Traditional Chinese and western medicine fusion diagnosis and treatment method based on intelligent large model

Abstract

The invention discloses a traditional Chinese and western medicine fusion diagnosis and treatment method based on an intelligent large model, and relates to the field of artificial intelligence and medical health, and the method comprises the steps: collecting traditional Chinese medicine four diagnosis information and western medicine diagnosis data of a patient, carrying out the standardized preprocessing through a unified interface, and obtaining structured input data; a traditional Chinese and western medicine diagnosis fusion inference framework is constructed by using a large model, cross-modal alignment and fusion are performed on a traditional Chinese medicine syndrome differentiation typing result and a western medicine diagnosis conclusion to form unified semantic representation, a causal robust modeling mechanism is introduced, and a personalized prescription diagnosis and treatment scheme is generated based on the unified semantic representation and causal robust features in combination with individual features of a patient. The problems that in the prior art, traditional Chinese and western medicine data splitting, individuation is insufficient, and intelligence and interpretability are poor are solved, the limitation that in existing traditional Chinese and western medicine diagnosis and treatment, data splitting, individuation is insufficient and intelligence is insufficient is broken through, and the technical effects that effectiveness, robustness and interpretability of hypertension diagnosis and treatment are improved are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence and medical health, in particular to a method for diagnosis and treatment of traditional Chinese medicine and western medicine based on intelligent large model. BACKGROUND

[0002] Hypertension is one of the chronic diseases with the highest incidence worldwide, and is known as the "first risk factor of cardiovascular and cerebrovascular diseases". Epidemiological studies have shown that the number of adult hypertensive patients in China has exceeded 300 million, and the prevalence rate is still on the rise. If not effectively controlled, hypertension often leads to serious complications such as myocardial infarction, stroke, and renal failure, causing a huge burden on personal health and public health systems. Western medicine diagnosis mainly relies on objective detection methods: (1) Blood pressure measurement: through ambulatory blood pressure monitoring (ABPM) and home blood pressure monitoring (HBPM) to determine whether the patient meets the diagnostic criteria for hypertension. (2) Laboratory tests: including blood biochemistry (lipids, blood sugar, renal function, liver function, electrolytes), urine routine, myocardial enzyme spectrum, etc., for assessing the risk of complications. (3) Imaging and functional examination: electrocardiogram, color Doppler ultrasound of the heart, fundus examination, carotid ultrasound, etc., for judging the damage to the target organs. (4) Western medicine treatment methods: mainly drug therapy, common drugs include diuretics, ACEI / ARB, calcium channel blockers, beta blockers, etc., supplemented by low-salt diet, exercise, smoking cessation and alcohol restriction, and other lifestyle interventions. Although western medicine has clear quantitative indicators and high standardization, it has the following shortcomings: individualized treatment is insufficient, different patients have significant differences in drug efficacy; it often focuses on "blood pressure indicators", but lacks long-term conditioning of the overall body function state; drug dependence is strong, and some patients still have difficulty in achieving stable blood pressure for a long time. Traditional Chinese medicine believes that hypertension belongs to the category of "dizziness" and "headache", and the core of its pathogenesis is "hyperactivity of liver yang, obstruction of phlegm turbidity, and imbalance of yin and yang". The diagnostic methods are: (1) Inspection: observe the complexion, tongue and tongue fur, and distinguish the function of blood and viscera; (2) Auscultation: listen to the patient's voice and breathing; (3) Inquiry: ask about medical history, diet, emotions, family history, etc.; (4) Pulse diagnosis: analyze the function of blood and viscera through pulse diagnosis. Treatment methods include: (1) Traditional Chinese medicine prescription conditioning, such as Tianma Gouteng Decoction and Zhen Gan Xifeng Decoction; (2) Acupuncture, massage and other methods to regulate meridian and blood; (3) Diet, rest, and emotional lifestyle guidance. The advantages of traditional Chinese medicine lie in the overall view and syndrome differentiation, emphasizing individualized treatment, and being able to provide long-term conditioning and prevention for patients. However, it has the following shortcomings: lack of standardized and objective indicators, with large differences in doctor's level; clinical efficacy evaluation relies mainly on experience, lacking a unified quantitative system; and there is a lack of data-level connection with modern western medicine test results.

