System and method for analyzing, managing and controlling tongue surface image anomaly based on multivariate data

Abstract

The invention discloses a multivariate data-based tongue surface image anomaly analysis management and control system and method, and relates to the technical field of medical image processing, and the system comprises the following components: a data collection module which collects a tongue surface image according to a standard specification by adopting professional equipment, collects basic information of a patient through an electronic questionnaire and a docking medical record system, and transmits the basic information to a database; according to the method, macroscopic and microscopic feature information in the lingual surface image is captured through a multi-scale feature extraction technology, so that lingual surface feature information can be described more comprehensively, meanwhile, a comprehensive feature vector is formed by combining multivariate data fusion, potential association among different data can be mined, and the accuracy of the lingual surface feature information is improved. According to the method, the correlation between the gene detection data and the lingual surface image features, especially the correlation between the gene detection data and the lingual surface image features, provides a richer basis for disease diagnosis, analyzes the comprehensive feature vectors through a statistical method and a machine learning algorithm, can more accurately judge whether the lingual surface has abnormal conditions, and greatly improves the accuracy and reliability of diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a tongue surface image abnormality analysis and control system and method based on multivariate data. Background Art

[0002] In the field of medical diagnosis, tongue diagnosis, as a traditional and important diagnostic method, has always received widespread attention. The characteristics of the tongue surface contain rich information about the human health status and can reflect the physiological and pathological status of the human body to a certain extent. With the rapid development of modern science and technology, image processing technology and deep learning technology have made significant progress, providing strong technical support for the use of tongue surface images for disease diagnosis and health assessment. Through image acquisition equipment and algorithms, high-quality tongue surface images can be obtained, and deep learning models can be used to extract and analyze the features in the image, thereby realizing auxiliary diagnosis of the disease. This diagnostic method based on tongue surface images has the advantages of being non-invasive and easy to operate. It provides a new way for early screening and diagnosis of diseases and has become an important direction of current research in medical image processing technology.

[0003] However, existing tongue image analysis methods have many limitations. On the one hand, most methods focus on single-scale features, describing tongue surface feature information only at a macroscopic or microscopic level. This makes it difficult to fully and deeply explore the rich information contained in tongue surface images, resulting in limited ability to identify tongue abnormalities and an inability to accurately capture subtle but potentially diagnostically important features. On the other hand, current systems lack comprehensive analysis and management of multivariate data. In medical diagnosis, in addition to tongue image features, basic patient information (such as age, gender, medical history, etc.) and genetic testing data also contain important diagnostic information. However, existing methods fail to fully integrate this relevant information to form a comprehensive and systematic diagnostic system, thus affecting the accuracy and reliability of diagnosis. In addition, genetic testing technology is playing an increasingly important role in disease diagnosis and personalized treatment, but few studies have combined genetic testing data with tongue image features to explore the potential correlation between the two, failing to provide a strong genetic basis for early disease prediction and personalized treatment. Summary of the Invention

[0004] The purpose of the present invention is to make up for the shortcomings of the existing technology and provide a tongue image abnormality analysis and control system and method based on multivariate data. It can comprehensively and meticulously describe the tongue feature information through multi-scale feature extraction technology, not only focusing on the macroscopic tongue morphology, color and other features, but also deeply exploring the texture, structure and other features at the micro level, thereby enhancing the ability to identify abnormal situations. At the same time, by fusing multivariate data, including tongue image features, genetic test data and patient basic information data, and adopting feature splicing and weighted fusion methods, different features are given corresponding weights according to the importance and correlation of various types of data to form a comprehensive feature vector. Through the analysis of the comprehensive feature vector, the potential correlation between genetic test data and tongue image features can be explored, providing a genetic basis for early prediction of diseases and personalized treatment.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: On the one hand, a tongue surface image abnormality analysis and control system based on multivariate data, the system includes the following components:

[0006] Data collection module: Tongue images are collected using professional equipment according to standard specifications. Basic patient information is collected through electronic questionnaires and connected to the medical record system. Biological samples are collected according to medical procedures and sent to professional institutions to obtain genetic testing data.

