Planning and evaluation system, computer equipment and storage media for TCM diagnosis and treatment

Abstract

This application relates to the field of data processing technology, and discloses a planning and evaluation system, computer equipment, and storage medium for traditional Chinese medicine (TCM) diagnosis and treatment. The system includes: a data acquisition module for acquiring multimodal monitoring data of the target subject during the current treatment phase; an effect prediction module for performing effect simulation and prediction based on the multimodal monitoring data and the target subject's current intervention plan to obtain adjustment results; and a feedback evaluation module for simulating and evaluating the current intervention plan based on the adjustment results to obtain feedback evaluation results. The feedback evaluation results provide quantitative references for treatment adjustments to the decision-making end, enabling the generation of subsequent intervention plans for the target subject in the next treatment phase. These three modules work together to constitute a planning and evaluation system for TCM diagnosis and treatment, providing physicians with real-time decision support and significantly improving the accuracy and adaptability of subsequent intervention plans.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a planning and evaluation system, computer equipment, and storage medium for traditional Chinese medicine diagnosis and treatment. Background Technology

[0002] The TCM diagnosis and treatment process in related technologies is usually based on the four diagnostic methods of the patient at a specific time of consultation, and a fixed diagnosis and treatment plan is formulated for the entire subsequent treatment course.

[0003] However, this data processing method, which relies solely on information from a single visit, makes it difficult to continuously monitor and quantify the dynamic changes in the patient's condition during the treatment process. This results in the inability to provide real-time and reliable decision-making references for adjusting the treatment plan, leading to a lack of timely personalized optimization of the plan.

[0004] Therefore, there is an urgent need for a solution that can continuously and dynamically evaluate and optimize fixed treatment plans in order to improve the accuracy and adaptability of decision-making in determining subsequent intervention plans. Summary of the Invention

[0005] This application aims to at least partially solve one of the technical problems in related technologies. To this end, this application proposes a planning and evaluation system, computer equipment, and storage medium for traditional Chinese medicine diagnosis and treatment. The main technical solutions adopted in this application include: Firstly, this application provides a plan evaluation system for TCM diagnosis and treatment. The system includes: a data acquisition module for acquiring multimodal monitoring data of the target subject during the current treatment phase; wherein the multimodal monitoring data includes dynamic physiological parameters and TCM four diagnostic parameters; dynamic physiological parameters refer to time-series data reflecting the target subject's physiological basis; TCM four diagnostic parameters refer to digital data characterizing the target subject's TCM physical signs; an effect prediction module for performing effect simulation prediction based on the multimodal monitoring data and the target subject's current intervention plan to obtain the plan adjustment results; and a feedback evaluation module for simulating and evaluating the current intervention plan based on the plan adjustment results to obtain feedback evaluation results; wherein the feedback evaluation results are used to provide quantitative treatment adjustment references for decision-making, so as to generate subsequent intervention plans for the target subject in the next treatment phase.

[0006] The data acquisition module addresses the discontinuity issue in traditional diagnosis and treatment by collecting dynamic physiological parameters and parameters from the four diagnostic methods of Traditional Chinese Medicine (TCM), thus improving data completeness and usability. The effect prediction module simulates the effects of the current intervention plan and generates quantitative adjustment suggestions, enhancing the relevance and scientific rigor of these suggestions. The feedback evaluation module, through simulation and quantitative assessment, generates a well-structured and evidence-based decision reference report, transforming the prediction results into actionable auxiliary information. Ultimately, these three modules work together to form a plan evaluation system for TCM diagnosis and treatment, enabling continuous dynamic evaluation and optimization of fixed treatment plans. This provides physicians with real-time decision support and significantly improves the accuracy and adaptability of subsequent intervention plans.

[0007] Optionally, the data acquisition module is also used to: continuously collect dynamic physiological parameters using wearable devices; continuously collect parameters of the four diagnostic methods of traditional Chinese medicine using dedicated acquisition devices; wherein the parameters of the four diagnostic methods of traditional Chinese medicine include observation parameter data, auscultation parameter data, inquiry parameter data, and palpation parameter data; perform structured processing on the historical text data of the target object to obtain historical parameter data; and integrate dynamic physiological parameters, parameters of the four diagnostic methods of traditional Chinese medicine, and historical parameter data to obtain multimodal monitoring data.

[0008] Wearable devices were used to continuously collect dynamic physiological parameters, enabling high-frequency and automated data acquisition. Dedicated data acquisition equipment was used to continuously collect parameters from the four diagnostic methods of Traditional Chinese Medicine (TCM), transforming traditional TCM diagnostic information into objective digital sequences, thus achieving the quantification and traceability of TCM information. Furthermore, by structuring historical text data, fragmented historical information was transformed into standardized historical parameter data, improving the usability of historical data. Finally, by integrating these three types of data, comprehensive multimodal monitoring data was obtained, providing comprehensive and accurate data input for subsequent plan evaluation.

[0009] Optionally, the effect prediction module is also used for: performing feature extraction and fusion processing based on multimodal monitoring data and the current intervention plan to obtain target profile data; wherein, the target profile data is used to reflect the current state of the target object in the current diagnosis and treatment stage; performing matching and reasoning based on a preset knowledge graph and the target profile data to obtain the state assessment result; and using the effect prediction model to optimize and solve based on the current intervention plan to obtain the plan adjustment result; wherein, the effect prediction model is a model constructed based on the current intervention plan data, under preset constraints, and with the state assessment result and the target profile data as input.

[0010] First, through feature extraction and fusion processing, target profile data that comprehensively reflects the target object's state and treatment background was obtained, laying a solid data foundation for subsequent analysis. Then, by simulating TCM diagnostic thinking through matching and reasoning based on a pre-defined knowledge graph and target profile data, a state assessment result capable of quantitatively evaluating the current health status was obtained. Finally, an effect prediction model was used for optimization, automatically exploring and recommending the optimal plan adjustment scheme under safety constraints. This improved the scientific rigor and safety of the adjustment suggestions, providing a reliable basis for adjustment schemes for feedback evaluation.

[0011] Optionally, the effect prediction module also includes a data fusion unit, used for: extracting multi-dimensional features based on multimodal monitoring data to obtain dynamic feature vectors; performing structural analysis and vectorization representation based on the current intervention plan to obtain planned feature vectors; wherein, the planned feature vectors are used to reflect the treatment intention for the target object; and performing attention fusion processing based on the dynamic feature vectors and planned feature vectors to obtain target profile data.

[0012] Multi-dimensional feature extraction was performed on multimodal monitoring data to obtain dynamic feature vectors that comprehensively describe the vital signs of the target object, providing rich underlying information for fusion analysis. Through structural analysis and vectorization, a planned feature vector that accurately encodes the treatment intention was obtained, providing clear guidance for the fusion process. Finally, attention-based fusion processing amplified the influence of features highly correlated with the current treatment intention while effectively suppressing the interference of irrelevant features. This resulted in target profile data that is not only comprehensive but also highlights key points, greatly improving the accuracy and relevance of subsequent status assessment and effect prediction.

