Auxiliary diagnosis and treatment method, device and equipment and storage medium

Abstract

The invention relates to an auxiliary diagnosis and treatment method, device and equipment and a storage medium, and the method comprises the steps: verifying the quality of medical data through an intelligent matching mechanism, and generating a supplementary signal containing standardized guidance for unqualified image data in real time; a cross-modal fusion technology is adopted to integrate the tongue image, the pulse image, the image and the inspection data, and a comprehensive feature vector of a unified dimension is constructed; outputting correlation mapping of the traditional Chinese medicine syndromes and the western medicine diseases based on the pathological reasoning model, and generating a drug treatment scheme and a non-drug treatment scheme through a treatment scheme reasoning model; and finally, dynamically optimizing the dialectical logic parameters and the treatment weight parameters according to user feedback. The method solves the problems of difficulty in multi-source medical data fusion, one-sided diagnosis basis and lack of curative effect feedback closed loop, significantly improves the accuracy and adaptability of personalized diagnosis and treatment, and provides standardized intelligent support for traditional Chinese and western medicine combined diagnosis and treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information data processing, and in particular to an auxiliary diagnosis and treatment method, device, equipment and storage medium. Background Art

[0002] With the deep integration of modern medicine and information technology, complementary diagnosis and treatment methods have become a key means of improving medical efficiency and quality. These methods integrate diverse medical data to build intelligent analytical models, providing decision support for physicians and helping patients obtain more precise health management plans. In Traditional Chinese Medicine (TCM), complementary diagnosis and treatment particularly emphasizes the integration of traditional syndrome differentiation theory with modern medical testing technologies. Through objective and standardized analysis, they strive to bridge the diagnostic gap between Chinese and Western medicine and provide patients with personalized, integrated treatment plans.

[0003] However, existing auxiliary diagnosis and treatment technologies still have significant limitations in practical applications. On the one hand, due to the huge differences between Chinese and Western medicine data systems, multi-source heterogeneous medical information such as tongue images, pulse conditions, imaging reports, and biochemical indicators are difficult to effectively integrate, resulting in a one-sided basis for diagnosis; on the other hand, traditional methods lack a dynamic optimization mechanism and are unable to continuously adjust the diagnostic logic and treatment plan weights based on patient feedback, resulting in a deviation between the recommended plan and the actual efficacy. In addition, existing technologies lack quality control over medical data. When the collected data does not meet the diagnostic requirements, there is a lack of effective real-time intervention methods, which can easily lead to the transmission of incorrect diagnosis chains downward. These defects not only reduce the reliability of diagnosis and treatment results, but also hinder the large-scale application of the integrated Chinese and Western medicine diagnosis and treatment model. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide an auxiliary diagnosis and treatment method, device, equipment and storage medium that can realize the intelligent integration of multi-source medical data, dynamically optimize the diagnosis and treatment model, and have the ability to close the loop of data quality.

[0005] The purpose of the present invention is achieved by the following scheme:

[0006] In a first aspect, the present invention provides an auxiliary diagnosis and treatment method, comprising the following steps:

[0007] S1: Obtain the user's diagnosis request and data to be diagnosed, and perform type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data, and test report data;

[0008] S2: Based on the medical imaging data and test report data of the data to be diagnosed, feature extraction and data fusion processing are performed on the target vital sign data. The tongue color features, pulse waveform frequency features and lesion location features are extracted, and the multi-source features are fused through the cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features;

[0009] S3: Input the comprehensive feature vector into the pre-trained pathological reasoning model for syndrome reasoning processing, perform reasoning analysis on the comprehensive feature vector, and output the TCM syndrome type and Western medicine disease association results. The Western medicine disease association results are used to indicate the modern medical diagnosis corresponding to the TCM syndrome;

[0010] S4: Input the TCM syndrome types and Western medicine disease association results into the pre-trained treatment plan inference model for processing, call the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtain personalized recommendation plans including drug treatment plans, diet therapy plans, and exercise therapy plans;

[0011] S5: Optimize the parameters of the pathology reasoning model and the treatment plan reasoning model based on the user's feedback data on the personalized recommendation plan, and update the parameters of the pathology reasoning model and the treatment plan reasoning model.

[0012] In a second aspect, the present invention provides an auxiliary diagnosis and treatment device, which is configured with the following modules:

[0013] The type matching and acquisition module is used to obtain the user's diagnosis request and data to be diagnosed, and perform type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data and test report data;

[0014] The data multi-source fusion module is used to perform feature extraction and data fusion processing on the target vital sign data based on the medical imaging data and test report data to be diagnosed. It extracts the tongue color features, pulse waveform frequency features and lesion location features, and fuses the multi-source features through the cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features;

[0015] The pathology reasoning module is used to input the comprehensive feature vector into the pre-trained pathology reasoning model for syndrome reasoning processing, perform reasoning analysis on the comprehensive feature vector, and output the TCM syndrome type and Western medicine disease association results. The Western medicine disease association results are used to indicate the modern medical diagnosis corresponding to the TCM syndrome;

[0016] The treatment plan recommendation module is used to input the TCM syndrome types and Western medicine disease association results into the pre-trained treatment plan inference model for processing, call the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtain personalized recommendations including drug treatment plans, diet therapy plans, and exercise therapy plans;

[0017] The model parameter optimization module is used to optimize the parameters of the pathology reasoning model and the treatment plan reasoning model based on the user's feedback data on the personalized recommendation plan, and update the parameters of the pathology reasoning model and the treatment plan reasoning model.

[0018] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any one of the above-mentioned auxiliary diagnosis and treatment methods when executing the computer program.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements any of the above-mentioned auxiliary diagnosis and treatment methods when executed by a processor.

[0020] In summary, the auxiliary diagnosis and treatment method provided by the present invention can achieve the synergistic gain effect of the fusion of Chinese and Western medicine data and dynamic optimization through the closed-loop diagnosis and treatment process design. At the data input end, through real-time verification and interactive guidance mechanism, effective control of medical data quality can be achieved, avoiding diagnostic deviations caused by non-standard image acquisition; at the feature processing layer, a multi-source sign deep integration strategy is adopted to achieve the organic unification of Chinese medicine features such as tongue and pulse with Western medicine data such as imaging and test indicators, so as to solve the problem of multimodal medical information islands. In the diagnostic reasoning stage, the two-way reasoning architecture based on the dynamic pathological model can achieve accurate mapping of Chinese medicine syndromes and modern medical diagnosis, overcoming the defect of disconnection between syndrome differentiation and disease differentiation in traditional methods; in the treatment plan generation link, through the drug and non-therapy collaborative recommendation mechanism, accurate adaptation of personalized plans can be achieved, significantly improving treatment compliance. Finally, with the help of a feedback-driven parameter optimization system, the continuous self-evolution capability of the diagnosis and treatment model is realized, so that the syndrome differentiation logic threshold and the treatment plan weight are dynamically adjusted with the clinical efficacy, fundamentally breaking through the technical bottleneck of the static model iteration lag.

[0021] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of an auxiliary diagnosis and treatment method provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of a process for generating a comprehensive feature vector according to an embodiment of the present application;

[0024] Figure 3 A schematic diagram of a process for generating a personalized recommendation solution provided in an embodiment of the present application;

[0025] Figure 4 A schematic structural diagram of an auxiliary diagnosis and treatment device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0026] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate preferred embodiments of the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the disclosure.

[0027] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] In one embodiment, Figure 1 As shown, an auxiliary diagnosis and treatment method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0029] S1: Obtain the user's diagnosis request and data to be diagnosed, and perform type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data and test report data.