[0003] However, in actual clinical application, there are still some limitations in the diagnosis and treatment mode of traditional Chinese and Western medicine, and a unified data fusion and collaborative diagnosis and treatment system has not yet been formed. Including: (1) Data fragmentation: Western medicine relies on quantitative indicators, and traditional Chinese medicine relies on experience-based syndrome differentiation, both lack information sharing and unified analysis platform; (2) Lack of intelligence: existing AI diagnosis and treatment mostly focuses on image recognition or single modality, lacking comprehensive analysis of multi-modal fusion of tongue diagnosis, pulse diagnosis, face diagnosis and western laboratory examination; (3) Lack of individualization: western medicine treatment focuses on drug dependence, and traditional Chinese medicine lacks standardization, making it difficult to form an integrated and precise diagnosis and treatment plan; (4) Lack of explainability: existing intelligent diagnosis and treatment models are often "black box", making it difficult to provide reasoning process consistent with clinical logic, reducing the trust of doctors and patients.

[0004] Therefore, there is an urgent need for an intelligent diagnosis and treatment method that integrates western objective detection and traditional Chinese whole-body syndrome differentiation, which can not only rely on the reasoning ability of large models to improve the comprehensiveness and explainability of diagnosis, but also provide personalized intervention plans for hypertension patients combining traditional Chinese and western medicine. SUMMARY

[0005] The application discloses a hypertension diagnosis and treatment technology based on intelligent large models. The technology faces the whole process management of hypertension, comprehensively collects western objective detection data (blood pressure, blood lipid / blood glucose / renal function laboratory indicators, electrocardiogram / image, etc.), traditional Chinese four diagnostic data (face diagnosis, tongue diagnosis, pulse diagnosis, and inquiry diagnosis), and individual characteristics (height, weight, age, family history, and lifestyle), and through a unified framework of multi-modal feature extraction-cross-modal alignment and fusion-knowledge retrieval enhanced chain reasoning (RAG+CoT)-causal robust representation learning-personalized prescription generation, an explainable traditional Chinese and western combined diagnosis and treatment suggestion (including drugs / Chinese medicine prescriptions / acupuncture / diet / exercise, etc.) is output. The application solves the problems of data fragmentation, lack of individualization, and poor intelligence and explainability in existing solutions, and has the advantages of standardization, explainability, and deployability.

[0006] In view of the above problems, the application provides a traditional Chinese and western combined diagnosis and treatment method based on intelligent large models, which comprises: Collecting traditional Chinese four diagnostic information and western diagnostic data of a patient, and performing standardized preprocessing through a unified interface to obtain structured input data, wherein the traditional Chinese four diagnostic information includes face diagnosis images, tongue diagnosis images, pulse diagnosis signals, and inquiry diagnosis texts, and the western diagnostic data includes blood pressure detection results, laboratory examination indicators, electrocardiograms, and imaging examination reports; A large model is used to construct a traditional Chinese and western combined diagnosis and inference framework, and traditional Chinese syndrome differentiation results and western diagnosis conclusions are cross-modally aligned and fused to form a unified semantic representation, wherein the traditional Chinese and western combined diagnosis and inference framework is used to simultaneously maintain the explainability of traditional Chinese whole-body syndrome differentiation logic and the objectivity of western quantitative indicators; By combining a unified semantic representation, a causal robust modeling mechanism is introduced to extract causal robust features. The causal robust modeling mechanism is used to adapt to different groups of people and collection scenarios, and reduce the impact of environmental noise and data bias. Personalized prescription treatment plans are generated based on the unified semantic representation and causal robust features, taking into account the individual characteristics of the patients.

[0007] Preferably, the standardization preprocessing includes image normalization, text vectorization, signal filtering, and numerical scaling transformation; In this process, image data is preprocessed using a convolutional neural network to remove differences in illumination and resolution; time-series pulse diagnosis signals are corrected for pulse frequency deviations through filtering and normalization; text information is semantically segmented and structured through natural language processing; and numerical indicators are standardized in terms of dimensions through minimum-maximum normalization.