[0013] Optionally, the data fusion unit further includes a data preprocessing subunit, used for: performing temporal convolution extraction processing based on dynamic physiological parameters to obtain temporal feature data; performing image and / or signal extraction processing based on the four diagnostic parameters of traditional Chinese medicine to obtain quantitative feature data; and concatenating the temporal feature data and the quantitative feature data to obtain a dynamic feature vector.

[0014] Temporal convolution extraction captures the long-term dependence and cyclical patterns of physiological indicators, improving the timeliness and trend representation ability of dynamic features. Image and / or signal extraction processing of parameters from the four diagnostic methods of Traditional Chinese Medicine (TCM) transforms traditional TCM diagnostic information into objective quantitative data, enhancing the standardization and computability of TCM features. Finally, stitching processing integrates temporal dynamic and static information, achieving organic integration of multi-dimensional monitoring data and providing high-quality data input for subsequent accurate condition profiling.

[0015] Optionally, the effect prediction module also includes a state reasoning unit, used for: calculating state similarity based on a preset knowledge graph and target profile data to obtain state similarity results; wherein, the state similarity results are used to reflect the degree of similarity between at least one candidate state in the preset knowledge graph and the current state; and using a multi-head attention model to perform state weight allocation and analysis based on the target profile data and state similarity results to obtain state evaluation results.

[0016] By utilizing state similarity calculation, candidate syndrome types highly correlated with the current state can be quickly identified, improving the efficiency of dialectical reasoning. Employing a multi-head attention model for multi-angle deep weight allocation and analysis simulates the multi-dimensional dialectical thinking of Traditional Chinese Medicine, strengthening the supporting role of key features, improving the accuracy and interpretability of state assessment results, and providing a reliable basis for subsequent optimization solutions.

[0017] Optionally, the effect prediction module also includes an optimization and solution unit, used for: adjusting the search space based on the parameters defined in the current intervention plan; constructing a prediction objective function based on the state assessment results; using a Bayesian optimization algorithm to iteratively optimize the prediction objective function in the parameter adjustment search space based on preset constraints to obtain the optimized objective function; and, if the optimized objective function satisfies the preset iteration conditions, using the updated intervention plan corresponding to the optimized objective function as the scheme adjustment result.

[0018] By defining parameters to adjust the search space, the specific scope and boundaries of the plan optimization were clarified, improving the targeting of the optimization solution. Constructing a predictive objective function quantified the overall benefits of the adjustment scheme, improving the accuracy of the optimization direction. Iterative optimization using a Bayesian optimization algorithm efficiently explored the optimal solution, reducing blind searches and improving solution efficiency. Optimization under preset constraints ensured the safety and compliance of the adjustment scheme, ultimately achieving the output of the solution with the optimal overall benefit.

[0019] Secondly, this application also provides a computer device, including the planning and evaluation system for TCM diagnosis and treatment as described above.

[0020] Thirdly, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the functions of the planning and evaluation system for TCM diagnosis and treatment described above.

[0021] Fourthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the functions of the planning and evaluation system for TCM diagnosis and treatment as described above. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a structural block diagram of a planning and evaluation system for TCM diagnosis and treatment provided according to an embodiment of this application; Figure 2a This is a flowchart of a method for determining the adjustment result according to yet another embodiment of this application; Figure 2b This is a structural block diagram of a planning and evaluation system for TCM diagnosis and treatment provided according to yet another embodiment of this application; Figure 3 This is an internal structural diagram of a computer device provided according to an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] It should be noted that in the traditional Chinese medicine treatment model, doctors primarily prescribe medication based on the "four diagnostic methods" (inspection, auscultation, palpation, and olfaction) at the time of consultation, and patients typically take medication according to a fixed regimen throughout the treatment course. This is a relatively static model, making it difficult to make timely and personalized adjustments based on the patient's daily physiological changes and recovery progress during treatment.

[0026] With the development of medical informatization and intelligent sensing technologies, digital TCM auxiliary systems (such as tongue diagnosis instruments and pulse diagnosis instruments) and health management platforms have emerged. These systems can collect and record some data, but they generally have two limitations: Firstly, these systems primarily focus on single-session or breakpoint-based auxiliary diagnosis, failing to form a closed loop with the treatment process, especially the continuous dynamic data during medication. In other words, these technologies face the problem of static treatment decisions and a disconnect in treatment course management. Because current TCM treatments typically determine fixed plans based on information from a single consultation, they lack the ability to assess and adjust in real-time over several weeks or months based on the patient's daily dynamic physiological indicators (such as sleep, heart rate, and metabolic data) and TCM signs (such as continuous changes in tongue and pulse). This results in treatment plans that cannot accurately match the patient's rapidly changing physical condition.

[0027] Secondly, while some studies involve medication reminders or prescription adjustments based on follow-up visits, their decision-making logic often relies on simple rules or doctors' manual judgment, lacking a core engine capable of automatically integrating multimodal dynamic data and simulating TCM diagnostic thinking for real-time intelligent analysis. This presents the challenge of intelligently integrating multimodal health data with TCM diagnostic decision-making. Although these systems can collect various types of data, they generally lack an intelligent core capable of deeply understanding TCM diagnostic logic and automatically integrating and analyzing multi-source heterogeneous information such as time-series physiological data, image data, and text medical records. This makes dynamic prescription adjustments still highly dependent on doctors' experience and manual judgment, hindering large-scale, real-time personalized treatment optimization.

[0028] Therefore, in addressing the crucial question of "how to utilize continuously collected home monitoring data to drive personalized dynamic optimization of traditional Chinese medicine prescriptions within the course of treatment," relevant technologies have not yet provided a systematic solution, resulting in the inability to fully leverage the precision and timeliness of personalized TCM treatment.

[0029] Based on this, according to the embodiments of this application, an embodiment of a planning and evaluation system for TCM diagnosis and treatment is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] This embodiment provides a planning and evaluation system 100 for traditional Chinese medicine diagnosis and treatment, such as... Figure 1 As shown, the system includes a data acquisition module 110, an effect prediction module 120, and a feedback evaluation module 130.

[0031] Among them, the data acquisition module 110 can refer to a module with data collection, processing and integration functions, which can be used to acquire multimodal monitoring data of the target object in the current diagnosis and treatment stage.

[0032] It should be noted that the target group can refer to an individual receiving TCM (Traditional Chinese Medicine) treatment services, such as a patient seeking medical attention. The current treatment phase can refer to a continuous treatment cycle designed for the target group. Specifically, after receiving TCM treatment, a continuous period of time for health intervention according to a predetermined plan can be set for the target group; this continuous period is the current treatment phase. For example, after the first consultation, a doctor might prescribe an initial prescription for a patient with spleen deficiency and dampness, sufficient to last for two weeks. These two weeks constitute the current treatment phase, during which the patient will follow the initial prescription and engage in health interventions such as taking medication or adjusting their lifestyle.