[0030] Specifically, when the system receives a user's diagnosis request, which may be a tongue diagnosis request, a face diagnosis request, or a hand diagnosis request, the system immediately starts the data processing process. The system first collects image data of the user's part to be diagnosed. These image data correspond to high-resolution images of the tongue, face, or hand, depending on the different diagnosis requests. The image data must meet specific optical parameter standards, such as light intensity between [X] lux and [Y] lux, and image resolution of not less than [M] pixels × [N] pixels, etc., to ensure the clarity and richness of the image details, providing a high-quality data foundation for subsequent processing. At the same time, the system synchronously obtains medical imaging data and test report data. Medical imaging data covers imaging data of multiple modalities such as X-ray, CT, and MRI, and test report data includes test results of biochemical indicators such as blood and urine. These data are stored in a standardized data format preset by the system to facilitate subsequent unified processing.

[0031] Based on the type of diagnosis request, the system calls the corresponding image recognition algorithm to match the type of image data to be diagnosed. For example, for tongue diagnosis requests, the system uses a tongue image recognition model based on a convolutional neural network (CNN). This model is trained with a large amount of labeled tongue image data and can accurately identify tongue images and distinguish tongue images from non-tongue images. After a successful match, the system further uses a preset feature acquisition algorithm to process the image. For the tongue image, it extracts key feature point information such as tongue color and tongue coating thickness to generate target vital sign data. The data is stored in a structured form and contains detailed information such as feature point coordinates and feature values, providing accurate quantitative basis for subsequent diagnostic analysis.

[0032] S2: Based on the medical imaging data and test report data of the data to be diagnosed, feature extraction and data fusion processing are performed on the target vital sign data, tongue color features, pulse waveform frequency features and lesion location features are extracted, and multi-source features are fused through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features.

[0033] Specifically, for medical imaging data, the system calls different analysis algorithms based on the image type. For example, for X-ray images, it uses a feature extraction algorithm based on edge detection to identify the lesion outline and location information. For CT and MRI images, it uses three-dimensional reconstruction technology combined with a feature point detection algorithm to accurately locate the lesion and extract its morphological characteristic parameters. These lesion location characteristic parameters include detailed information such as the lesion's three-dimensional coordinates, size, and shape, and are stored in a data structure preset by the system. At the same time, the system parses the test report data and extracts the values ​​of key biochemical indicators, such as white blood cell counts, blood sugar levels, and various protein levels in the blood. These biochemical indicator data are also stored in a structured form to facilitate subsequent fusion processing.

[0034] For tongue image data, the system uses a special color feature extraction algorithm. This algorithm is based on color space conversion (such as RGB to HSV) and cluster analysis technology to accurately quantify the color characteristics of the tongue, subdivide the tongue color into multiple levels such as light red, light white, and light purple, and express its saturation and brightness and other parameters in numerical form; for pulse waveform frequency feature extraction, the system uses signal processing technology to perform Fourier transform and other operations on the pulse waveform data, analyze its frequency components, and extract key feature parameters such as the main frequency and bandwidth. These parameters can reflect the speed, strength and other characteristics of the pulse.

[0035] Subsequently, the system uses a cross-modal attention weighted algorithm, which is built on a deep learning framework. By learning the correlation and importance weights between different data modalities (images, videos, test reports, etc.), the system fuses the extracted multi-source features. During the fusion process, the algorithm automatically assigns attention weights. For example, when the tongue color feature is highly correlated with certain inflammatory indicators in the test report, the system will increase the weight of these two types of features in the fusion process, thereby generating a comprehensive feature vector containing multi-dimensional features. This vector integrates multi-source heterogeneous information such as tongue color, pulse waveform frequency, lesion location, and biochemical indicators, and comprehensively and accurately characterizes the patient's health status in the form of a vector, providing comprehensive data input for subsequent pathological reasoning. The entire fusion process is automatically executed by the system according to the preset algorithm to ensure the objectivity and accuracy of data processing.

[0036] S3: Input the comprehensive feature vector into the pre-trained pathological reasoning model for syndrome reasoning processing, perform reasoning analysis on the comprehensive feature vector, and output the TCM syndrome type and Western medicine disease association results. The Western medicine disease association results are used to indicate the modern medical diagnosis corresponding to the TCM syndrome.

[0037] Specifically, the pathology reasoning model is based on a deep neural network architecture. After being jointly trained with a large amount of clinical case data and diagnostic knowledge from integrated traditional Chinese and Western medicine, it has the ability to perform deep reasoning and analysis on comprehensive feature vectors. The pathology reasoning model internally contains multiple neural network layers. For example, the input layer receives the comprehensive feature vector, the hidden layer performs nonlinear transformation and feature learning on the features through activation functions and weight matrices, and the output layer generates the corresponding diagnostic results. During the reasoning process, the model first normalizes the input comprehensive feature vector to ensure that feature data of different dimensions are compared and analyzed within the same numerical range.

[0038] Preferably, the pathology reasoning model performs association analysis on each feature in the feature vector based on the internal preset Chinese and Western medicine diagnostic knowledge graph. For example, when the tongue color feature is pale and the pulse waveform frequency feature shows a deep and thin pulse, the model combines the lesion location features and abnormal biochemical indicators, and through the forward propagation calculation of the neural network, infers that the corresponding Chinese medicine syndrome type is "Qi and Blood Deficiency Syndrome". At the same time, based on the Western medicine disease knowledge associated with the Chinese medicine syndrome, the corresponding Western medicine disease association results are output, such as indicating the possibility of modern medical diagnosis such as "malnutrition anemia".

[0039] During its training, this pathology inference model incorporates extensive experience from integrated Chinese and Western medicine diagnosis and treatment. This model effectively bridges the gap between Chinese and Western medicine diagnoses, providing physicians with a comprehensive diagnostic reference. Its inference results are presented as probabilities, such as the probability of a TCM syndrome type being "Qi and Blood Deficiency Syndrome" being [P]%, and the probability of a Western medicine disease-associated result being "Nutritional Anemia" being [Q]%, helping physicians more intuitively understand the credibility of the diagnosis. Based on the model's output, the system automatically generates a structured diagnostic report containing the TCM syndrome type, Western medicine disease-associated results, and the corresponding diagnostic evidence and explanations. This provides an accurate diagnostic basis for subsequent treatment recommendations. The entire inference process is automatically executed by the system based on the model's algorithm, ensuring the objectivity and consistency of the diagnostic results, without any subjective judgment.

[0040] S4: Input the TCM syndrome types and Western medicine disease association results into the pre-trained treatment plan reasoning model for processing, call the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtain personalized recommendation plans including drug treatment plans, diet therapy plans and exercise therapy plans.

[0041] Specifically, the treatment plan inference model is built based on reinforcement learning and knowledge graph technology. Its knowledge graph integrates rich information from the Traditional Chinese Medicine (TCM) prescription database and the non-drug therapy knowledge base. The TCM prescription database contains thousands of classic TCM prescriptions and detailed information on their constituent drugs, efficacy, and indications. The composition of each prescription clearly indicates the name, dosage range, and preparation method of each drug. The non-drug therapy knowledge base covers various non-drug therapies, including dietary therapy and exercise therapy. Dietary therapy plans include the nutritional content, therapeutic efficacy, applicable symptoms, and specific recipes of different ingredients. Exercise therapy plans include detailed information on the intensity, frequency, applicable population, and expected efficacy of various exercise methods, such as Tai Chi, Ba Duan Jin, and aerobic exercise.

[0042] After receiving the diagnosis results, the treatment plan inference model first performs semantic matching and association queries in the knowledge graph based on the TCM syndrome type and the Western medicine disease association results. For example, when the diagnosis result is the TCM syndrome type "liver depression and qi stagnation syndrome" and the Western medicine disease association result is "mild depression", the model searches the TCM prescription library for TCM prescriptions that have a high correlation with both "liver depression and qi stagnation syndrome" and "mild depression", such as Chaihu Shugan San, etc. At the same time, it analyzes the drug composition and dosage range of the prescription, and fine-tunes it based on the patient's individual characteristics such as age and weight to generate a personalized drug treatment plan. In terms of non-drug therapies, the model selects foods from the dietary therapy plan that have the effects of soothing the liver and regulating qi, relieving depression and calming the mind, such as rose tea and green radish drink, and provides specific recipes and consumption recommendations; selects exercise methods suitable for relieving depression from the exercise therapy plan, such as Tai Chi and yoga, recommends specific exercise frequency and duration, and generates an exercise therapy plan that includes exercise movements and precautions.