[0007] Feature extraction and preprocessing module: Extracts macroscopic and microscopic features from tongue images and performs normalization and dimensionality reduction preprocessing. It also encodes variant sites and normalizes expression levels for genetic testing data, and encodes and formats basic patient information.

[0008] Multivariate data fusion module: This module sequentially combines the pre-processed tongue image, genetic testing, and patient basic information data features, and performs weighted fusion based on data importance and relevance to form a comprehensive feature vector.

[0009] Association analysis and abnormality diagnosis module: This module uses data analysis methods to explore the potential associations between genes and tongue surface features in the fused comprehensive feature vector. Classification technology is used to determine whether the tongue surface is abnormal based on the comprehensive feature vector, and the annotated data is used to train and optimize the model.

[0010] Management and feedback module: When tongue abnormalities are detected, a report containing various results and suggestions is generated and sent to the doctor. A patient health record is established to track and manage data. Recheck reminders are set according to the condition and strategies are adjusted based on new data. Feedback is collected to optimize the system.

[0011] Furthermore, the feature extraction and preprocessing module processes the collected tongue surface images and extracts feature information of the tongue surface at different scales. On the one hand, the macroscopic features of the tongue surface are extracted, including the overall shape of the tongue and the approximate distribution of the color. On the other hand, the microscopic features of the tongue surface are extracted, including the texture details of the tongue coating and the distribution characteristics of the microvessels on the tongue surface. The features extracted at different scales are fused, and the extracted tongue surface image features are preprocessed. At the same time, the genetic detection data are preprocessed. Specifically, a specific coding conversion is performed on the gene mutation sites to convert them into a form that is convenient for computer processing and analysis, and the gene expression level data is normalized to eliminate the scale differences between different gene expression amounts. In addition, the collected basic information of the patients is encoded and formatted.

[0012] Furthermore, the feature extraction and preprocessing module extracts the feature information of the tongue surface at different scales. Specifically, let the tongue surface image I(x, y), where (x, y) is the coordinate of the pixel in the image, define Gaussian kernels of different scales Among them, σ i Represents the standard deviation of the Gaussian kernel at the i-th scale, where i = 1, 2, ..., n is the number of scales. Gaussian filtering of different scales is performed on the tongue image I(x, y) to obtain smoothed images at different scales. The calculation formula is: Where * represents the convolution operation, which calculates the gradient amplitude of the smoothed image at different scales and gradient direction The calculation formula is: And combine the gradient magnitude and gradient direction at different scales into a feature vector Right now Get the feature vectors at different scales.

[0013] Furthermore, the feature extraction and preprocessing module fuses the features extracted at different scales, and the fusion formula is: Among them, F merged is the fused tongue image feature vector, F i is the tongue surface image feature vector extracted at the i-th scale, which is extracted by applying convolution kernels or filtering operations of different scales to the tongue surface image, w i is the importance weight of the i-th scale feature, which is determined based on the feature correlation analysis and feature importance ranking method, reflecting the contribution of the scale feature to the overall feature, and n is the number of scales extracted.

[0014] Furthermore, the multivariate data fusion module sequentially stitches and fuses the pre-processed tongue surface image, genetic testing, and patient basic information data features. The stitching and fusion formula is: Among them, Ftotal is the integrated feature vector after fusion, F′ i is the feature vector of the i-th type of data after preprocessing, H i is the information entropy of the i-th type of data features, which is obtained by calculating the probability distribution of each type of data features. max It is the maximum value of the information entropy of all data features, and m is the number of data categories, that is, m = 3, which are tongue surface image, gene and basic information respectively.

[0015] Furthermore, the association analysis and abnormality diagnosis module uses data analysis methods to conduct in-depth analysis of the comprehensive feature vector obtained after processing, quantifies the degree of association between genes and tongue surface features through correlation calculation, finds the correspondence between gene mutations and specific tongue surface abnormalities, and calculates the abnormality diagnosis risk score based on the fused comprehensive feature vector using the abnormality diagnosis model.