[0033] Multimodal monitoring data refers to a collection of data that integrates various dimensions of a target object, comprehensively reflecting its health status. Specifically, multimodal monitoring data can include dynamic physiological parameters and parameters from the four diagnostic methods of Traditional Chinese Medicine. Dynamic physiological parameters refer to time-series data that reflects the target object's physiological basis, i.e., continuous data that changes over time and reflects basic bodily functions, such as heart rate, sleep duration, activity level, or metabolic-related data.

[0034] The parameters of the four diagnostic methods in Traditional Chinese Medicine (TCM) refer to the digital data that can characterize the physical signs of a target object in TCM, namely, the data collected using the four diagnostic methods of TCM: observation, auscultation and olfaction, inquiry and palpation.

[0035] Specifically, the parameters of the four diagnostic methods in Traditional Chinese Medicine include observation, auscultation and olfaction, inquiry, and palpation.

[0036] Among them, the diagnostic parameters can refer to the data obtained by observing the external physical characteristics of the target object and digitally processing them, such as the RGB values ​​of tongue color, the proportion of tongue coating coverage area, and the density of tongue cracks extracted from the tongue image collected by a dedicated tongue imager.

[0037] Auscultation parameter data can refer to data obtained by collecting the sound and / or odor of a target object and processing it digitally, such as acoustic feature data such as speech frequency and energy changes obtained by analyzing mobile phone recordings, and breath component concentration data obtained by electronic nose.

[0038] Consultation parameter data can refer to data obtained by asking the target subject about their symptoms and / or feelings and then processing it in a structured manner, such as quantitative scoring data on fatigue level and stool characteristics collected through structured questionnaires.

[0039] Palpation parameter data can refer to data obtained by sensing the pulse and other vital signs of the target object and processing it digitally. For example, the pulse waveform signals of the three parts of the pulse (cun, guan, and chi) collected by the intelligent pulse diagnosis instrument, as well as the characteristic values ​​such as pulse rate, pulse force, and rhythm extracted from them.

[0040] Specifically, the data acquisition module 110 can also be used to continuously collect dynamic physiological parameters using wearable devices and to continuously collect parameters from the four diagnostic methods of traditional Chinese medicine using dedicated acquisition devices. Subsequently, structured processing can be performed on the historical text data of the target object to obtain historical parameter data. Finally, the dynamic physiological parameters, the four diagnostic methods of traditional Chinese medicine, and the historical parameter data are integrated to obtain multimodal monitoring data.

[0041] Wearable devices can refer to portable smart devices that can be worn on the body of a target and can continuously collect at least one type of physiological data in real time, such as smart bracelets or smartwatches. Dedicated data collection devices can refer to professional devices specifically designed for the four diagnostic methods of traditional Chinese medicine, capable of accurately collecting vital sign data of a target, such as tongue diagnostic instruments, pulse diagnostic instruments, electronic noses, or audio acquisition devices.

[0042] For example, the data acquisition module 110 includes a wearable device and a dedicated data acquisition device, both of which are connected to the system in real time. Taking a smart bracelet as an example, the system can connect to the smart bracelet via Bluetooth to acquire heart rate data and sleep stage data recorded by the smart bracelet at a fixed frequency (such as one hour or one minute). At the same time, various dedicated data acquisition devices can be deployed in the target's home environment to guide the target to perform the four diagnostic methods of traditional Chinese medicine ("inspection, auscultation, inquiry, and palpation") at a fixed frequency (such as taking a picture of the tongue with a tongue imager, collecting pulse signals with a smart pulse pad, and describing symptoms through a structured questionnaire app), and upload the image and waveform data to the system to achieve continuous acquisition of parameters of the four diagnostic methods of traditional Chinese medicine.

[0043] Next, the data acquisition module 110 can also be used to perform structured processing on the historical text data of the target object to obtain historical parameter data.

[0044] Historical text data can refer to text-based data related to health care that was generated before the current stage of diagnosis and treatment, such as electronic medical records, previous prescription records, previous treatment logs, and historical examination reports stored in the electronic medical record system for the target object.

[0045] Specifically, the data acquisition module first identifies and extracts key medical entities from the text using natural language processing technology. Then, it performs normalization and association analysis on the extracted entities to transform them into regular and standardized data fields. Finally, these fields are integrated to obtain historical parameter data. This historical parameter data can refer to a standardized set of parameters that reflects the target object's past health status and treatment history.

[0046] For example, taking the historical electronic medical records of the target object as historical text data, the named entity recognition model can first be used to scan the historical electronic medical records, that is, to extract core entities such as past syndrome types (e.g., "liver stagnation and spleen deficiency"), past prescriptions (e.g., "Poria cocos 15g, Atractylodes macrocephala 10g"), key symptoms (e.g., "insomnia, abdominal distension"), and examination results (e.g., "heart rate 65 bpm"). Then, the syndrome types are converted into corresponding encoding vectors, that is, the Chinese medicine entities are associated with the drug properties knowledge base, thereby organizing the mapped and associated entities and their attributes (e.g., the history of symptom severity and the frequency of drug use) into a structured standard data table to form historical parameter data.

[0047] Furthermore, the data acquisition module 110 can also be used to integrate dynamic physiological parameters, traditional Chinese medicine diagnostic parameters, and historical parameter data to obtain multimodal monitoring data.

[0048] For example, the formats of dynamic physiological parameters, TCM diagnostic parameters, and historical parameters can first be standardized to unify the units of numerical data and the encoding format of text data. Then, outliers, such as abnormal heart rate data and missing values ​​incidentally collected by smart bracelets, can be removed using sliding windows or recognition algorithms. Next, a unified timeline index is established, and all data timelines are aligned sequentially. For example, various data features at the same time point (such as the average heart rate of the day, the thickness of the tongue coating on the day, and the "spleen deficiency" label in the medical history) are concatenated into a multi-dimensional feature vector. Finally, these feature vectors from all time points are organized sequentially to form a dataset containing multi-dimensional information, i.e., multimodal monitoring data, thus comprehensively reflecting the health status and historical background of the target subject at the current stage of diagnosis and treatment.

[0049] Wearable devices were used to continuously collect dynamic physiological parameters, enabling high-frequency and automated data acquisition. Dedicated data acquisition equipment was used to continuously collect parameters from the four diagnostic methods of Traditional Chinese Medicine (TCM), transforming traditional TCM diagnostic information into objective digital sequences, thus achieving the quantification and traceability of TCM information. Furthermore, by structuring historical text data, fragmented historical information was transformed into standardized historical parameter data, improving the usability of historical data. Finally, by integrating these three types of data, comprehensive multimodal monitoring data was obtained, providing comprehensive and accurate data input for subsequent plan evaluation.