[0043] When generating recommendations, the treatment plan reasoning model comprehensively considers the synergy between medication, diet, and exercise therapy to ensure comprehensive and personalized recommendations. The system presents the generated personalized recommendations as a structured document, detailing medication, diet, and exercise treatments to provide patients with comprehensive, specific, and easy-to-use treatment guidance. The entire recommendation process is automatically executed by the system based on preset algorithms and knowledge base content, ensuring the objectivity and scientific nature of the recommendations, without any subjective bias. Medication plans include the name of the Chinese herbal formula, its components, dosage, and decoction method. Dietary plans include a list of ingredients, preparation methods, and frequency of consumption. Exercise plans include the name of the exercise, key movements, and an exercise plan.

[0044] S5: Optimize the parameters of the pathology reasoning model and the treatment plan reasoning model based on the user's feedback data on the personalized recommendation plan, and update the parameters of the pathology reasoning model and the treatment plan reasoning model.

[0045] Specifically, when users provide feedback on the personalized recommendation plans provided by the system, the feedback data is diverse in form and rich in content, including quantitative evaluation of treatment effects, detailed description of adverse reactions during treatment, compliance with the recommended plans, and the user's subjective feelings during the treatment process.

[0046] After receiving this feedback data, the system first performs preprocessing, including data cleaning, unified format conversion, and semantic analysis, to ensure the accuracy and usability of the feedback data. Subsequently, the system initiates the parameter optimization process and adjusts the relevant parameters of the pathology reasoning model and the treatment plan reasoning model based on the preset machine learning optimization algorithm. During the optimization process, the system constructs a loss function, using the treatment effect evaluation in the feedback data as the main target variable, and adverse reactions and compliance as constraints. Through iterative calculations, the model parameters are gradually adjusted to minimize the deviation between the predicted diagnostic results and the actual treatment effects, while reducing the probability of adverse reactions and improving the acceptability and feasibility of the plan.

[0047] For example, if feedback data indicates that a certain patient group commonly experiences stomach discomfort after using a specific Traditional Chinese Medicine (TCM) formula, the system will adjust the weighting parameters for the syndrome characteristics of this patient group in the pathology inference model accordingly. It will also reduce the recommendation weight for this TCM formula in the treatment inference model, while increasing the probability of recommending alternative formulas. After multiple rounds of optimization iterations, the system generates updated model parameters and replaces the original ones, using the optimized model as the basis for new diagnostic and treatment recommendations. The system regularly verifies and evaluates the updated model, using methods such as cross-validation and new clinical data to test its accuracy and reliability in real-world applications. Through this continuous optimization mechanism, the system continuously improves the accuracy of patient health diagnoses and the rationality of treatment recommendations, better meeting patients' medical needs and enhancing the quality and reliability of auxiliary diagnosis and treatment. The entire optimization process is automatically executed by the system based on pre-set algorithms and procedures, ensuring the objectivity and scientific nature of parameter updates and the absence of any subjective intervention.

[0048] In summary, the auxiliary diagnosis and treatment method provided by the present invention can achieve the synergistic gain effect of the fusion of Chinese and Western medicine data and dynamic optimization through the closed-loop diagnosis and treatment process design. At the data input end, through real-time verification and interactive guidance mechanism, effective control of medical data quality can be achieved, avoiding diagnostic deviations caused by non-standard image acquisition; at the feature processing layer, a multi-source sign deep integration strategy is adopted to achieve the organic unification of Chinese medicine features such as tongue and pulse with Western medicine data such as imaging and test indicators, so as to solve the problem of multimodal medical information islands. In the diagnostic reasoning stage, the two-way reasoning architecture based on the dynamic pathological model can achieve accurate mapping of Chinese medicine syndromes and modern medical diagnosis, overcoming the defect of disconnection between syndrome differentiation and disease differentiation in traditional methods; in the treatment plan generation link, through the drug and non-therapy collaborative recommendation mechanism, accurate adaptation of personalized plans can be achieved, significantly improving treatment compliance. Finally, with the help of a feedback-driven parameter optimization system, the continuous self-evolution capability of the diagnosis and treatment model is realized, so that the syndrome differentiation logic threshold and the treatment plan weight are dynamically adjusted with the clinical efficacy, fundamentally breaking through the technical bottleneck of the static model iteration lag.

[0049] In one embodiment, S1 of an auxiliary diagnosis and treatment method provided by the present invention specifically includes the following steps:

[0050] S11: Obtain the user's diagnosis request and the data to be diagnosed, perform image type recognition processing on the image data to be diagnosed through a preset image recognition model, determine whether the image data meets the requirements of the diagnosis request, and generate an image matching result.

[0051] Specifically, the system processes image data using a pre-set image recognition model. This model, built on deep learning technology and trained on a large amount of diverse image data, possesses powerful image type recognition capabilities. The system inputs the image data to be diagnosed into the model, which then analyzes the image's multi-dimensional information, including color distribution, texture characteristics, and shape, using its internal pre-set feature extraction algorithms.

[0052] At the same time, the system will obtain the specific content of the diagnosis request, such as a tongue diagnosis request, a facial diagnosis request, or a hand diagnosis request, to clarify the specific image type required for this diagnosis and the corresponding anatomical features. The model compares the analysis results of the image data with the requirements of the diagnosis request to determine whether the image data meets the type and quality standards required for diagnosis. This judgment process is based on preset image type matching rules and quality assessment indicators, such as whether the image resolution reaches [X] pixels or more, whether the contrast meets [Y] levels, etc. Ultimately, the system generates an image matching result, which contains the judgment information of whether it matches and the relevant matching measurement value, providing a basis for subsequent data processing.

[0053] S12: Process the image matching results that do not meet the diagnosis request, generate a data supplement signal and send it to the user terminal. The data supplement signal is used to instruct the user terminal to display a dynamic schematic diagram of the target acquisition part and a standardized shooting guide to the user through a display medium.

[0054] Specifically, the system generates a data supplement signal based on preset logical rules. The signal follows a specific communication protocol to ensure that it can be accurately transmitted to the user terminal. The data supplement signal carries detailed information about the target acquisition part. The system will retrieve the corresponding target acquisition part dynamic diagram from the pre-stored dynamic diagram library according to the part corresponding to the diagnosis request, such as the tongue dynamic diagram corresponding to the tongue diagnosis request, to show the shape and position changes of the tongue in different states, so that the user can accurately understand the acquisition requirements. At the same time, the system will also select guide content that matches the target acquisition part from the standardized shooting guide library. The guide content contains detailed information such as lighting requirements, shooting angle recommendations, and shooting distance recommendations during shooting. It is displayed to the user in the form of pictures and texts through the user terminal's mobile phone screen, tablet computer and other display media to ensure that the user can re-collect image data in accordance with standard requirements and improve the accuracy and availability of image data.

[0055] S13: Based on the convolutional neural network, the image matching results that meet the diagnostic request are processed to identify the locations of key anatomical landmarks, extract the tongue color features, pulse waveform frequency features and lesion location features, and fuse the multi-source features through the cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features as the target vital sign data.