[0016] Furthermore, the association analysis and abnormality diagnosis module quantifies the degree of association between genes and tongue surface features by calculating the degree of association, and the calculation formula for the degree of association is: Among them, R g-t is the correlation index between genetic characteristics and tongue surface characteristics. The larger the value, the stronger the correlation. ij is the element in row i and column j of the gene feature matrix, which comes from the gene feature after multivariate data fusion, b ij is the element in the i-th row and j-th column of the tongue feature matrix, which comes from the tongue feature matrix after multivariate data fusion. p and q are the number of rows and columns of the feature matrix, and d g-t is the distance between the gene feature vector and the tongue feature vector in the feature space, and its calculation formula is: Among them, x i and y i are the i-th component of the gene feature vector and the tongue feature vector, n is the number of feature vectors, and D0 is the distance decay coefficient, which controls the influence of distance on the association degree.

[0017] Furthermore, the association analysis and abnormality diagnosis module calculates the abnormality diagnosis risk score based on the fused comprehensive feature vector using the abnormality diagnosis model, and the calculation formula is: Among them, R risk It is the risk score of tongue abnormality, and its value range is [0, 1]. The closer it is to 1, the greater the possibility of abnormality. is the i-th component of the fused comprehensive feature vector, w i is the weight of the i-th feature component, b is the bias term, and s is the dimension of the comprehensive feature vector.

[0018] On the other hand, a tongue surface image abnormality analysis and control method based on multivariate data, the specific steps of the method are:

[0019] Data collection: Use professional image acquisition equipment to capture tongue images according to standard specifications, collect basic patient information through electronic questionnaires and medical record systems, and collect genetic testing samples and obtain test data;

[0020] Feature extraction and preprocessing: Input tongue surface images into the deep learning model, extract multi-scale features and perform preprocessing, and perform corresponding preprocessing on genetic testing data and patient basic information data;

[0021] Multivariate data fusion: The pre-processed data features are spliced and weighted to form a comprehensive feature vector;

[0022] Association analysis and abnormality diagnosis: Statistical and machine learning methods are used to analyze comprehensive feature vectors, explore the association between genes and tongue surface features, and calculate abnormality diagnosis risk scores to determine whether the tongue surface is abnormal;

[0023] Management and feedback: If an abnormality is detected, an abnormality report will be generated and sent to the doctor. At the same time, the patient's health data will be tracked and managed, and regular review reminders will be given. The management and control strategy will be adjusted based on the review results and genetic information.

[0024] Compared with the existing technology, this tongue surface image abnormality analysis and control system and method based on multivariate data has the following beneficial effects:

[0025] 1. The present invention uses multi-scale feature extraction technology to capture macroscopic and microscopic feature information in tongue surface images, thereby being able to more comprehensively describe tongue surface feature information. At the same time, by combining multivariate data fusion to form a comprehensive feature vector, it is possible to explore potential associations between different data, especially the association between genetic testing data and tongue surface image features, providing a richer basis for disease diagnosis. By analyzing the comprehensive feature vector through statistical methods and machine learning algorithms, it is possible to more accurately determine whether there are abnormalities on the tongue surface, greatly improving the accuracy and reliability of diagnosis, and helping doctors to detect signs of disease earlier.

[0026] 2. The present invention combines genetic testing data with tongue image features to provide a strong basis for early prediction of diseases and personalized treatment from a genetic perspective. By analyzing the potential correlation between genes and abnormal tongue manifestations, doctors can better understand the patient's genetic background and disease risk, take preventive measures in advance, achieve accurate prevention and treatment of diseases, and improve the patient's quality of life and health level.