[0008] Preferably, the traditional Chinese medicine and Western medicine diagnostic fusion inference framework is implemented by constructing a cross-modal alignment mechanism, wherein the cross-modal alignment mechanism divides the traditional Chinese medicine features and Western medicine features into different subspaces and uses attention mechanism and contrastive learning to achieve alignment and weighted fusion; wherein, subspace alignment ensures that the semantic features of traditional Chinese medicine and the numerical features of Western medicine can be expressed in a unified latent space.

[0009] Preferably, the causal robust modeling mechanism includes a causal representation learning module, which extracts robust causal features through adversarial debiasing, feature perturbation screening, and causal independence testing to maintain stable prediction performance under different populations and different collection conditions. Adversarial debiasing is used to weaken spurious correlations caused by sampling bias and individual differences; feature perturbation screening verifies the stability of features by introducing noise intervention during the training phase; and causal independence testing is used to filter out spurious correlation features that have no direct causal effect on the prediction.

[0010] Preferably, the personalized prescription treatment plan includes: a Traditional Chinese Medicine (TCM) formula module, which matches classic or empirical formulas based on syndrome differentiation, outputs the constituent drugs, recommended dosages, and compatibility principles, and explains the indications and contraindications. The module can dynamically adjust the dosage and herbal compatibility based on the patient's age and constitution; an acupuncture treatment module, which recommends acupoint selection based on syndrome differentiation and constitution, outputs key acupoints, treatment methods, and treatment courses, and supports individualized adjustments; the acupuncture treatment module dynamically modifies the stimulation frequency and needle retention time based on the patient's tolerance; a dietary regulation module, which recommends a dietary structure based on the patient's constitution, outputs food categories, pairing principles, and contraindications, and adjusts dietary recommendations based on the patient's lifestyle and regional differences; and a lifestyle guidance module, which provides suggestions on exercise, rest, and emotional regulation based on individual information to improve patient compliance and recovery effects, including recommending appropriate exercise forms and intensities, adjusting rest schedules, and providing psychological counseling methods.

[0011] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention collects patients' information from the four diagnostic methods of Traditional Chinese Medicine (TCM) and Western medicine diagnostic data. Standardized preprocessing is performed through a unified interface to obtain structured input data. The TCM diagnostic information includes facial images, tongue images, pulse signals, and medical history text. The Western medicine diagnostic data includes blood pressure test results, laboratory test indicators, electrocardiograms, and imaging reports. A large-scale model is used to construct a TCM-Western medicine diagnostic fusion inference framework. The TCM syndrome differentiation results and Western medicine diagnostic conclusions are cross-modal aligned and fused to form a unified semantic representation. This framework maintains both the interpretability of the overall TCM syndrome differentiation logic and the objectivity of Western medicine quantitative indicators. Furthermore, a causal robust modeling mechanism is introduced, combined with the unified semantic representation, to extract causal robust features. This mechanism adapts to different populations and data collection scenarios, reducing the impact of environmental noise and data bias. Finally, personalized prescription treatment plans are generated based on the unified semantic representation and causal robust features, combined with individual patient characteristics. This achieves the technical effect of improving the comprehensiveness and interpretability of diagnosis by leveraging the reasoning capabilities of a large-scale model, while providing personalized intervention plans combining TCM and Western medicine for hypertension patients. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1A flowchart illustrating a method for integrated traditional Chinese and Western medicine diagnosis and treatment based on an intelligent large model, provided in an embodiment of the present invention; Detailed Implementation

[0014] This invention provides a diagnostic and treatment method integrating traditional Chinese and Western medicine based on an intelligent large model, which is used to address the problems of fragmented data between traditional Chinese and Western medicine, insufficient personalization, and poor intelligence and interpretability in existing technologies.