[0056] Specifically, the convolutional neural network model, after being trained with professional medical image data, can accurately identify the locations of key anatomical landmarks in an image. For example, in a tongue image, it can accurately identify the coordinate positions of the tongue tip, tongue edge, tongue center, and tongue root. In a facial diagnosis image, it can locate key facial features such as the cheekbone, nose tip, and chin. While identifying key anatomical landmarks, the system activates a feature extraction algorithm. Based on the tongue color characteristics, it uses color space separation technology to decompose the image into different color channels. Cluster analysis and color statistics methods are used to quantify the proportion and distribution of red, white, purple and other components of the tongue color, generating accurate tongue color feature parameters. For the pulse waveform frequency characteristics, the system uses signal processing technology to pre-process the pulse image to remove noise interference. It then extracts the frequency components of the pulse waveform through methods such as fast Fourier transform and calculates feature parameters such as the main frequency and bandwidth.

[0057] In terms of lesion location feature extraction, the system combines medical image analysis technology and the location information of anatomical landmarks to accurately determine the three-dimensional coordinates and morphological features of the lesion in the image. Subsequently, the system uses a cross-modal attention weighted algorithm, which is built on a multimodal data fusion framework. By learning the correlation and importance weights between different feature types such as color, waveform frequency, and position, the extracted multi-source features are fused. During the fusion process, the algorithm automatically assigns attention weights so that each feature can be effectively integrated according to its contribution to the diagnosis. Ultimately, a comprehensive feature vector containing multidimensional features is generated as the target vital sign data, providing comprehensive and accurate quantitative information for subsequent diagnostic analysis. The entire process is automatically executed by the system based on preset algorithms and procedures to ensure the objectivity and accuracy of data processing.

[0058] In one embodiment, Figure 2 As shown, S2 of an auxiliary diagnosis and treatment method provided by the present invention specifically includes the following steps:

[0059] S21: Perform feature extraction processing on the medical imaging data to be diagnosed, analyze the changing trend of organ morphological abnormality indicators, and generate an image feature set containing quantitative features of the organ structure.

[0060] Specifically, the system uses professional image analysis algorithms, such as semantic segmentation and three-dimensional reconstruction technology based on deep learning, to perform feature extraction and processing on medical imaging data to be diagnosed. The system can identify and extract morphological features such as organ size, shape, edge clarity, and internal structural uniformity, such as calculating the volume of lung nodules in lung CT images and the morphological irregularity of the liver in liver MRI images. The system will also analyze the changing trends of these features over time, and through curve fitting and statistical analysis methods, determine whether the organ morphological abnormality indicators deviate from the normal range and the rate of change. Ultimately, the system integrates these quantified morphological features into an image feature set, which includes structural quantitative features such as organ volume, surface area, and shape regularity, providing accurate imaging basis for subsequent diagnosis. The entire process is automatically executed by the system according to the preset algorithm to ensure the objectivity and accuracy of data processing, without any subjective intervention.

[0061] S22: Quantify the test report data of the diagnostic data, mark the clinical risk level of the abnormal biochemical indicators, and generate a test feature set including a test abnormality score matrix. The test feature set is used to indicate the progression stage of the physiological function disorder.

[0062] Specifically, the system sets normal reference ranges and risk level assessment standards for each indicator based on the built-in medical knowledge base and clinical guidelines. For abnormal indicators that exceed the normal range, the system calculates their clinical risk level through a specific risk assessment model. For example, based on the degree of increase in blood sugar levels, combined with factors such as the patient's age and medical history, it assesses the risk level of diabetic complications. The system integrates these quantified indicators and corresponding risk levels into a test feature set to form an abnormality scoring matrix. This matrix presents the degree of abnormality and risk level of each biochemical indicator in matrix form. By analyzing the long-term change trend of these indicators and using statistical analysis methods to determine the progression stage of physiological dysfunction, it provides doctors with a quantitative assessment of the patient's physiological state to assist in judging the progression of the disease. During the processing process, the system operates strictly in accordance with preset rules and algorithms to ensure the objectivity and consistency of data processing.

[0063] S23: Perform data fusion processing on the target vital sign data, imaging feature set and test feature set, balance the clinical value weights of different data sources, and generate a comprehensive feature vector that integrates multi-dimensional vital sign indicators.

[0064] Specifically, the system determines the relative importance of imaging features, test features, and physical sign data in diagnosing specific diseases based on the results of a large number of clinical case analyses. For example, when diagnosing heart disease, the test features of the electrocardiogram have a higher weight. During the fusion process, the system integrates the key features of each data set through mathematical methods such as weighted summation to form a comprehensive feature vector. This vector contains multi-dimensional physical sign indicators with adjusted weights, which comprehensively reflect the patient's health status. The integrated data is processed by dimensionality reduction to remove redundant information, highlight key features, optimize model input, and improve diagnostic efficiency. Preferably, the comprehensive feature vector that integrates multi-dimensional physical sign indicators is generated by the following steps:

[0065] S231: Perform feature alignment processing on the target vital sign data, imaging feature set, and test feature set to unify the quantitative benchmarks of different data sources, convert the measurement units according to clinical diagnostic standards, and generate a standardized feature group with a unified dimension.

[0066] Specifically, when the system performs feature alignment processing on the target vital sign data, imaging feature set, and test feature set, it first performs a detailed analysis and comparison of the features of different data sources. Based on the preset clinical diagnostic standards, the system determines a unified quantitative benchmark and unifies the measurement units of different data sources through a specific conversion algorithm. For example, the organ volume in the imaging features is converted from cubic centimeters to standard medical units, and the biochemical indicators in the test report are converted from the reporting units of different laboratories to international standard units. The system automatically calls the preset conversion rules and algorithms to perform precise unit conversion and quantitative adjustment on each feature, ensuring that all features are compared and integrated under the same dimension, and generating a standardized feature group with a unified dimension. The entire alignment process is automatically performed by the system according to preset rules and algorithms to ensure the objectivity and consistency of data processing. There is no subjective intervention. All operations are based on pre-defined procedures and algorithms, providing a standardized data foundation for subsequent feature fusion.

[0067] S232: Perform weight allocation processing on the standardized feature groups, evaluate the clinical importance levels of different physical signs and features, and determine the contribution weight distribution that reflects the priority of TCM syndrome differentiation.

[0068] Specifically, the system assigns weights to standardized feature groups. Drawing on a built-in medical knowledge base and extensive clinical diagnostic case analysis, it assesses the clinical importance of different physical features in diagnosing specific diseases. Based on the theoretical framework of Traditional Chinese Medicine (TCM) syndrome differentiation, the system determines a contribution weight distribution that reflects the priorities of TCM syndrome differentiation. For example, the system identifies the high weight of tongue color in diagnosing spleen and stomach diseases, and the critical weight of lesion location in diagnosing neoplastic diseases. Using a pre-set weight calculation model, the system automatically calculates the contribution weight of each feature, taking into account its clinical relevance and diagnostic sensitivity. This weight allocation process adheres to the principles of TCM syndrome differentiation and treatment, ensuring that the weight distribution accurately reflects the priorities of TCM syndrome differentiation while also addressing the needs of Western medicine diagnosis. The system strictly adheres to pre-set rules and algorithms during the weight allocation process, ensuring objectivity and scientific rationality. No subjective judgment or selection is involved. All weights are determined based on the system's internal algorithms and data, providing a precise weighting basis for subsequent feature fusion.

[0069] S233: Based on the contribution weight distribution, the standardized feature group is subjected to feature fusion processing, the diagnostic value of multi-source physical sign indicators is integrated, and a comprehensive feature vector is generated.

[0070] Specifically, when the system performs feature fusion processing on the standardized feature group based on the contribution weight distribution, it adopts feature fusion algorithms, such as weighted average and principal component analysis, to integrate the vital signs indicators from different sources. The system fuses the various features in the standardized feature group by weighted summation according to the contribution weight of each feature, highlighting the influence of high-weight features and reducing the interference of low-weight features. During the fusion process, the system automatically adjusts the weight parameters to ensure that the fused feature vector can fully reflect the patient's health status while retaining the diagnostic value of key features. The system normalizes the fused feature vector so that its numerical range adapts to the input requirements of the subsequent diagnostic model and generates the final comprehensive feature vector. The entire feature fusion process is automatically executed by the system according to the preset fusion algorithm to ensure the objectivity and reliability of the fusion results. There is no subjective intervention behavior. All operations are based on pre-defined procedures and algorithms to provide comprehensive and accurate multi-dimensional vital signs data for subsequent diagnostic reasoning.