[0027] Other advantages, objects and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art based on an examination of the following or may be learned from the practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0029] Figure 1 This is a schematic diagram of the structure of a tongue surface image abnormality analysis and control system based on multivariate data;

[0030] Figure 2 The figure is a flowchart of a tongue surface image abnormality analysis and control method based on multivariate data. DETAILED DESCRIPTION

[0031] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0032] Example 1

[0033] 500 suspected diabetic patients aged between 30 and 60 years old with different lifestyle habits (such as dietary preferences, exercise volume, etc.) and family medical history (with or without a family history of diabetes) were selected as research subjects. A high-resolution (resolution not less than 300dpi) medical-grade camera was used, equipped with professional oral lighting equipment with adjustable color temperature and brightness. Before collecting tongue surface images, the patients were asked to rinse their mouths with clean water to remove food debris and odor in the mouth and keep the mouth clean. According to the standard shooting process, tongue surface images were taken from multiple angles such as the front and side, with 3-5 images taken at each angle. Finally, the image with the highest clarity, uniform lighting, complete tongue surface and no reflection was selected for subsequent analysis. When shooting, the shooting distance (10-15 cm from the tongue surface) and angle (perpendicular to the tongue surface) were strictly controlled. The patient's age, gender, family history of diabetes, dietary habits (whether high-sugar, high-fat diet), amount of exercise (weekly exercise time and intensity) and other basic information are collected. The age of onset, severity of the disease and other information of diabetic patients in the family are recorded in detail to facilitate subsequent analysis of the impact of family genetic factors on the onset of diabetes. At the same time, a 5ml fasting venous blood sample is collected from the patient and sent to a qualified professional genetic testing institution for testing. The test items include multiple gene loci closely related to the onset of diabetes, such as the mutation of genes such as TCF7L2, PPARG, KCNJ11, and the expression levels of genes related to insulin resistance. This ensures that the genetic testing process strictly follows the standardized operating procedures to ensure the accuracy and reliability of the test results.

[0034] The collected tongue surface image is processed to extract the characteristic information of the tongue surface at different scales. Specifically, the tongue surface image I(x, y) is assumed, where (x, y) is the coordinate of the pixel in the image, and Gaussian kernels of different scales are defined. where σ i Represents the standard deviation of the Gaussian kernel at the i-th scale, where i = 1, 2, ..., n is the number of scales. Gaussian filtering of different scales is performed on the tongue image I(x, y) to obtain smoothed images at different scales. The calculation formula is: Where * represents the convolution operation, which calculates the gradient amplitude of the smoothed image at different scales and gradient direction The calculation formula is: And combine the gradient magnitude and gradient direction at different scales into a feature vector Right now The feature vectors at different scales are obtained. On the one hand, the macroscopic features of the tongue surface are extracted, including the overall shape of the tongue and the approximate distribution of the color. On the other hand, the microscopic features of the tongue surface are extracted, including the texture details of the tongue coating and the distribution characteristics of the microvessels on the tongue surface. The features extracted at different scales are fused. The fusion formula is: Among them, F merged is the fused tongue image feature vector, F i is the tongue surface image feature vector extracted at the i-th scale, which is extracted by applying convolution kernels or filtering operations of different scales to the tongue surface image, w i is the importance weight of the i-th scale feature, determined based on feature correlation analysis and feature importance ranking, reflecting the contribution of the scale feature to the overall feature set. n is the number of scales extracted. The extracted multi-scale features are normalized, and the feature values are mapped to the interval [0, 1] to eliminate scale differences between different features. Principal component analysis (PCA) is used to reduce the dimensionality of the features, retaining the main components that are most representative of early diabetes diagnosis, reducing the dimensionality and computational complexity of the data while avoiding the impact of correlation between features on subsequent analysis. The genetic testing data are discretized and normalized before preprocessing. Gene variant sites are encoded according to variant type and location and converted into numerical data. Quantile normalization is used for gene expression level data to ensure that they have the same distribution characteristics, facilitating fusion analysis with other data. The basic information of the patients is encoded and formatted. Gender information is binary encoded (0 for male and 1 for female). Information such as dietary habits and exercise amount is quantized and converted into numerical data. Numerical information such as age is standardized to have a mean of 0 and a variance of 1.