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0016] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0017] Examples, such as Figure 1 As shown, this invention provides a method for integrated traditional Chinese and Western medicine diagnosis and treatment based on an intelligent large-scale model, the method comprising: The patient's TCM four diagnostic information and Western medicine diagnostic data are collected and standardized preprocessed through a unified interface to obtain structured input data. The TCM four diagnostic information includes facial diagnosis images, tongue diagnosis images, pulse diagnosis signals, and consultation text. The Western medicine diagnostic data includes blood pressure test results, laboratory test indicators, electrocardiogram, and imaging examination reports.

[0018] The standardization preprocessing includes image normalization, text vectorization, signal filtering, and numerical scaling transformation. In this process, image data is preprocessed using a convolutional neural network to remove differences in illumination and resolution; time-series pulse diagnosis signals are corrected for pulse frequency deviations through filtering and normalization; text information is semantically segmented and structured through natural language processing; and numerical indicators are standardized in terms of dimensions through minimum-maximum normalization.

[0019] In one possible implementation, the Western medicine testing data comes from: patient-uploaded blood test reports and physical examination reports (PDF or image format). Processing method: Based on existing mature OCR (Optical Character Recognition) tools and pre-trained NLP models in the medical field, the Western medicine diagnostic report text is parsed to directly extract structured indicators such as blood pressure, blood lipids (LDL, HDL, TC, TG), blood glucose, kidney function (creatinine, uric acid), and electrocardiogram parameters. Stable indicator recognition and extraction can be achieved without additional training. Representation format: ,in For indicator dimensions.

[0020] Traditional Chinese medicine image data includes: tongue diagnosis images: patients upload photos of their tongue surface (RGB / infrared), and tongue texture and coating features are extracted; facial diagnosis images: patients upload facial images (RGB / infrared), and skin color and blood circulation features are extracted; pulse diagnosis signals: pulse waveforms are recorded using a high-precision pressure sensor array to obtain time-series data. ,in For sampling duration; Consultation text: Obtain information such as lifestyle habits, medical history, and family history through speech recognition and human-computer interaction question and answer.

[0021] Individual characteristic data includes patient height, weight, age, gender, family medical history, dietary structure, exercise habits, smoking and drinking status, and occupation; representation format: encoded as a numerical and categorical vector, denoted as... .

[0022] The specific process of data standardization is as follows: Numerical data Z-score standardization is adopted. in, Original features and These are the mean and standard deviation, respectively. This is a standardized result.

[0023] The standardization process for image data (tongue diagnosis, facial diagnosis) involves: unifying the resolution and color space, normalizing to [0,1], and performing enhancements such as cropping, rotation, and brightness correction. The standardization process for time-series signals (pulse waveform) involves: bandpass filtering for noise reduction, removing baseline drift, and normalizing the amplitude to [−1,1]. The standardization process for text data (medical consultation) involves: inputting the segmented text into a pre-trained medical language model and embedding it as a fixed-dimensional vector.

[0024] A large model is used to construct a diagnostic fusion inference framework for traditional Chinese medicine and Western medicine. The results of TCM syndrome differentiation and Western medicine diagnosis are aligned and fused across modalities to form a unified semantic representation. The diagnostic fusion inference framework is used to simultaneously maintain the interpretability of the overall syndrome differentiation logic of TCM and the objectivity of Western medicine quantitative indicators. The embodiments of the present invention also include: The proposed integrated inference framework for TCM and Western medicine diagnosis is achieved by constructing a cross-modal alignment mechanism. This mechanism divides TCM features and Western medicine features into different subspaces and uses attention mechanisms and contrastive learning to achieve alignment and weighted fusion. Among them, subspace alignment ensures that the semantic features of traditional Chinese medicine can be expressed in a unified latent space with the numerical features of Western medicine.