[0071] The auxiliary diagnosis and treatment method provided above can achieve deep coordination of the diagnostic basis of traditional Chinese medicine and Western medicine through a multi-level medical data processing mechanism. In the image feature extraction stage, by quantifying organ morphological abnormality indicators, it is possible to objectively characterize tissue structural lesions and provide visual support for TCM viscera differentiation. In the test data analysis stage, by annotating biochemical risk levels and constructing a scoring matrix, it is possible to achieve standardized grading of the degree of physiological dysfunction, which is used to make up for the lack of quantitative basis in traditional syndrome differentiation. Finally, through a dynamic weight fusion mechanism, after balancing the clinical value of TCM features such as tongue and pulse conditions with Western medicine data such as imaging tests, a correlation mapping of cross-modal physical signs indicators can be established, solving the technical obstacle of difficult unified evaluation of multi-source medical data.

[0072] The entire process can achieve three core benefits: first, it can convert organ structural changes identified by Western medical imaging into an objective basis for TCM syndrome judgment, thereby strengthening the correlation between "disease differentiation" and "syndrome differentiation"; second, it can quantify the stage of disease progression through risk grading of test indicators, providing evidence-based support for the principle of "treating symptoms in an emergency and treating the root cause in a non-critical manner"; third, the adaptive fusion strategy based on clinical value weights can break through the integration barriers between TCM qualitative descriptions and Western medicine quantitative data, forming a comprehensive diagnostic vector with clinical interpretability, and establishing a cross-scale medical evidence chain for subsequent precise syndrome differentiation.

[0073] In one embodiment, S3 of an auxiliary diagnosis and treatment method provided by the present invention specifically includes the following steps:

[0074] S31: Perform syndrome feature mapping on the comprehensive feature vector, divide the deficiency, excess, cold and heat pathological attribute categories, and generate syndrome feature distribution including syndrome quantitative score.

[0075] Specifically, after obtaining the comprehensive feature vector, the system constructs a deep learning-based semantic mapping network. This network, trained on a large number of TCM classics, clinical cases, and modern medical data, is capable of understanding and translating the pathological features of integrated Chinese and Western medicine. The system then compares the multidimensional feature components of the comprehensive feature vector against predefined TCM syndrome templates. These templates cover typical characteristic descriptions of common syndromes such as "Qi deficiency," "blood stasis," "cold and dampness," and "heat toxicity." Each template is associated with a series of quantifiable characteristic parameters. For example, a "Qi deficiency" syndrome template might include characteristic parameters such as a weak pulse, a pale and plump tongue, and an elevated fatigue index. The system calculates a similarity measure between the comprehensive feature vector and each syndrome template, using the cosine similarity algorithm combined with Euclidean distance to quantify the intensity of each syndrome feature in the patient, generating a syndrome feature distribution containing a quantitative syndrome score. Each syndrome score reflects the degree of match between the patient's current health status and the specific syndrome, providing a quantitative basis for subsequent syndrome diagnosis.

[0076] S32: Classify the distribution of syndrome characteristics, divide the syndrome types according to the TCM syndrome diagnosis standard, and generate TCM syndrome types.

[0077] Specifically, the system uses a syndrome classification model built based on decision trees and Bayesian classifiers based on the TCM syndrome diagnostic criteria. This model integrates authoritative textbooks such as "Diagnostics of Traditional Chinese Medicine" and the consensus of clinical experts, and matches the quantitative scores in the syndrome feature distribution with the syndrome diagnostic criteria one by one. For example, if the patient's "Qi deficiency" syndrome quantitative score exceeds the set threshold and is accompanied by a significant increase in the "blood stasis" syndrome score, the system will classify the patient as "Qi deficiency and blood stasis syndrome" based on the diagnostic criteria of "Qi deficiency and blood stasis syndrome". The system will comprehensively consider the interaction and synergistic effects of multiple syndrome features, determine the most likely TCM syndrome type through probabilistic reasoning, and generate the corresponding diagnosis results. This process not only relies on the strength of a single syndrome feature, but also takes into account the mutual correlation and conversion rules between syndromes to ensure the accuracy and comprehensiveness of the diagnostic results, providing clinicians with a clear direction for TCM syndrome diagnosis.

[0078] S33: Perform disease association analysis on the distribution of syndrome characteristics, locate the corresponding relationship of target organ lesions, and determine the correlation mapping relationship between TCM syndromes and Western medicine diseases.

[0079] Specifically, the system calls a pre-built knowledge graph for the association of Chinese and Western medicine diseases. This graph integrates clinical guidelines, medical literature, and expert experience, and stores the complex associations between Chinese medicine syndromes and Western medicine diseases in the form of a graph structure. The system uses the key syndrome features in the syndrome feature distribution as query nodes and conducts a deep traversal search in the knowledge graph. For example, for "Qi deficiency and blood stasis syndrome", the system will search the knowledge graph for associated Western medicine disease nodes, such as "ischemic cardiovascular and cerebrovascular disease" and "chronic fatigue syndrome". By analyzing the weights of the edges in the graph and the credibility of the association paths, the system determines the association mapping relationship between Chinese medicine syndromes and Western medicine diseases. This process involves a comprehensive analysis of a large amount of clinical data and experimental research results to ensure that the established associations have a scientific basis and clinical reference value, providing strong support for the diagnosis of diseases that integrate Chinese and Western medicine.

[0080] S34: Perform clinical risk analysis on the association mapping relationship, quantify the degree of organ function damage, and generate Western medicine disease association results including disease codes and risk levels.

[0081] Specifically, the system is based on a Western medicine disease risk assessment model that incorporates multi-dimensional information such as disease severity, prognosis, and complication risk. The system first invokes an organ function impairment assessment algorithm, combining organ morphological abnormalities and biochemical abnormalities in the comprehensive feature vector to quantify the extent of target organ function impairment in the patient. For example, for patients with ischemic cardiovascular and cerebrovascular disease, the system analyzes indicators such as changes in myocardial wall thickness and decreased ventricular ejection fraction in cardiac imaging data, as well as elevated myocardial enzyme levels in blood tests, to calculate a quantitative score for cardiac function impairment. Furthermore, the system categorizes disease risk levels into low, medium, and high levels based on clinical risk grading standards. Based on the organ function impairment score and the potential risk of disease progression, the system generates a Western medicine disease association result, including a disease code and risk level. Disease codes adhere to the internationally recognized ICD coding standards to ensure standardization and universality of the results, while the risk level provides clinicians with a visual indication of the severity of the disease, assisting them in developing more precise treatment strategies and interventions.

[0082] In one embodiment, Figure 3 As shown, S4 of an auxiliary diagnosis and treatment method provided by the present invention specifically includes the following steps:

[0083] S41: Match TCM prescriptions to TCM syndrome types, screen core medicinal material combinations and their compatibility relationships, and generate initial drug treatment plans.

[0084] Specifically, when the system matches TCM prescriptions to TCM syndrome types, it first calls upon a preset TCM prescription knowledge base, which contains detailed information on thousands of classic TCM prescriptions and their core herbal combinations, compatibility relationships, and indications. Based on the correspondence between TCM syndrome types and TCM prescriptions, the system screens out the core herbal combination that best matches the current syndrome type through feature matching and similarity calculations. For example, for "Qi Deficiency Syndrome," the system matches prescriptions containing core herbs such as Astragalus, Codonopsis, and Atractylodes.