[0035] The pre-processed tongue image features, gene detection data features and patient basic information data features are spliced in a certain order to form a long feature vector. In order to better integrate different types of data, the order of feature splicing is determined according to the relevance of the data and its importance for early diagnosis of diabetes. The tongue image features are spliced first, then the gene detection data features, and finally the patient basic information data features. The weights calculated based on information entropy are used. Weighted fusion is performed to form a comprehensive feature vector. When calculating the information entropy, the probability distribution and uncertainty of different data features are taken into account. Through multiple experiments and data analysis, the distribution of weights is optimized so that the fused comprehensive feature vector can more accurately reflect the potential relationship between the patient's health status and the onset of diabetes.

[0036] Using association analysis algorithms Analyzing the association between genes and tongue surface characteristics, in-depth analysis found that certain variations in the TCF7L2 gene were significantly associated with sparse distribution and abnormal morphology of tongue microvessels, and this association was more obvious in patients with a family history of diabetes. At the same time, there was also a certain correlation between the expression level of the PPARG gene and changes in tongue color (such as reddish or dark). Using the abnormal diagnosis model The comprehensive feature vectors were analyzed to determine whether there were any abnormalities on the tongue surface related to diabetes. The results showed that the system detected abnormalities on the tongue surface in 80 patients. After further clinical examination (including blood sugar testing, glycosylated hemoglobin testing, etc.), 70 of them were diagnosed as early-stage diabetes patients, with a diagnostic accuracy rate of 87.5%. At the same time, follow-up observations were conducted on 10 patients who were not diagnosed, and it was found that 3 of them gradually developed symptoms of diabetes in subsequent examinations.

[0037] For patients with tongue abnormalities, the system generates an abnormality report and sends it to the doctor, which includes the results of tongue feature analysis (detailed description of the tongue's color, texture, microvascular distribution and other characteristics and their differences from normal conditions), gene and tongue feature association analysis results (pointing out the specific association pattern between related gene variants and tongue features), abnormality type and personalized treatment recommendations based on genetic information (such as for patients with abnormalities in insulin resistance-related genes, it is recommended to increase insulin sensitivity drug treatment or adjust the diet structure to reduce insulin resistance).

[0038] Doctors develop personalized treatment plans based on the reports. For patients diagnosed with early diabetes, in addition to drug treatment, they also develop detailed diet plans (such as controlling carbohydrate intake, increasing dietary fiber intake, etc.) and exercise plans (such as at least 150 minutes of moderate-intensity aerobic exercise per week). For patients who are undiagnosed but have potential risks, they are advised to improve their lifestyle habits and undergo regular checkups.

[0039] The system tracks and manages the patient's health data and regularly reminds the patient to have a follow-up examination (the review cycle is set at 1-3 months depending on the patient's specific situation). During the follow-up examination, tongue images, genetic test data and basic information data are collected again to analyze changes in the patient's condition. Based on the review results, the treatment plan is adjusted in a timely manner, such as adjusting drug dosages, optimizing diet and exercise plans, etc.

[0040] Example 2

[0041] 600 patients with suspected coronary heart disease aged between 40 and 70 years old, with different lifestyle habits (such as smoking, drinking, exercise frequency and intensity, etc.), family medical history (whether there is a family history of coronary heart disease), and underlying diseases (such as hypertension, hyperlipidemia, etc.) were selected as research subjects. This age group is a high-incidence stage of coronary heart disease. Covering different lifestyle habits and underlying disease conditions can more comprehensively capture factors related to coronary heart disease. A medical-grade camera with high resolution and high color reproduction is used, combined with professional oral lighting equipment to ensure that the tongue image is clear and complete. Before collection, the patient is asked to keep the mouth clean and rest for 10-15 minutes to stabilize the physiological state. When shooting, the patient is asked to naturally extend the tongue, and tongue images are collected from multiple angles (front, left and right sides), taking 3-5 pictures at each angle, and finally selecting the best quality image. At the same time, the shooting distance and lighting conditions are controlled to ensure the consistency of the image.