[0050] Furthermore, the plan evaluation system 100 for TCM diagnosis and treatment also includes an effect prediction module 120. This effect prediction module 120 can be a module with feature extraction, data fusion, reasoning analysis, and optimization solution functions, which can be used to simulate and predict the effect based on multimodal monitoring data and the current intervention plan of the target object, so as to obtain the result of plan adjustment.

[0051] It should be noted that the current intervention plan can refer to a health intervention program developed at the beginning of the current treatment phase based on the target individual's initial health status. This program can guide the target individual in their conditioning or health management during the current treatment phase. For example, the current intervention plan could be a traditional Chinese medicine prescription issued by a doctor to the target individual, including drug combinations, dosages, and methods of decoction and administration; or it could be a health management plan that includes dietary recommendations and lifestyle adjustments.

[0052] Specifically, firstly, deep feature extraction can be performed on multimodal monitoring data to obtain dynamic feature vectors that characterize the target subject's basic physiological state and TCM physical characteristics. Simultaneously, the current intervention plan is structurally analyzed and vectorized to obtain a plan feature vector reflecting the treatment intention. Then, the two types of vectors can be fused to form a comprehensive target profile data that reflects the target subject's current health status and treatment background.

[0053] Subsequently, based on a pre-defined TCM knowledge graph, the target profile data can be matched and subjected to dialectical reasoning to obtain a status assessment result that reflects the syndrome status of the target individual. Finally, the status assessment result, the target profile data, and the current intervention plan are used as inputs and fed into a pre-trained effect prediction model.

[0054] The model has established a mapping relationship between input features and expected efficacy by learning a large amount of "state-intervention plan-efficacy" sample data in advance. It can simulate the changes in efficacy and risk under different parameter adjustment directions (such as increasing or decreasing drug dosage or adjusting lifestyle).

[0055] Finally, by continuously adjusting the model parameters, a set of parameter adjustment suggestions that optimize the expected therapeutic effect while keeping the risks under control is output as the final treatment plan adjustment result.

[0056] The adjustment result of this plan can be a set of specific parameter modification suggestions calculated by the algorithm. Its essence is reference information after data processing, and it does not constitute a clinical diagnosis or treatment prescription. The adjustment parameters it contains may involve diagnosis and treatment related content such as the compatibility or dosage of traditional Chinese medicine, or it may cover life intervention related content such as work and rest arrangements and dietary suggestions.

[0057] Furthermore, the plan evaluation system 100 for TCM diagnosis and treatment also includes a feedback evaluation module 130. This feedback evaluation module 130 can be a module with simulation, quantitative evaluation and result integration functions, which can be used to simulate and evaluate the current intervention plan based on the results of the plan adjustment to obtain feedback evaluation results.

[0058] Specifically, firstly, based on the adjusted intervention plan, the system can construct a simulated intervention plan containing the adjusted parameters in the computing environment. Then, using an effect prediction model, it simulates and extrapolates the potential changes in efficacy and risks in the next treatment stage after implementing the plan. Next, based on the changes in efficacy and risks, a multi-dimensional effect indicator evaluation is performed, calculating the rate of change in syndrome scores, symptom scores, or physiological indicators. Finally, the system integrates these extrapolated indicator evaluation results, performs weighted calculations and risk-benefit analysis according to preset evaluation rules, and generates a comprehensive quantitative evaluation report as feedback. For example, taking the current intervention plan as "Poria cocos 15g, Atractylodes macrocephala 10g, one dose daily" with a treatment period of 2 weeks, if the plan adjustment result is "Poria cocos increased to 18g", then the feedback evaluation module 130 first updates the current intervention plan based on the plan adjustment result, generating a simulated intervention plan. Then, this simulated plan and the current state of the target subject are input again into the effect prediction model to predict the possible effect at the end of the next treatment phase (two weeks later) after implementing the new plan. For example, the target subject's "spleen deficiency and dampness" syndrome shows an improvement trend. Next, the rate of change of various effect indicators is calculated to obtain the indicator evaluation results. The indicator evaluation results can be specific values. For example, calculating the syndrome score reduction rate, we get an indicator evaluation result of a decrease from 45 points to 32 points, a reduction rate of 28.9%; calculating the physiological indicator improvement rate, we get an indicator evaluation result such as a 20% increase in the tongue coating thick and greasy index; calculating the probability of key symptom relief, we get an indicator evaluation result of a 15% increase in the probability of relief from sticky stool; calculating the probability of risk occurrence, we get an indicator evaluation result of a risk of excessive drug effect of less than 5%. Subsequently, the evaluation results of the above indicators can be weighted and integrated according to the preset weights of various effect indicators (which can be determined by summarizing historical data or by asking senior physicians to score), forming a structured feedback evaluation result.

[0059] The feedback evaluation results can refer to a set of data that can quantitatively reflect the effects and risks after the adjustment. It can be in the form of a structured data report and can be used to provide quantitative references for treatment adjustments to the decision-making end, so as to generate subsequent intervention plans for the target subjects in the next stage of treatment.

[0060] Understandably, the decision-making end can refer to the entity involved in diagnosis and treatment decisions, or it can be the user interface of a professionally qualified physician. Treatment adjustment reference refers to the quantitative indicators, adjustment suggestions, and risk warnings included in the feedback assessment results. The next treatment stage refers to the next consecutive treatment cycle following the current stage. The subsequent intervention plan refers to the health management or intervention program developed for the target individual in the next treatment stage, based on the feedback assessment results. In other words, the feedback assessment results are directly presented on the doctor's user interface, allowing the doctor to combine these results with their own clinical experience to comprehensively evaluate the target individual's current state and treatment effectiveness, make a final decision, and then develop a subsequent intervention plan for the next treatment stage.

[0061] Optionally, this assessment system may also include an information presentation module, which can perform format conversion based on the feedback assessment results, thereby allowing the feedback assessment results to be displayed on the system interface in the form of visual charts or structured reports. At the same time, it generates prompt logs, including specific adjustment suggestions for the plan, the assessment results of various effect indicators, and the reasoning chain and reasoning basis of the entire process from the target profile data used as the basis for reasoning to the final feedback assessment results, so that doctors can view them at any time on the user interface at their decision-making end.

[0062] It should be noted that all adjustment suggestions and feedback evaluation results generated by this system are supplementary reference information and cannot be directly used as the basis for medical decisions. The final diagnosis and prescription adjustment must be made by a licensed physician after a comprehensive assessment of the patient's condition. In other words, this implementation seeks protection for a purely computer-based data processing system. It processes objective data (image pixels and digital signals), and the output is another set of data (adjustment parameters and evaluation reports). This result itself does not constitute a clinically meaningful diagnosis or treatment prescription, but is merely reference information calculated using complex algorithms. Whether and how to adopt this information depends entirely on the physician's independent professional judgment.