[0085] The system automatically analyzes the properties, flavors, meridians, efficacy, and compatibility of these herbs, ensuring that their synergistic effects are effective in targeting the patient's syndrome type. During the matching process, the system adheres to the TCM principles of monarch, minister, assistant, and envoy compatibility to optimize herbal combinations and generate an initial drug treatment plan. The system operates according to preset rules and algorithms, ensuring the objectivity and scientific nature of the matching between syndrome types and TCM prescriptions. All operations are based on predefined procedures and algorithms, providing precise drug recommendations for subsequent treatment optimization.

[0086] S42: Map the results of Western medicine disease association to non-drug therapies, match acupoint stimulation plans and their operation specifications, and generate exercise treatment plans.

[0087] Specifically, when mapping Western medicine disease-related results to non-drug therapies, the system utilizes a pre-set non-drug therapy knowledge base, which covers a variety of non-drug therapies, including detailed information on acupoint stimulation, exercise therapy, and dietary therapy. Based on the disease code and risk level in the Western medicine disease-related results, the system uses a matching algorithm to identify the non-drug therapies associated with the disease. For acupoint stimulation protocols, the system matches the corresponding acupoints and their corresponding procedural specifications. For example, for "mild depression," the system matches acupoints such as Shenmen and Sanyinjiao, and provides corresponding acupoint stimulation protocols, including detailed parameters such as stimulation intensity, frequency, and duration. During the matching process, the system integrates modern medical research findings with Traditional Chinese Medicine meridian theory to ensure the scientific and effective nature of the acupoint stimulation protocols. The system operates according to pre-set rules and algorithms, ensuring the objectivity and reliability of the non-drug therapy mapping. No subjective judgment or selection is involved. All operations are based on pre-defined procedures and algorithms, generating exercise therapy and acupoint stimulation protocols with detailed procedural specifications, providing patients with a comprehensive range of treatment options.

[0088] S43: Optimize the initial drug treatment plan, adjust the medicinal ingredients based on the physical contraindications, and generate an integrated treatment plan.

[0089] Specifically, when the system optimizes the initial drug treatment plan, it calls the preset physical contraindication knowledge base based on the patient's physical contraindication characteristics and the risk level in the disease association results. The knowledge base contains a list of contraindication medicinal materials and adjustment suggestions for different physical types (such as yang deficiency constitution, yin deficiency constitution, etc.). The system automatically screens the contraindication medicinal materials in the initial drug treatment plan based on the patient's physical information, and optimizes the medicinal material components according to the adjustment suggestions. For example, for patients with yang deficiency constitution, the system will reduce the dosage of cold medicinal materials or replace them with warm medicinal materials to avoid aggravating the physical bias. During the optimization process, the system fully considers the synergistic effects of drugs and potential drug interactions to ensure the safety and effectiveness of the treatment plan. The system operates according to preset rules and algorithms to ensure the objectivity and scientific nature of the optimization process. There is no subjective intervention behavior. All adjustments are based on pre-defined procedures and algorithms to generate a final integrated treatment plan to provide patients with personalized and safe drug treatment plans. Preferably, the integrated treatment plan is generated by the following steps:

[0090] S431: Perform constitution feature mapping processing on TCM syndrome types, identify constitution classification features such as Qi deficiency, dampness and heat, and determine the user's constitution type.

[0091] Specifically, when the system performs constitution feature mapping processing on TCM syndrome types, it calls the preset constitution feature mapping model. This model integrates the theoretical system of TCM constitution classification and covers common constitution types, such as Qi deficiency constitution, damp-heat constitution, etc. The system performs pattern recognition and feature matching based on the correspondence between the patient's syndrome feature distribution and the constitution feature mapping model to identify the classification features related to the patient's constitution. The model will perform feature comparison to identify the constitution type with the highest matching degree with characteristics such as Qi deficiency constitution, damp-heat constitution, and finally determine the user's constitution type. The system strictly follows the preset algorithms and rules to ensure the objectivity and accuracy of the mapping process. All operations are based on pre-defined procedures and algorithms, and there is no subjective intervention.

[0092] S432: Based on the user's physical condition, the initial drug treatment plan is screened for contraindications, inappropriate medicinal components are filtered out, and an optimized drug treatment plan is generated.

[0093] Specifically, the system calls a preset constitution contraindication knowledge base, which contains a list of contraindicated medicinal materials and adjustment suggestions corresponding to different constitution types. Based on the user's constitution type and contraindication knowledge base, the system automatically screens inappropriate medicinal material components in the initial drug treatment plan. The system filters out those medicinal materials that are not suitable for the user's constitution or may cause adverse reactions, and optimizes the drug plan based on the adjustment suggestions in the knowledge base to generate an optimized drug treatment plan. During the screening process, the system operates strictly in accordance with the preset rules and algorithms to ensure the objectivity and scientific nature of the optimization process, without any subjective judgment or selection behavior.

[0094] S433: Supplement the optimized drug treatment plan with dietary therapy, add a combination of medicinal and edible ingredients to generate an integrated treatment plan.

[0095] Specifically, the system calls the preset diet therapy knowledge base, which covers detailed information on food ingredients with medicinal and edible properties and their conditioning effects. Based on the user's constitution type and syndrome type, the system screens out food combinations with targeted conditioning effects from the knowledge base. The system analyzes the nature, flavor, meridians and conditioning effects of these ingredients to ensure that they can complement the drug treatment plan and jointly promote the patient's recovery. The system automatically adds these food combinations to the optimized drug treatment plan to generate an integrated treatment plan. In the process of diet therapy supplementation, the system follows the principles of traditional Chinese medicine diet therapy and modern nutrition theory to ensure the rationality and effectiveness of the food combination. All operations are based on pre-defined procedures and algorithms. There is no subjective intervention behavior, providing patients with comprehensive and personalized treatment plans.

[0096] S44: Prioritize the integrated treatment plan and exercise therapy plan, set the implementation order of the treatment measures, and generate personalized recommendation plans.

[0097] Specifically, the system analyzes the mechanism of action and time characteristics of drug therapy and non-drug therapy in alleviating disease symptoms and improving physiological functions. For patients with acute diseases or severe symptoms, the system gives priority to recommending treatment measures that can take effect quickly, such as emergency drugs in drug treatment or emergency acupoint plans in acupoint stimulation. For patients with chronic diseases or in rehabilitation, the system pays more attention to long-term efficacy and improvement of quality of life, and gives priority to recommending exercise therapy plans and Chinese medicine conditioning plans. During the sorting process, the system comprehensively considers the feasibility of treatment measures and patient compliance to ensure the rationality and acceptability of treatment plans. The system operates according to preset rules and algorithms to ensure the objectivity and scientific nature of storage. There is no subjective judgment or selection behavior. All operations are based on pre-defined procedures and algorithms to generate personalized recommendation plans with detailed implementation sequences, providing patients with comprehensive, personalized and orderly treatment guidance.

[0098] The auxiliary diagnosis and treatment method provided by this application can achieve deep coordination and individualized adaptation of Chinese and Western medical treatment methods through a step-by-step treatment plan generation mechanism. In the drug plan construction stage, the core medicinal material compatibility relationship is screened based on the TCM syndrome type, which can achieve the goal of precise medication corresponding to the prescription and syndrome, and ensure the effective implementation of the prescription principle of monarch, minister, assistant and envoy; in the non-drug therapy design link, the acupuncture point stimulation plan is mapped by the Western medicine disease association results, which can achieve the organic integration of modern medical diagnosis and traditional external treatment technology to break through the efficacy limitations of single therapy. Subsequently, the initial drug plan is dynamically optimized through the characteristics of physical contraindications, and a medicinal material component adjustment mechanism tailored to the individual can be established to solve the defect of traditional medication ignoring individual differences; finally, with the help of treatment priority sorting strategy, by setting the implementation sequence of drugs and non-drug therapies, a dynamic intervention path of "treating symptoms in emergencies and treating the root causes in slowness" can be formed.