[0025] Because the diagnosis of hypertension involves multimodal data such as blood indicators, imaging data, pulse waveforms, and text descriptions, relying on only a single modality can lead to biased and inaccurate diagnostic results. Therefore, it is necessary to design corresponding feature extractors for different types of data to convert the raw data into a unified representation vector for subsequent fusion and inference. This invention uses deep learning models as encoders for different modalities: global visual features of tongue diagnosis images are extracted using a Visual Transformer (ViT); local and texture features of facial diagnosis images are extracted using a Convolutional Neural Network (CNN); temporal dependencies of pulse diagnosis signals are extracted using a Temporal Convolutional Network (TCN); laboratory and individual numerical indicators are modeled using a Multilayer Perceptron (MLP); and the consultation text is semantically embedded using a BERT model in the medical field.

[0026] The resulting embedding vector is used for unified representation: .

[0027] in, It is a visual Transformer model used to extract features from tongue diagnosis images; It is a convolutional neural network used to extract features from facial diagnosis images; It is a temporal convolutional neural network used to extract pulse signal features; It is a multi-layer fully connected network used to extract features from numerical data; This is a medical natural language processing model used to extract semantic features from medical consultation texts. and The original image of the tongue and the extracted features; and The original image and extracted features are used for the face diagnosis. and This includes the raw time-series data of the pulse and the extracted features. and This refers to the raw data and extracted features derived from Western medical diagnostic reports; and This includes the original consultation information and the extracted features; The original data of individual characteristics and the extracted features. For features extracted from various data sources, we constrain them to the same dimension. .

[0028] Traditional Chinese medicine (TCM) and Western medicine data come from different sources and have significantly different semantic spaces. Without cross-modal alignment, models struggle to capture the correlations between different modalities, leading to fragmented diagnostic results. Therefore, it is necessary to construct a unified semantic space to enable TCM and Western medicine data to complement each other, thereby achieving a more complete diagnostic basis. This invention maps different modalities to a unified space using a subspace projection network and employs contrastive learning (InfoNCE) as a constraint to make the TCM and Western medicine features of the same patient spatially similar. Simultaneously, an attention mechanism is used to dynamically weight the importance of different modalities, resulting in a fused representation.

[0029] First, the extracted features are projected into two subspaces: Traditional Chinese Medicine and Western Medicine. ; in, It is a semantic projection network for traditional Chinese medicine, consisting of a multi-layer fully connected network and a multi-head attention mechanism; It is a semantic projection network for Western medicine, with the same structure as the semantic projection network for traditional Chinese medicine, but the parameters are not shared; and These are outputs demonstrating the integration of Traditional Chinese Medicine (TCM) and Western medicine. To improve stability, and L2 normalization is performed during the subsequent calculations: , Next, feature alignment constraints are applied across traditional Chinese and Western medicine modalities, using the InfoNCE approach in contrastive learning to ensure that the same person... and More similar, while differentiating themselves from others distance: , in, , This refers to the number of mini-batches during network training. For anchor sample index, Enumerate positive / negative samples (usually) (negative) The temperature coefficient is used to control the sharpness of the contrast distribution. This invention employs... The similarity calculation function uses cosine similarity. .

[0030] Then, attention-weighted fusion representation is performed: , in, This is a modal set (tongue diagnosis, facial diagnosis, pulse diagnosis, Western medicine laboratory indicators, text, individual characteristics, etc.); For the first Feature representation of each modality The learnable attention scoring vector is randomly initialized early in training and then continuously updated through backpropagation to automatically learn the importance of features from different modalities. Specifically, With each modal feature The inner product is calculated to obtain the saliency score of the modality, which is then normalized using softmax to form the attention weights. Finally, a unified fusion representation is obtained through weighted summation. This allows the system to dynamically and adaptively allocate the contribution of different modalities, thus more reasonably reflecting the diagnostic and therapeutic value of Chinese and Western medicine data.

[0031] By combining a unified semantic representation, a causal robust modeling mechanism is introduced to extract causal robust features. The causal robust modeling mechanism is used to adapt to different groups of people and collection scenarios, and reduce the impact of environmental noise and data bias. The embodiments of the present invention also include: The causal robust modeling mechanism includes a causal representation learning module, which extracts robust causal features through adversarial debiasing, feature perturbation screening, and causal independence testing, in order to maintain stable prediction performance under different populations and different data collection conditions. Among them, adversarial debiasing is used to weaken spurious correlations caused by sampling bias and individual differences; feature perturbation screening verifies the stability of features by introducing noise intervention during the training phase; and causal independence testing is used to screen out spurious correlation features that have no direct causal effect on prediction.