[0042] Detailed basic information such as the patient's age, gender, family medical history (recording the age of onset, severity of the disease, etc. of patients with coronary heart disease in the family), lifestyle habits (number of cigarettes smoked per day, amount of alcohol consumed, weekly exercise time and method, etc.), duration of underlying disease and treatment status are collected. In addition, the patient's symptoms (such as the frequency, duration, and inducing factors of chest pain and chest tightness) are also collected. A 5ml venous blood sample is collected from the patient and sent to a professional genetic testing institution to detect genes related to the onset of coronary heart disease, such as the LDLR gene (related to blood lipid metabolism), the APOE gene (affecting lipoprotein metabolism), the MTHFR gene (related to homocysteine metabolism), etc., as well as the expression levels of related genes to ensure that the genetic testing process is carried out strictly in accordance with standard operating procedures.

[0043] Advanced image processing technology is used to extract multi-scale features from the collected tongue images. Not only the macroscopic features of the tongue surface, such as the shape of the tongue (whether it is fat, thin, etc.) and color (red, light, purple, etc.) are extracted, but also the microscopic features, such as the texture of the tongue coating (rough, fine, flaking, etc.) and the morphology of the tongue microvessels (whether it is tortuous, dilated, etc.), are extracted. A variety of image processing algorithms and deep learning models are used to capture tongue surface feature information from different levels and angles. The extracted tongue surface image features are normalized to have a unified format and range. A dimensionality reduction algorithm is used to remove redundant features and retain the most valuable features for the early diagnosis of coronary heart disease. For genetic testing data, discretization and normalization are performed, and gene mutation sites and expression levels are converted into numerical forms that are easy to analyze. The patient's basic information is encoded and formatted, and classification information (such as gender, family medical history, etc.) is converted into digital codes. Numerical information (such as age, exercise time, etc.) is standardized.

[0044] The preprocessed tongue image features, genetic test data features, and patient basic information data features are spliced together in a certain logical order to form a comprehensive feature set. The multi-scale features of the tongue image are first spliced together, then combined with the genetic test data features, and finally integrated with the patient basic information data features, so that different types of data can complement and verify each other. According to the importance and relevance of various types of data in the early diagnosis of coronary heart disease, different data features are assigned corresponding weights. Through a large number of experiments and data analysis, a reasonable weight distribution scheme is determined, and the spliced feature set is weightedly fused to form a comprehensive feature vector to comprehensively reflect the potential relationship between the patient's physical condition and coronary heart disease.

[0045] Using professional data analysis methods, we deeply explored the potential correlation between genetic testing data and tongue image features. The analysis found that certain variations in the LDLR gene are associated with the tortuosity of the tongue microvessels and the greasy feeling of the tongue coating. The expression level of the APOE gene is related to the degree of pale purple color of the tongue. At the same time, combined with the patient's basic information, such as long-term smoking and hypertension, we further explored the interaction between these factors and genes and tongue features.

[0046] Using anomaly diagnosis models The comprehensive feature vectors were analyzed to determine whether patients had early signs of coronary heart disease. Abnormalities on the tongue surface were detected in 100 patients. After further clinical examinations (such as electrocardiogram, cardiac ultrasound, coronary angiography, etc.), 80 of them were diagnosed with early coronary heart disease, with a diagnostic accuracy rate of 80%.

[0047] For patients with tongue abnormalities, the system generates a detailed abnormality report, which includes the results of tongue feature analysis (detailed description of the various features of the tongue and the differences from the normal tongue), the results of the gene-tongue feature association analysis (pointing out the specific association pattern between the relevant genes and tongue features), the abnormality type and individualized treatment recommendations based on genetic information (for example, for patients with LDLR gene mutations, it is recommended to adjust the blood lipid treatment plan, increase the dosage of statins or combine with other lipid-lowering drugs; for patients with tongue microvascular tortuosity, it is recommended to improve blood circulation, increase exercise and take blood-activating and blood-stasis-removing drugs, etc.).