[0063] In the above implementation, the data acquisition module 110 solves the problem of data discontinuity in traditional diagnosis and treatment by collecting dynamic physiological parameters and parameters from the four diagnostic methods of Traditional Chinese Medicine, thereby improving the completeness and usability of the data. The effect prediction module 120 improves the pertinence and scientific nature of the intervention plan adjustment suggestions by simulating the effect of the current intervention plan and generating quantitative adjustment suggestions. The feedback evaluation module 130 generates a well-structured and evidence-based decision reference report through simulation and quantitative evaluation, thereby transforming the prediction results into actionable auxiliary information. Finally, these three modules work together to form a plan evaluation system for Traditional Chinese Medicine diagnosis and treatment, thereby realizing continuous dynamic evaluation and optimization of fixed treatment plans, providing physicians with real-time, quantitative decision support, and significantly improving the accuracy and adaptability of subsequent intervention plans.

[0064] In some implementation methods, please refer to the appendix. Figure 2a The effect prediction module 120 is also used for: S210. Based on multimodal monitoring data and the current intervention plan, feature extraction and fusion processing are performed to obtain target profile data.

[0065] The target profile data refers to a structured data set that comprehensively represents the current health status and treatment background of the target subject. Specifically, it can be used to reflect the current state of the target subject at the current treatment stage, which includes the target subject's real-time physical condition and the suitability of the target subject with the current intervention plan. For example, if the target subject is a patient with spleen deficiency and dampness, and the current treatment stage is two weeks and the target subject is on the 7th day of the stage, then the target profile data for that day may include: basic physiological function data (such as an average heart rate of 68 beats / min and a sleep depth of 1.3 hours), TCM physical characteristics data (such as a pale red tongue with a quantitative value of 0.5; a thick and greasy tongue coating index of 0.7), and specific implementation data of the current intervention plan (such as needing to take 15 grams of Poria cocos and 10 grams of Atractylodes macrocephala on that day).

[0066] Specifically, please refer to the appendix. Figure 2b The effect prediction module 120 also includes a data fusion unit 210, which is used to: firstly extract multi-dimensional features based on multimodal monitoring data to obtain dynamic feature vectors; then perform structural analysis and vectorization representation based on the current intervention plan to obtain plan feature vectors; and finally perform attention fusion processing based on dynamic feature vectors and plan feature vectors to obtain target profile data.

[0067] The data fusion unit 210 can be a unit with multi-source data feature extraction, vector transformation, and fusion functions, capable of integrating monitoring data and intervention plan information into unified representation data. The dynamic feature vector can be a high-dimensional data vector that quantifies the dynamic changes in the target object's physiological basis and TCM physical signs. For example, the dynamic feature vector can be a high-dimensional array of [heart rate time-series feature value 0.68, sleep trend feature 0.52, tongue color quantification value 0.5, tongue coating thickness index 0.7, and pulse rate feature 0.75], capable of comprehensively covering dynamic health information.

[0068] Specifically, the data fusion unit 210 may further include a data preprocessing subunit 211, which may be a subunit with feature processing, extraction and splicing functions. It can be used to perform temporal convolution extraction processing based on dynamic physiological parameters to obtain temporal feature data; then perform image and / or signal extraction processing based on the four diagnostic parameters of traditional Chinese medicine to obtain quantitative feature data; and finally splice the temporal feature data and quantitative feature data to obtain a dynamic feature vector.

[0069] Among them, time-series feature data can refer to time-representative feature data extracted from dynamic physiological parameters, which can reflect the quantitative characteristics of the evolution trend of physiological indicators over time. For example, taking heart rate data in dynamic physiological parameters as an example, the data preprocessing subunit 211 can input the heart rate time-series data of the past 7 days into the time-series convolutional network. By expanding the causal convolutional structure, it automatically learns and captures the diurnal rhythm changes of heart rate (such as lower heart rate at night or daytime fluctuations) and long-term trends (such as the overall heart rate gradually stabilizing), and finally outputs a set of time-series feature data that characterizes the complex dynamics of heart rate, such as [heart rate cycle stability 0.72; heart rate trend slope 0.15], so as to more accurately reflect the dynamic changes of the physiological baseline state.

[0070] Quantitative feature data refers to characteristic data that can accurately describe TCM (Traditional Chinese Medicine) signs using digital parameters, quantifying traditional subjective TCM signs into objective values. For example, for TCM diagnostic parameters presented primarily in image form, such as tongue image data in the inspection parameter data, image extraction processing can be performed. This can involve using convolutional neural networks to segment and analyze the tongue image, extracting the tongue's color histogram, texture features of the tongue coating (such as the contrast of the gray-level co-occurrence matrix), and tongue morphology parameters (such as the teeth mark index), to obtain image-level quantitative feature data. For TCM diagnostic parameters presented in signal form, such as pulse waveform data in the palpation parameter data, image extraction processing can be performed. This can involve converting the signal into a spectral image using Fourier transform, extracting signal features such as frequency and energy, to obtain signal-level quantitative feature data. For TCM diagnostic parameters containing comprehensive data existing in both image and signal forms, such as cough recording data and breath concentration distribution images in the auscultation parameter data, image and signal extraction processing can be performed simultaneously, ultimately fusing them to obtain quantitative feature data that integrates both image and signal features.

[0071] After obtaining the time-series feature data and the quantized feature data, they can be concatenated to obtain a dynamic feature vector. For example, the time-series feature data and the quantized feature data can be directly concatenated in the dimensional direction to form a longer, high-dimensional vector, which is the dynamic feature vector containing comprehensive dynamic information about the target object.

[0072] Temporal convolution extraction captures the long-term dependence and cyclical patterns of physiological indicators, improving the timeliness and trend representation ability of dynamic features. Image and / or signal extraction processing of parameters from the four diagnostic methods of Traditional Chinese Medicine (TCM) transforms traditional TCM diagnostic information into objective quantitative data, enhancing the standardization and computability of TCM features. Finally, stitching processing integrates temporal dynamic and static information, achieving organic integration of multi-dimensional monitoring data and providing high-quality data input for subsequent accurate condition profiling.

[0073] Furthermore, after obtaining the dynamic feature vector, the data fusion unit 210 can also perform structural analysis and vectorization representation based on the current intervention plan to obtain the plan feature vector.

[0074] The plan feature vector can be a standardized data vector containing the most representative information of the current intervention plan, reflecting the therapeutic intent towards the target population. Specifically, it can take the form of a numerical vector, clearly demonstrating the core purpose, drug combination logic, or lifestyle intervention direction of the current intervention plan. For example, if the current intervention plan is "Poria cocos 15g (principal drug, for strengthening the spleen) and Atractylodes macrocephala 10g (assistant drug, for drying dampness); one dose daily, taken after meals," then the plan feature vector can be a high-dimensional vector in the form of [Poria cocos properties + 15g dosage; Atractylodes macrocephala properties + 10g dosage; intention to strengthen the spleen and dry dampness; taken after meals], intuitively reflecting the current intervention plan's therapeutic intent to strengthen the spleen and eliminate dampness.