[0032] Because the sources of multimodal data in traditional Chinese medicine and Western medicine are complex, and there is noise interference (such as differences in image lighting and acquisition equipment), individual differences (such as age, weight, and family history), and non-critical variables (such as temporary dietary habits), directly using fusion features is not feasible. Modeling and prediction results can be unstable and biased. Therefore, causal robust modeling is needed to eliminate spurious correlations and highlight causal correlation characteristics, ensuring consistency and reliability of prediction results across different populations and data collection conditions. This invention obtains fusion features... Subsequently, to ensure the stability of diagnostic results and the individualization of prescription recommendations, a causal robust modeling and prescription optimization module was further constructed.

[0033] First, it is mapped to robust causal features through the causal representation module: , in, For the causal robustness characteristics that are ultimately obtained; The causal feature extraction module includes: (1) preprocessing to reduce bias: by comparing the feature distribution of different populations or under different collection conditions, adversarial training or statistical constraints (such as maximum mean difference MMD) are used to reduce data bias; (2) feature perturbation simulation: random perturbation (Gaussian noise) is performed on the input features to screen out causal features that still maintain stable predictive effect under perturbation; (3) causal screening: using mutual information constraints or simplified causal discrimination criteria (such as conditional independence test) to screen out false features that are not related to the label.

[0034] The final diagnosis was: , in, For the diagnosis and treatment results, Classification / regression head (can be a softmax classifier or an MLP regressor).

[0035] Next, in the prediction results Based on this, individual patient characteristics are introduced. [Age, gender, family history, work-rest schedule, occupation], e.g., 53-year-old male, Qi and Yin deficiency constitution, history of surgery, caregiver. A candidate prescription set is then constructed. , Each of them They all include traditional Chinese medicine solutions, lifestyle plans, and dietary plans.

[0036] Finally, based on the prediction results and individual characteristics, the optimal prescription is selected: , in, Considerations: (1) Efficacy indicators: measuring the effectiveness of the prescription in predicting disease status. The improvement effect can be obtained from clinical statistical models or experience databases, such as the magnitude of blood pressure reduction, the proportion of tumor recurrence risk reduction, etc.; (2) Safety indicators: examine whether the prescription has side effects or contraindications, for example, when the patient has liver and kidney dysfunction, avoid certain drugs or Chinese medicines; (3) Compliance indicators: consider whether the patient can follow the prescription for a long time, such as drug cost, number of times to take the medicine, difficulty of dietary adjustment, and frequency of acupuncture.

[0037] therefore, It can be written as: , in, As punishment for insufficient therapeutic effect, As punishment for security risks, As a punishment for insufficient compliance, These are weighting coefficients, set according to clinical needs.

[0038] Through the above steps, this system can eliminate noise and bias by using causal modeling to ensure the robustness of diagnostic results; based on the predicted results, combined with the individual characteristics of patients, it outputs personalized prescriptions that combine traditional Chinese and Western medicine; thus realizing a complete closed loop of "diagnosis - causal robustness - prescription optimization" and ensuring the reliability and individualization of clinical applications.

[0039] Personalized prescription treatment plans are generated based on the unified semantic representation and causal robust features, taking into account the individual characteristics of the patients.

[0040] The personalized prescription treatment plan includes: The Traditional Chinese Medicine (TCM) prescription module is used to match classic or empirical prescriptions based on syndrome differentiation, output the constituent drugs, recommended dosage and compatibility principles, and explain the indications and contraindications. The dosage and drug compatibility can be dynamically adjusted according to the patient's age and physical differences. The acupuncture treatment module is used to recommend acupoint selection schemes based on syndrome types and constitutions, output key acupoints, treatment methods and courses of treatment, and support individualized adjustments; the acupuncture treatment module dynamically modifies the stimulation frequency and needle retention time according to the patient's tolerance. The dietary management module is used to recommend dietary structures based on the patient's constitution, output food categories, pairing principles and taboos, and adjust dietary recommendations in combination with the patient's lifestyle and regional differences. The lifestyle guidance module is used to provide suggestions on exercise, daily routines, and emotional well-being based on individual information, thereby improving patient compliance and rehabilitation outcomes. This includes recommending appropriate forms and intensities of exercise, adjusting daily routines, and providing psychological support.