[0099] The method of the present application can achieve triple clinical benefits: first, it combines the prescription compatibility guided by the holistic view of traditional Chinese medicine with the non-drug therapy supported by modern evidence-based medicine to build a complementary treatment system; second, it achieves dual guarantees of drug safety and effectiveness through physical contraindication screening, thereby reducing drug-induced risks; third, it sets the execution order of treatment measures based on the stage of disease development, which can optimize the allocation of medical resources and improve patient compliance, providing an operational implementation plan for integrated medical practice.

[0100] Preferably, if Figure 4 As shown, the present invention provides an auxiliary diagnosis and treatment device 600, which is configured with the following modules:

[0101] Type matching and acquisition module 610 is used to obtain the user's diagnosis request and data to be diagnosed, and perform type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data, and test report data;

[0102] The data multi-source fusion module 620 is used to perform feature extraction and data fusion processing on the target vital sign data based on the medical imaging data and test report data to be diagnosed, extract tongue color features, pulse waveform frequency features, and lesion location features, and fuse the multi-source features through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features;

[0103] Pathology reasoning module 630 is used to input the comprehensive feature vector into a pre-trained pathology reasoning model for syndrome reasoning processing, perform reasoning analysis on the comprehensive feature vector, and output the TCM syndrome type and Western medicine disease association results. The Western medicine disease association results are used to indicate the modern medical diagnosis corresponding to the TCM syndrome;

[0104] The treatment plan recommendation module 640 is used to input the TCM syndrome types and Western medicine disease association results into the pre-trained treatment plan inference model for processing, call the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtain personalized recommendations including drug treatment plans, diet therapy plans, and exercise therapy plans;

[0105] The model parameter optimization module 650 is used to optimize the parameters of the pathology reasoning model and the treatment plan reasoning model based on the user's feedback data on the personalized recommendation plan, and update the parameters of the pathology reasoning model and the treatment plan reasoning model.

[0106] In summary, the auxiliary diagnosis and treatment device provided by the present invention can achieve the synergistic gain effect of the fusion of Chinese and Western medicine data and dynamic optimization through the closed-loop diagnosis and treatment process design. At the data input end, through real-time verification and interactive guidance mechanism, effective control of medical data quality can be achieved, avoiding diagnostic deviations caused by non-standard image acquisition; at the feature processing layer, a multi-source sign deep integration strategy is adopted to achieve the organic unification of Chinese medicine features such as tongue and pulse with Western medicine data such as imaging and test indicators, so as to solve the problem of multimodal medical information islands. In the diagnostic reasoning stage, the two-way reasoning architecture based on the dynamic pathological model can achieve accurate mapping of Chinese medicine syndromes and modern medical diagnosis, overcoming the defect of disconnection between syndrome differentiation and disease differentiation in traditional methods; in the treatment plan generation link, through the drug and non-therapy collaborative recommendation mechanism, accurate adaptation of personalized plans can be achieved, significantly improving treatment compliance. Finally, with the help of a feedback-driven parameter optimization system, the continuous self-evolution capability of the diagnosis and treatment model is realized, so that the syndrome differentiation logic threshold and the treatment plan weight are dynamically adjusted with the clinical efficacy, fundamentally breaking through the technical bottleneck of the static model iteration lag.

[0107] Through a full-chain design of quality control-feature fusion-intelligent reasoning-solution optimization-closed-loop evolution, this device can achieve a systematic improvement in diagnostic accuracy, a substantial improvement in the acceptance of treatment plans, and the coordinated promotion of standardization of Chinese and Western medicine diagnosis and treatment, providing scalable technical support for the development of integrative medicine.

[0108] Preferably, the type matching and acquisition module 610 provided in this application is configured with the following units:

[0109] An image type recognition unit is used to obtain a diagnosis request and data to be diagnosed, determine whether the image data meets the diagnosis request requirements through a preset image recognition model, and generate an image matching result;

[0110] A data supplement signal generating unit is used to process image matching results that do not meet the diagnosis request, generate a data supplement signal and send it to the user terminal. The data supplement signal is used to instruct the user terminal to display a dynamic schematic diagram of the target acquisition part and a standardized shooting guide to the user through a display medium;

[0111] The target vital sign data generation unit is used to process the image matching results that meet the diagnostic request based on the convolutional neural network, identify the positions of key anatomical landmarks, extract the tongue color features, pulse waveform frequency features and lesion location features, and fuse multi-source features through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features as the target vital sign data.

[0112] Preferably, the data multi-source fusion module 620 provided in this application is configured with the following units:

[0113] An image feature set generation unit is used to perform feature extraction processing on the medical image data to be diagnosed, analyze the changing trend of organ morphological abnormality indicators, and generate an image feature set containing quantitative features of organ structure;

[0114] A test feature set generation unit is used to perform index quantification processing on the test report data of the diagnostic data, mark the clinical risk level of abnormal biochemical indicators, and generate a test feature set including a test abnormality score matrix. The test feature set is used to indicate the progression stage of physiological function disorder;

[0115] The multi-dimensional feature fusion unit is used to perform data fusion processing on the target vital sign data, imaging feature set and test feature set, balance the clinical value weights of different data sources, and generate a comprehensive feature vector that integrates multi-dimensional vital sign indicators.

[0116] The multi-dimensional feature fusion unit includes a standard feature generation subunit, a contribution weight allocation subunit, and a multi-source fusion execution subunit. The standard feature generation subunit is used to perform feature alignment processing on the target vital sign data, imaging feature set, and test feature set, unify the quantitative benchmarks of different data sources, convert measurement units according to clinical diagnostic standards, and generate a standardized feature group with a unified dimension; the contribution weight allocation subunit is used to perform weight allocation processing on the standardized feature group, evaluate the clinical importance level of different vital sign features, and determine the contribution weight distribution reflecting the priority of TCM syndrome differentiation; the multi-source fusion execution subunit is used to perform feature fusion processing on the standardized feature group based on the contribution weight distribution, integrate the diagnostic value of multi-source vital sign indicators, and generate a comprehensive feature vector.

[0117] Preferably, the pathology reasoning module 630 provided in this application is configured with the following units:

[0118] The syndrome attribute division unit is used to perform syndrome feature mapping processing on the comprehensive feature vector, divide the deficiency and excess, cold and heat pathological attribute categories, and generate syndrome feature distribution including syndrome quantitative score;

[0119] The TCM syndrome classification unit is used to classify the distribution of syndrome characteristics, divide syndrome types according to TCM syndrome diagnosis standards, and generate TCM syndrome types;

[0120] The Chinese and Western medicine correlation analysis unit is used to perform disease correlation analysis on the distribution of syndrome characteristics, locate the corresponding relationship between target organ lesions, and determine the correlation mapping relationship between Chinese medicine syndromes and Western medicine diseases;

[0121] The Western medicine disease risk assessment unit is used to perform clinical risk analysis on the association mapping relationship, quantify the degree of organ function damage, and generate Western medicine disease association results including disease codes and risk levels.

[0122] Preferably, the solution recommendation module 640 provided in this application is configured with the following units:

[0123] The Chinese medicine prescription matching unit is used to match Chinese medicine prescriptions according to TCM syndrome types, screen core medicinal material combinations and their compatibility relationships, and generate initial drug treatment plans;

[0124] The non-drug therapy mapping unit is used to perform non-drug therapy mapping processing on the Western medicine disease association results, match the acupoint stimulation plan and its operation specifications, and generate an exercise therapy plan;

[0125] The drug regimen optimization unit is used to optimize the initial drug treatment plan, adjust the medicinal material components based on the physical contraindications, and generate an integrated treatment plan;

[0126] The treatment priority sorting unit is used to prioritize the integrated treatment plan and exercise therapy plan, set the implementation order of treatment measures, and generate personalized recommendation plans.