[0075] Specifically, the current intervention plan is first analyzed in a structured manner to identify its core elements, such as the names of Chinese herbal medicines, dosages, compatibility relationships, methods of administration, or lifestyle intervention requirements. These extracted elements are then mapped to a pre-defined knowledge base, assigning each element a corresponding code and weight, and transforming them into standardized vectors. Finally, the vectors of each element are weighted and fused to obtain the plan's feature vector.

[0076] For example, taking the current intervention plan as "Poria cocos 15g, Atractylodes macrocephala 10g, Alisma plantago-aquatica 6g; one dose daily, taken in the morning and evening," the process first involves structural analysis to extract core elements such as the herbs (Poria cocos, Atractylodes macrocephala, and Alisma plantago-aquatica), dosage (15g, 10g, and 6g), compatibility (principal herb: Poria cocos, assistant herb: Atractylodes macrocephala, adjuvant herb: Alisma plantago-aquatica), and administration (one dose daily, taken in the morning and evening). Next, using a pre-defined knowledge base of traditional Chinese medicine attributes, each herb is associated and transformed into a vector containing its properties, meridian tropism, and efficacy (e.g., Poria cocos vector: [sweet and bland, spleen meridian, diuretic and dampness-removing]). The dosage is converted into standardized numerical codes (e.g., 15g is encoded as 0.5), and the compatibility and administration are encoded into vectors according to pre-defined rules. Finally, based on the roles of the drugs (e.g., 0.5 for the principal drug, 0.3 for the assistant drug, and 0.2 for the adjuvant drug), the drug vector and dosage vector are weighted and summed. Then, they are concatenated with the compatibility vector and the administration method vector to form the final plan feature vector, which fully and in a standardized way reflects the treatment intention and execution requirements of the current intervention plan.

[0077] Finally, the data fusion unit 210 can also perform attention fusion processing based on the dynamic feature vector and the planned feature vector to obtain target profile data.

[0078] Specifically, firstly, the dimensions of the dynamic feature vector and the planned feature vector can be unified. Then, a multi-head self-attention mechanism is used to calculate the correlation of all features in both vectors and assign them different weights. Finally, the two types of vectors are fused based on the weights to obtain the target profile data.

[0079] For example, consider a dynamic feature vector with [heart rate time-series feature value 0.68, sleep trend feature 0.52, tongue color quantification value 0.5, tongue coating thickness index 0.7, and pulse rate feature 0.75], and a planned feature vector with [Poria cocos properties + 15g dosage; Atractylodes macrocephala properties + 10g dosage; spleen-strengthening and dampness-drying intent; to be taken after meals]. First, the two types of vectors are processed to unify their dimensions, ensuring consistent data format and collaborative computation. Then, the unified vectors are input into an attention fusion layer, which calculates the correlation between the diagnostic intent in the planned feature vector and each feature in the dynamic feature vector. For instance, the model, based on a diagnostic knowledge graph, finds that the diagnostic intent of "spleen-strengthening and dampness-drying" is often accompanied by the physiological indicator "thick tongue coating," but its correlation with the physiological indicator "sleep trend" is not high. The model extracts these correlation characteristics during calculation, assigning higher weights to the physiological indicator of "thick, greasy tongue coating" and lower weights to the physiological indicator of "sleep trend." This allows the model to focus on features that better reflect the diagnostic intent when using the target profile data for status assessment and effect prediction. After performing correlation analysis and assigning weights to all features in the dynamic feature vector, the weights of each feature can be labeled on that feature to integrate and obtain the target profile data.

[0080] Multi-dimensional feature extraction was performed on multimodal monitoring data to obtain dynamic feature vectors that comprehensively describe the vital signs of the target object, providing rich underlying information for fusion analysis. Through structural analysis and vectorization, a planned feature vector that accurately encodes the treatment intention was obtained, providing clear guidance for the fusion process. Finally, attention-based fusion processing amplified the influence of features highly correlated with the current treatment intention while effectively suppressing the interference of irrelevant features. This resulted in target profile data that is not only comprehensive but also highlights key points, greatly improving the accuracy and relevance of subsequent status assessment and effect prediction.

[0081] S220. Based on the preset knowledge graph and target profile data, perform matching and reasoning to obtain the state assessment result.

[0082] The pre-defined knowledge graph can refer to a digital knowledge base embedded with TCM diagnostic logic. This means that scattered knowledge of syndromes, symptoms, medications, and treatments in TCM theory is logically organized into a knowledge system that allows computers to query and match information. The status assessment result can refer to a digital judgment of the target object's current TCM syndrome type or health status. Specifically, this status assessment result can include the target object's syndrome probability distribution, syndrome evolution trend, and key supporting evidence. For example, the status assessment result could be: "Syndrome probability: Spleen deficiency with dampness 65%, Liver stagnation with spleen deficiency 20%, Damp-heat retention 15%; Evolution trend: Increasing trend of spleen deficiency with dampness; Supporting evidence: Thick, greasy tongue coating index 0.7 and slippery, rapid pulse characteristic 0.75."

[0083] Specifically, the effect prediction module 120 may also include a state reasoning unit 220, which may be a unit with knowledge graph query, similarity calculation and multi-head attention reasoning functions. It can be used to calculate state similarity based on preset knowledge graph and target profile data to obtain state similarity results; and use a multi-head attention model to allocate and analyze state weights based on target profile data and state similarity results to obtain state evaluation results.

[0084] Among them, the state similarity result refers to a set of quantitative values ​​that can be used to reflect the degree of similarity between at least one candidate state in the preset knowledge graph and the current state.

[0085] Understandably, the pre-defined knowledge graph defines many candidate states, such as typical TCM syndrome states like spleen deficiency with dampness, liver stagnation with spleen deficiency, and heart-kidney disharmony. The target profile data contains the current state, reflecting the target object's actual health status. State similarity calculation can be understood as using algorithms (such as cosine similarity or pattern matching) to calculate the degree of fit between the target object's actual health status and each typical syndrome state, thus using different quantitative scores to represent the level of similarity. The set of these quantitative scores constitutes the state similarity result, indicating which standard syndromes the current state is most closely related to.

[0086] After obtaining the state similarity results, a multi-head attention model can be used to perform state weight allocation and analysis based on the target profile data and the state similarity results to obtain the state evaluation results.

[0087] It should be noted that a multi-head attention model can refer to a neural network model that includes multiple parallel attention heads. Specifically, each attention head in this model focuses on capturing the correlation between different types of features and candidate states in the target profile data, thereby simulating the diagnostic thinking of "four diagnostic methods combined" in traditional Chinese medicine.