[0041] In one embodiment, after completing causal robust modeling and personalized prescription optimization, the system needs to output the final results in an intuitive and structured manner to ensure that doctors and patients can clearly understand the diagnostic conclusions and recommended treatment plans. This includes: a Traditional Chinese Medicine (TCM) prescription module: automatically matching corresponding classic or empirical prescriptions based on the syndrome differentiation results; output including the prescription name, constituent drugs, recommended dosage, and compatibility principles; explaining the indications and contraindications of the prescription through the knowledge reasoning capabilities of the large model; an acupuncture treatment module: recommending acupuncture point selection plans based on TCM syndrome types and the patient's constitution; output including key acupoints, treatment methods, and treatment course suggestions; dynamically adjusting the stimulation frequency and needle retention time according to the patient's condition; and a dietary regulation module: recommending corresponding dietary structures based on the TCM principle of "medicine and food sharing the same origin" and the patient's constitution; output including recommended food categories, principles for combining dietary therapy prescriptions, and dietary restrictions; ensuring nutritional balance and assisting in the efficacy of TCM and acupuncture. Lifestyle guidance module: Based on the patient's age, physical condition and lifestyle habits, suggestions are made on exercise, rest and emotional regulation; the output content includes exercise type, exercise intensity, daily routine and emotion management methods; it helps to improve patient compliance and overall rehabilitation effect.

[0042] In summary, the embodiments of the present invention have at least the following technical effects: 1. This invention is the first to integrate the information from the four diagnostic methods of traditional Chinese medicine (face diagnosis, tongue diagnosis, pulse diagnosis, and inquiry) with objective test data from Western medicine (blood pressure, blood lipids, blood sugar, kidney function, electrocardiogram, imaging examinations, etc.) within the same framework, breaking through the problems of data fragmentation and difficulty in complementarity in existing methods, and realizing the unified representation of multimodal information.

[0043] 2. Through cross-modal alignment and fusion mechanisms, this invention utilizes attention mechanisms and contrastive learning constraints to enable the features of traditional Chinese medicine and Western medicine to complement and enhance each other in a unified semantic space. This maintains the integrity and interpretability of the diagnostic logic of traditional Chinese medicine while taking into account the objective quantitative features of Western medicine diagnostic indicators, thereby improving the accuracy and consistency of diagnostic results.

[0044] 3. Based on multimodal fusion features, this invention introduces a causal robust modeling mechanism, which can effectively eliminate noise interference and spurious correlations caused by individual differences, extract stable causal features, improve the generalization ability and reliability of the diagnostic model under different populations and different collection conditions, and avoid the dependence of traditional methods on non-causal features.

[0045] 4. This invention further incorporates the patient's individual characteristics (age, gender, family history, lifestyle, etc.) and automatically generates a personalized prescription plan that combines traditional Chinese and Western medicine. This plan includes traditional Chinese medicine prescriptions, acupuncture plans, dietary conditioning, and lifestyle guidance, and can provide differentiated and precise treatment suggestions for different patients, thereby improving the feasibility and compliance of clinical applications.

[0046] 5. Through the large model reasoning mechanism, the diagnosis and treatment plan generated by this invention has good interpretability, and can clearly explain the basis for recommendation, indications and contraindications. It overcomes the "black box" defect of existing intelligent diagnosis and treatment systems, enhances the trust of doctors and patients in the results of intelligent diagnosis and treatment, and promotes the intelligent and standardized development of integrated traditional Chinese and Western medicine diagnosis and treatment.