[0127] Preferably, the drug regimen optimization unit provided in this application includes a constitution type identification subunit, a drug contraindication screening subunit, and a dietary supplementation scheme generation subunit. Among them, the constitution type identification subunit is used to perform constitution feature mapping processing on the TCM syndrome type, identify constitution classification characteristics such as qi deficiency and dampness and heat, and determine the user's constitution type; the drug contraindication screening subunit performs contraindication screening processing on the initial drug treatment plan based on the user's constitution type, filters out inappropriate medicinal material components, and generates an optimized drug treatment plan; the dietary supplementation scheme generation subunit performs dietary supplementation processing on the optimized drug treatment plan, adds a combination of medicinal and edible ingredients, and generates an integrated treatment plan.

[0128] In one embodiment, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned auxiliary diagnosis and treatment method when executing the computer program.

[0129] In one embodiment, the present application also provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned auxiliary diagnosis and treatment method when executed by a processor.

[0130] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0131] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0132] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should be included within the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. An auxiliary diagnosis and treatment method, characterized in that: The following steps are involved: S1: Obtaining a user's diagnosis request and data to be diagnosed, and performing type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data, and test report data; S2: performing feature extraction and data fusion processing on the target vital sign data based on the medical imaging data and test report data of the data to be diagnosed, extracting tongue color features, pulse waveform frequency features, and lesion location features, and fusing multi-source features through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features; S3: Inputting the comprehensive feature vector into a pre-trained pathological reasoning model for syndrome reasoning processing, performing reasoning analysis on the comprehensive feature vector, and outputting a TCM syndrome type and a Western medicine disease association result, wherein the Western medicine disease association result is used to indicate a modern medical diagnosis corresponding to the TCM syndrome; S4: Inputting the TCM syndrome type and Western medicine disease association results into a pre-trained treatment plan reasoning model for processing, calling the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtaining personalized recommendation plans including drug treatment plans, diet therapy plans, and exercise therapy plans; S5: Optimizing parameters of the pathology reasoning model and the treatment plan reasoning model based on user feedback data on the personalized recommendation plan, and updating parameters of the pathology reasoning model and the treatment plan reasoning model.

2. The method according to claim 1, characterized in that Said S1 comprises: S11: Obtaining a user's diagnosis request and data to be diagnosed, performing image type recognition processing on the image data of the data to be diagnosed using a preset image recognition model, determining whether the image data meets the requirements of the diagnosis request, and generating an image matching result; S12: Processing the image matching results that do not meet the diagnosis request, generating a data supplement signal and sending it to the user terminal, wherein the data supplement signal is used to instruct the user terminal to display a dynamic diagram of the target acquisition part and a standardized shooting guide to the user through a display medium; S13: Based on the convolutional neural network, the image matching results that meet the diagnostic request are processed to identify the positions of key anatomical landmarks, extract tongue color features, pulse waveform frequency features and lesion location features, and fuse multi-source features through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features as target vital sign data.

3. The method according to claim 1, characterized in that The S2 includes: S21: performing feature extraction processing on the medical imaging data to be diagnosed, analyzing the changing trend of organ morphological abnormality indicators, and generating an image feature set containing quantitative features of organ structure; S22: performing index quantification processing on the test report data of the data to be diagnosed, marking the clinical risk level of the abnormal biochemical index, and generating a test feature set including a test abnormality score matrix, wherein the test feature set is used to indicate the progression stage of the physiological function disorder; S23: performing data fusion processing on the target vital sign data, the imaging feature set, and the test feature set, balancing the clinical value weights of different data sources, and generating a comprehensive feature vector that integrates multi-dimensional vital sign indicators.

4. The method according to claim 3, characterized in that The S23 includes: S231: performing feature alignment processing on the target vital sign data, the imaging feature set, and the test feature set, unifying the quantitative benchmarks of different data sources, converting measurement units according to clinical diagnostic standards, and generating a standardized feature set with a unified dimension; S232: performing weight distribution processing on the standardized feature group, evaluating the clinical importance level of different physical signs and features, and determining the contribution weight distribution reflecting the priority of TCM syndrome differentiation; S233: Performing feature fusion processing on the standardized feature group based on the contribution weight distribution, integrating the diagnostic value of multi-source physical sign indicators, and generating a comprehensive feature vector.

5. The method according to claim 1, wherein The S3 includes: S31: performing syndrome feature mapping processing on the comprehensive feature vector, dividing the deficiency-excess cold-heat pathological attribute categories, and generating a syndrome feature distribution including a syndrome quantitative score; S32: Classifying the syndrome characteristic distribution, dividing the syndrome types according to the TCM syndrome diagnosis standard, and generating TCM syndrome types; S33: performing disease association analysis on the syndrome characteristic distribution, locating the corresponding relationship between target organ lesions, and determining the association mapping relationship between TCM syndromes and Western medicine diseases; S34: Performing clinical risk analysis on the association mapping relationship, quantifying the degree of organ function damage, and generating a Western medicine disease association result including a disease code and risk level.

6. The method according to any one of claims 1 to 5, characterized in that The S4 includes: S41: performing Chinese medicine prescription matching processing on the TCM syndrome type, screening core medicinal material combinations and their compatibility relationships, and generating an initial drug treatment plan; S42: performing non-drug therapy mapping processing on the Western medicine disease association results, matching acupoint stimulation schemes and operation specifications, and generating an exercise therapy plan; S43: Optimizing the initial drug treatment plan, adjusting the medicinal material components based on the physical contraindications, and generating an integrated treatment plan; S44: Prioritize the integrated treatment plan and exercise therapy plan, set the implementation order of the treatment measures, and generate personalized recommendation plans.

7. The method according to claim 6, characterized in that The S43 includes: S431: Performing constitution feature mapping processing on the TCM syndrome type, identifying constitution classification features such as qi deficiency, dampness and heat, and determining the user's constitution type; S432: Perform contraindication screening on the initial drug treatment plan based on the user's physical condition, filter out inappropriate medicinal components, and generate an optimized drug treatment plan; S433: Supplement the optimized drug treatment plan with dietary therapy, add a combination of food and medicine ingredients to generate an integrated treatment plan.

8. An auxiliary diagnosis and treatment device, characterized in that: The device comprises: A type matching and acquisition module is used to obtain the user's diagnosis request and data to be diagnosed, and perform type matching and feature acquisition processing on the image data of the data to be diagnosed based on the diagnosis request to generate target vital sign data. The data to be diagnosed includes image data of the user's part to be diagnosed, medical imaging data, and test report data; A data multi-source fusion module is used to perform feature extraction and data fusion processing on the target vital sign data based on the medical imaging data and test report data of the data to be diagnosed, extract tongue color features, pulse waveform frequency features and lesion location features, and fuse multi-source features through a cross-modal attention weighted algorithm to generate a comprehensive feature vector containing multi-dimensional features; a pathology reasoning module, configured to input the comprehensive feature vector into a pre-trained pathology reasoning model for syndrome reasoning processing, perform reasoning analysis on the comprehensive feature vector, and output a TCM syndrome type and a Western medicine disease association result, wherein the Western medicine disease association result is used to indicate a modern medical diagnosis corresponding to the TCM syndrome; A treatment plan recommendation module is used to input the TCM syndrome types and Western medicine disease association results into a pre-trained treatment plan inference model for processing, call the TCM prescription library and non-drug therapy knowledge base to recommend treatment plans, and obtain personalized recommendation plans including drug treatment plans, diet therapy plans, and exercise therapy plans; The model parameter optimization module is used to optimize the parameters of the pathology reasoning model and the treatment plan reasoning model based on the user's feedback data on the personalized recommendation plan, and update the parameters of the pathology reasoning model and the treatment plan reasoning model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images