[0088] For example, suppose the model contains four attention heads, each focusing on analyzing the correlation between tongue appearance and syndrome type, pulse appearance and syndrome type, symptom description and syndrome type, and physiological indicators and syndrome type, respectively. Each head calculates independently and outputs the weights of its corresponding features and candidate syndrome types. For instance, if the attention head focusing on tongue appearance and syndrome type correlation finds a high match between "thick, greasy tongue coating 0.7" and the typical feature "spleen deficiency with dampness," then a higher weight can be assigned to this syndrome type. Similarly, if the attention head focusing on symptoms and syndrome type correlation finds a high match between "loose stools 0.6 and fatigue 0.7" and the typical feature "spleen and stomach damp-heat," then a higher weight can be assigned to this syndrome type. Subsequently, the model concatenates the weight results from each attention head and integrates them through a feedforward network to form a normalized probability distribution. It then labels the key supporting features and their contributions used for inference, forming a complete state assessment result.

[0089] By utilizing state similarity calculation, candidate syndrome types highly correlated with the current state can be quickly identified, improving the efficiency of dialectical reasoning. Employing a multi-head attention model for multi-angle deep weight allocation and analysis simulates the multi-dimensional dialectical thinking of Traditional Chinese Medicine, strengthening the supporting role of key features, improving the accuracy and interpretability of state assessment results, and providing a reliable basis for subsequent optimization solutions.

[0090] S230. Utilize the effect prediction model to optimize and solve the current intervention plan to obtain the results of the plan adjustment.

[0091] Among them, the effect prediction model can refer to an intelligent prediction model built using machine learning methods. It is pre-trained using a large number of historical samples containing health status, intervention plans and efficacy feedback, and can predict the adjusted efficacy and risks based on the current status of the target object and the current intervention plan.

[0092] It should be noted that, since actual diagnosis and treatment must adhere to basic principles such as drug safety, compatibility rules, and drug balance, and to avoid medication risks or deviations from the treatment direction due to unconstrained adjustments, a preset constraint condition needs to be set during the simulation of TCM intervention plan adjustments to ensure the safety and rationality of the adjustment plan. This preset constraint condition can refer to a series of boundary restrictions and rules imposed on the intervention plan adjustments, formulated based on TCM theory, pharmacopoeia standards, and clinical experience. It may include safe drug dosage ranges, compatibility rules (such as prohibiting combinations of "eighteen incompatibilities" and "nineteen incompatibilities"), drug balance requirements (such as limiting the ratio of heat-clearing and dampness-resolving drugs in prescriptions for damp-heat syndromes), and limitations on the magnitude of single adjustments (such as dosage increases or decreases not exceeding ±20% of the original dosage), etc.

[0093] Furthermore, this effect prediction model can be understood as a model constructed using current intervention plan data as a benchmark, under preset constraints, and with state assessment results and target profile data as inputs. That is, the core task of the effect prediction model is to adjust the current intervention plan within its allowable adjustment range (i.e., the space defined by preset constraints) to match a better health status outcome, based on this benchmark. The model's input data can include two parts: first, the state assessment results, used to clarify the target subject's current core health problems; and second, target profile data, used to provide comprehensive individual background information. The model learns the mapping relationship between "health status, intervention plan, and efficacy feedback" in historical samples to predict, under the current state, what kind of efficacy feedback is most likely to be presented by adjusting certain parameters in the intervention plan, and what state the target subject's state is most likely to update to under the predicted efficacy, thus outputting the optimal adjustment recommendations.

[0094] Specifically, the effect prediction module 120 may further include an optimization solution unit 230. This optimization solution unit 230 may refer to a computational unit responsible for searching for the optimal solution in a complex parameter space. It can be used to adjust the search space based on the parameters defined by the current intervention plan, construct a prediction objective function based on the state assessment results, and use a Bayesian optimization algorithm to iteratively optimize the prediction objective function based on preset constraints in the parameter adjustment search space to obtain the optimized objective function. If the optimized objective function satisfies the preset iteration conditions, the updated intervention plan corresponding to the optimized objective function is used as the scheme adjustment result.

[0095] The parameter adjustment search space can refer to the mathematical space consisting of all possible values ​​of each parameter, within the limits of preset constraints and based on the current intervention plan. The objective function can be a mathematical function that can be quantitatively calculated to measure the merits of the adjustment plan.

[0096] For example, suppose the current intervention plan is "Poria cocos 15g, Atractylodes macrocephala 10g, Alisma plantago-aquatica 6g", with the preset constraint that "the dosage of each herb can be adjusted within ±20%, excluding all combinations containing incompatibilities". Then the parameter adjustment search space can be defined as "the dosage of Poria cocos can be adjusted to 12-18g, the dosage of Atractylodes macrocephala can be adjusted to 8-12g, and the dosage of Alisma plantago-aquatica can be adjusted to 4.8-7.2g", so that the model can perform any dosage combination within this range.

[0097] The objective function for prediction can be a weighted function that integrates multiple efficacy indicators (such as efficacy indicators and risk indicators), that is, a mapping function with specific dosage combinations as independent variables and comprehensive benefit value as the dependent variable. Efficacy indicators may include the reduction in the probability of the current syndrome type and the relief rate of key symptoms, while risk indicators may include the drug overdose index and the probability of adverse reactions.

[0098] For example, using "Poria cocos 18g, Atractylodes macrocephala 9g, Alisma plantago-aquatica 6g" as the input dosage combination, combined with the current state assessment results (e.g., the probability of spleen deficiency and dampness excess is 0.68), the model first calculates that the probability of spleen deficiency and dampness excess has decreased to 0.52, a reduction of 0.16, showing an improving trend; the key symptom relief rate has increased by 12%; the drug property bias index is 0.03; and the adverse reaction probability is 4%. Finally, through weighted calculation (e.g., the weight of efficacy indicators is positive 0.7, and the weight of risk indicators is negative 0.3), the final output is a comprehensive benefit value. This value is the objective function value corresponding to the dosage combination, which can quantitatively reflect the comprehensive advantages and disadvantages of the dosage combination. The task of the optimization solution unit 230 is to find the dosage combination that maximizes this output value (representing the best efficacy) and satisfies the constraints.

[0099] Furthermore, after completing the function construction of the model, the optimization solution unit 230 can also use the Bayesian optimization algorithm to iteratively optimize the prediction objective function based on preset constraints in the parameter adjustment search space to obtain the optimized objective function. If the optimized objective function satisfies the preset iteration conditions, the updated intervention plan corresponding to the optimized objective function is used as the scheme adjustment result.