[0047] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0048] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0049] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A diagnostic and treatment method integrating traditional Chinese and Western medicine based on an intelligent large-scale model, characterized in that, The method includes: The patient's TCM four diagnostic information and Western medicine diagnostic data are collected and standardized preprocessed through a unified interface to obtain structured input data. The TCM four diagnostic information includes facial diagnosis images, tongue diagnosis images, pulse diagnosis signals, and consultation text. The Western medicine diagnostic data includes blood pressure test results, laboratory test indicators, electrocardiogram, and imaging examination reports. A large model is used to construct a diagnostic fusion inference framework for traditional Chinese medicine and Western medicine. The results of TCM syndrome differentiation and Western medicine diagnosis are aligned and fused across modalities to form a unified semantic representation. The diagnostic fusion inference framework is used to simultaneously maintain the interpretability of the overall syndrome differentiation logic of TCM and the objectivity of Western medicine quantitative indicators. By combining a unified semantic representation, a causal robust modeling mechanism is introduced to extract causal robust features. The causal robust modeling mechanism is used to adapt to different groups of people and collection scenarios, and reduce the impact of environmental noise and data bias. Personalized prescription treatment plans are generated based on the unified semantic representation and causal robust features, taking into account the individual characteristics of the patients.

2. The integrated traditional Chinese and Western medicine diagnosis and treatment method based on an intelligent large model as described in claim 1, characterized in that, The standardization preprocessing includes image normalization, text vectorization, signal filtering, and numerical scaling transformation. In this process, image data is preprocessed using a convolutional neural network to remove differences in illumination and resolution; time-series pulse diagnosis signals are corrected for pulse frequency deviations through filtering and normalization; text information is semantically segmented and structured through natural language processing; and numerical indicators are standardized in terms of dimensions through minimum-maximum normalization.

3. The integrated traditional Chinese and Western medicine diagnosis and treatment method based on an intelligent large model as described in claim 1, characterized in that, include: The proposed integrated inference framework for TCM and Western medicine diagnosis is achieved by constructing a cross-modal alignment mechanism. This mechanism divides TCM features and Western medicine features into different subspaces and uses attention mechanisms and contrastive learning to achieve alignment and weighted fusion. Among them, subspace alignment ensures that the semantic features of traditional Chinese medicine can be expressed in a unified latent space with the numerical features of Western medicine.

4. The integrated traditional Chinese and Western medicine diagnosis and treatment method based on an intelligent large model as described in claim 1, characterized in that, include: The causal robust modeling mechanism includes a causal representation learning module, which extracts robust causal features through adversarial debiasing, feature perturbation screening, and causal independence testing, in order to maintain stable prediction performance under different populations and different data collection conditions. Among them, adversarial debiasing is used to weaken spurious correlations caused by sampling bias and individual differences; feature perturbation screening verifies the stability of features by introducing noise intervention during the training phase; and causal independence testing is used to screen out spurious correlation features that have no direct causal effect on prediction.

5. The integrated traditional Chinese and Western medicine diagnosis and treatment method based on an intelligent large model as described in claim 1, characterized in that, The personalized prescription treatment plan includes: The Traditional Chinese Medicine (TCM) prescription module is used to match classic or empirical prescriptions based on syndrome differentiation, output the constituent drugs, recommended dosage and compatibility principles, and explain the indications and contraindications. The dosage and drug compatibility can be dynamically adjusted according to the patient's age and physical differences. The acupuncture treatment module is used to recommend acupoint selection schemes based on syndrome types and constitutions, output key acupoints, treatment methods and courses of treatment, and support individualized adjustments; the acupuncture treatment module dynamically modifies the stimulation frequency and needle retention time according to the patient's tolerance. The dietary management module is used to recommend dietary structures based on the patient's constitution, output food categories, pairing principles and taboos, and adjust dietary recommendations in combination with the patient's lifestyle and regional differences. The lifestyle guidance module is used to provide suggestions on exercise, daily routines, and emotional well-being based on individual information, thereby improving patient compliance and rehabilitation outcomes. This includes recommending appropriate forms and intensities of exercise, adjusting daily routines, and providing psychological support.

Images