, moss thickness index, tooth mark depth, tooth mark number and tongue body fat index of the multi-dimensional feature vector, and calculate the contribution percentage of each dimension to the comprehensive deviation; A data combining unit is configured to combine the comprehensive deviation and all contribution percentages to generate a structured individualized deviation profile, and the individualized deviation profile is used to generate an individual conditioning scheme subsequently.

Images