[0048] Doctors develop personalized treatment plans based on the reports. In addition to drug treatment, they also develop diet plans for patients (such as controlling fat and cholesterol intake, and increasing vegetable and fruit intake), exercise plans (such as at least 150 minutes of aerobic exercise per week, such as brisk walking, jogging, etc.), and lifestyle adjustment suggestions (such as quitting smoking, limiting alcohol consumption, and maintaining a good mood, etc.).

[0049] The system conducts long-term tracking and management of patients' health data, and regularly reminds patients to undergo follow-up examinations (the review cycle is set at 1-3 months, depending on the severity of the disease and the treatment plan). During the follow-up examination, tongue images, genetic test data, and basic information data are collected again to evaluate the treatment effect and changes in the disease. Based on the review results, the treatment plan is adjusted in a timely manner, such as adjusting drug dosages, changing treatment drugs, or further optimizing lifestyle recommendations.

[0050] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can, without departing from the scope of the technical solution of the present invention, make some changes or modifications to equivalent embodiments using the technical contents disclosed above. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A tongue surface image abnormality analysis and control system based on multivariate data, characterized in that: The system includes the following components: Data collection module: Tongue images are collected using professional equipment according to standard specifications. Basic patient information is collected through electronic questionnaires and connected to the medical record system. Biological samples are collected according to medical procedures and sent to professional institutions to obtain genetic testing data. Feature extraction and preprocessing module: Extracts macroscopic and microscopic features from tongue images and performs normalization and dimensionality reduction preprocessing. It also encodes variant sites and normalizes expression levels for genetic testing data, and encodes and formats basic patient information. Multivariate data fusion module: This module sequentially combines the pre-processed tongue image, genetic testing, and patient basic information data features, and performs weighted fusion based on data importance and relevance to form a comprehensive feature vector. Association analysis and abnormality diagnosis module: This module uses data analysis methods to explore the potential associations between genes and tongue surface features in the fused comprehensive feature vector. Classification technology is used to determine whether the tongue surface is abnormal based on the comprehensive feature vector, and the annotated data is used to train and optimize the model. Management and feedback module: When tongue abnormalities are detected, a report containing various results and suggestions is generated and sent to the doctor. A patient health record is established to track and manage data. Recheck reminders are set according to the condition and strategies are adjusted based on new data. Feedback is collected to optimize the system.

2. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 1, characterized in that: The feature extraction and preprocessing module processes the collected tongue surface images and extracts feature information of the tongue surface at different scales. On the one hand, the macroscopic features of the tongue surface are extracted, including the overall shape of the tongue body and the approximate distribution of the color. On the other hand, the microscopic features of the tongue surface are extracted, including the texture details of the tongue coating and the distribution characteristics of the microvessels on the tongue surface. The features extracted at different scales are fused, and the extracted tongue surface image features are preprocessed. At the same time, the genetic detection data are preprocessed. Specifically, the gene mutation sites are subjected to specific coding conversion and converted into a form that is convenient for computer processing and analysis. The gene expression level data are normalized to eliminate the scale differences between different gene expression amounts. In addition, the collected basic information of the patients is encoded and formatted.

3. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 2, characterized in that: The feature extraction and preprocessing module extracts the feature information of the tongue surface at different scales. Specifically, let the tongue surface image I(x, y), where I(x, y) is the coordinate of the pixel in the image, and define Gaussian kernels of different scales Among them, σ i Represents the standard deviation of the Gaussian kernel at the i-th scale, where i = 1, 2, ..., n is the number of scales. Gaussian filtering of different scales is performed on the tongue image I(x, y) to obtain smoothed images at different scales. The calculation formula is: Where * represents the convolution operation, which calculates the gradient amplitude of the smoothed image at different scales and gradient direction The calculation formula is: And combine the gradient magnitude and gradient direction at different scales into a feature vector That is Get the feature vectors at different scales.

4. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 2, characterized in that: The feature extraction and preprocessing module fuses the features extracted at different scales, and the fusion formula is: Among them, F merged is the fused tongue image feature vector, F i is the tongue surface image feature vector extracted at the i-th scale, which is extracted by applying convolution kernels or filtering operations of different scales to the tongue surface image, w i is the importance weight of the i-th scale feature, which is determined based on the feature correlation analysis and feature importance ranking method, reflecting the contribution of the scale feature to the overall feature, and n is the number of scales extracted.

5. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 1, characterized in that: The multivariate data fusion module sequentially splices and fuses the pre-processed tongue surface image, genetic testing, and patient basic information data features. The splicing and fusion formula is: Among them, F total is the integrated feature vector after fusion, F′ i is the feature vector of the i-th type of data after preprocessing, H i is the information entropy of the i-th type of data features, which is obtained by calculating the probability distribution of each type of data features. max It is the maximum value of the information entropy of all data features, and m is the number of data categories, that is, m = 3, which are tongue surface image, gene and basic information respectively.

6. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 1, characterized in that: The association analysis and abnormality diagnosis module uses data analysis methods to conduct an in-depth analysis of the comprehensive feature vector obtained after processing, quantifies the degree of association between genes and tongue surface features through correlation calculation, finds the correspondence between gene variation and specific tongue surface abnormalities, and calculates the abnormality diagnosis risk score based on the fused comprehensive feature vector using the abnormality diagnosis model.

7. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 6, characterized in that: The association analysis and abnormality diagnosis module quantifies the degree of association between genes and tongue surface features by calculating the degree of association. The calculation formula for the degree of association is: Among them, R g-t is the correlation index between genetic characteristics and tongue surface characteristics. The larger the value, the stronger the correlation. ij is the element in row i and column j of the gene feature matrix, which comes from the gene feature after multivariate data fusion, b ij is the element in the i-th row and j-th column of the tongue feature matrix, which comes from the tongue feature matrix after multivariate data fusion. p and q are the number of rows and columns of the feature matrix, and d g-t is the distance between the gene feature vector and the tongue feature vector in the feature space, and its calculation formula is: Among them, x i and y i are the i-th component of the gene feature vector and the tongue feature vector, n is the number of feature vectors, and D0 is the distance decay coefficient, which controls the influence of distance on the association degree.

8. The tongue surface image abnormality analysis and control system based on multivariate data according to claim 1, characterized in that: The association analysis and abnormality diagnosis module calculates the abnormality diagnosis risk score based on the fused comprehensive feature vector using the abnormality diagnosis model. The calculation formula is: Among them, R risk It is the risk score of tongue abnormality, and its value range is [0, 1]. The closer it is to 1, the greater the possibility of abnormality. is the i-th component of the fused comprehensive feature vector, w i is the weight of the i-th feature component, b is the bias term, and s is the dimension of the comprehensive feature vector.

9. A method for analyzing and controlling abnormalities in tongue surface images based on multivariate data, the method being applicable to a system for analyzing and controlling abnormalities in tongue surface images based on multivariate data as claimed in any one of claims 1 to 8, characterized in that: The specific steps of the method are: Data collection: Use professional image acquisition equipment to capture tongue images according to standard specifications, collect basic patient information through electronic questionnaires and medical record systems, and collect genetic testing samples and obtain test data; Feature extraction and preprocessing: Input tongue surface images into the deep learning model, extract multi-scale features and perform preprocessing, and perform corresponding preprocessing on genetic testing data and patient basic information data; Multivariate data fusion: The pre-processed data features are spliced and weighted to form a comprehensive feature vector; Association analysis and abnormality diagnosis: Statistical and machine learning methods are used to analyze comprehensive feature vectors, explore the association between genes and tongue surface features, and calculate abnormality diagnosis risk scores to determine whether the tongue surface is abnormal; Management and feedback: If an abnormality is detected, an abnormality report will be generated and sent to the doctor. At the same time, the patient's health data will be tracked and managed, and regular review reminders will be given. The management and control strategy will be adjusted based on the review results and genetic information.

Images