[0100] It should be noted that the Bayesian optimization algorithm first constructs a probabilistic surrogate model for the unknown predictive objective function based on a small number of initial sampling points. Then, according to a sampling function (such as the desired improvement in EI), it selects the next most promising sampling point (i.e., a set of dosage combinations) in the parameter search space for experimentation, that is, it calculates the objective function value at that point using the effect prediction model. Next, it updates the model with this new sample to make it closer to the true predictive objective function. This process is iterated until the objective function value tends to stabilize, resulting in the optimized objective function. This optimized objective function can refer to the objective function that achieves the optimal overall benefit after multiple iterations of optimization.

[0101] It is understandable that the optimized objective function corresponds to an updated intervention plan. This updated intervention plan can refer to a specific intervention adjustment scheme that enables the optimized objective function to achieve the optimal value. That is, it is a set of dosage combinations that meets the constraints and has the highest overall benefit after model optimization and screening.

[0102] For example, taking the current intervention plan of 15g Poria cocos and 10g Atractylodes macrocephala as an example, with a 65% probability of spleen deficiency and dampness syndrome in the status assessment results, after completing the function construction of the model, the model starts the Bayesian optimization algorithm. The first step randomly selects the following three dosage combinations: dosage combination one: 12g Poria cocos and 8g Atractylodes macrocephala; dosage combination two: 16g Poria cocos and 11g Atractylodes macrocephala; dosage combination three: 13g Poria cocos and 9g Atractylodes macrocephala. Through the effect prediction model, their objective function values ​​are calculated to be 0.58, 0.61, and 0.55, respectively. Based on the results, it is speculated that there may be better effects near dosage combination two (function value of 0.61), so predictions continue near this dosage combination (e.g., 17g Poria cocos and 12g Atractylodes macrocephala). After multiple iterations, the effect prediction model determined that the objective function value was maximized when the input was a certain dose combination (18 grams of Poria cocos and 8.5 grams of Atractylodes macrocephala). This dose combination can be used as the updated intervention plan, and the adjusted plan results are output (Poria cocos increased by 3 grams and Atractylodes macrocephala decreased by 1.5 grams).

[0103] It must be emphasized again that the results of this adjustment plan are a set of specific parameter modification suggestions calculated by an algorithm. They do not constitute a clinical diagnosis or treatment prescription, nor can they be directly used as a basis for medical decisions. Essentially, they are merely reference information calculated using a complex algorithm; whether to adopt this information and how to adopt it depends entirely on the independent professional judgment of the physician.

[0104] By defining parameters to adjust the search space, the specific scope and boundaries of the plan optimization were clarified, improving the targeting of the optimization solution. Constructing a predictive objective function quantified the overall benefits of the adjustment scheme, improving the accuracy of the optimization direction. Iterative optimization using a Bayesian optimization algorithm efficiently explored the optimal solution, reducing blind searches and improving solution efficiency. Optimization under preset constraints ensured the safety and compliance of the adjustment scheme, ultimately achieving the output of the solution with the optimal overall benefit.

[0105] In the above implementation, firstly, through feature extraction and fusion processing, target profile data that comprehensively reflects the target object's state and treatment background is obtained, laying a solid data foundation for subsequent analysis. Then, by simulating TCM diagnostic thinking through matching and reasoning based on a pre-set knowledge graph and target profile data, a state assessment result that can quantitatively evaluate the current health status is obtained. Finally, an effect prediction model is used for optimization, automatically exploring and recommending the optimal plan adjustment scheme under safety constraints, improving the scientific rigor and safety of the adjustment suggestions, and providing a reliable basis for adjustment schemes for feedback evaluation.

[0106] It should be understood that the TCM-oriented planning and evaluation system in this embodiment is presented in the form of functional units. Here, a unit refers to an ASIC (Application Specific Integrated Circuit), a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the aforementioned functions. Each module in the TCM-oriented planning and evaluation system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0107] This specification also provides a method for planning and evaluating traditional Chinese medicine diagnosis and treatment, which includes the following steps: S302. Continuously collect dynamic physiological parameters using wearable devices.

[0108] Among them, dynamic physiological parameters refer to time-series data that can reflect the physiological basis of the target object.

[0109] S304. Continuously collect parameters of the four diagnostic methods of traditional Chinese medicine using dedicated acquisition equipment.

[0110] Among them, the parameters of the four diagnostic methods in traditional Chinese medicine refer to the digital data that can characterize the physical signs of the target object in traditional Chinese medicine, and the parameters of the four diagnostic methods in traditional Chinese medicine include the parameters of inspection, auscultation and olfaction, inquiry and palpation.

[0111] S306. Based on the historical text data of the target object, perform structured processing to obtain historical parameter data.

[0112] S308. Temporal convolution extraction is performed based on dynamic physiological parameters to obtain temporal feature data.

[0113] S310. Based on the parameters of the four diagnostic methods of traditional Chinese medicine, perform image and / or signal extraction processing to obtain quantitative feature data.

[0114] S312. The time-series feature data and quantized feature data are concatenated to obtain a dynamic feature vector.

[0115] S314. Based on the current intervention plan, perform structural analysis and vectorization to obtain the plan feature vector. The plan feature vector reflects the treatment intention for the target individual.

[0116] S316. Attention fusion processing is performed based on dynamic feature vectors and planned feature vectors to obtain target profile data.

[0117] Among them, the target profile data is used to reflect the current state of the target object at the current stage of diagnosis and treatment.

[0118] S318. Calculate state similarity based on the preset knowledge graph and target profile data to obtain state similarity results. The state similarity results reflect the degree of similarity between at least one candidate state in the preset knowledge graph and the current state.

[0119] S320. Utilize a multi-head attention model to perform state weight allocation and analysis based on target profile data and state similarity results to obtain state evaluation results.

[0120] S322. Adjust the search space based on the parameters defined in the current intervention plan, and construct the prediction objective function based on the state assessment results.

[0121] S324. Using the Bayesian optimization algorithm, the objective function is iteratively optimized based on preset constraints in the parameter adjustment search space to obtain the optimized objective function.

[0122] S326. If the optimized objective function meets the preset iteration conditions, the updated intervention plan corresponding to the optimized objective function shall be used as the result of the scheme adjustment.

[0123] S328. Based on the results of the program adjustment, conduct simulation and evaluation of the current intervention plan to obtain feedback evaluation results.

[0124] The feedback evaluation results are used to provide quantitative references for adjusting diagnosis and treatment for decision-makers, so as to generate subsequent intervention plans for the target subjects in the next stage of diagnosis and treatment.

[0125] For specific limitations regarding the planning and evaluation method for TCM diagnosis and treatment, please refer to the limitations of the planning and evaluation system for TCM diagnosis and treatment mentioned above, which will not be repeated here. It should be understood that although the steps in the flowchart above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction for the execution of these steps; they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple steps or stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in rotation with other steps or at least some of the steps or stages within other steps.

[0126] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 3 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processor can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations. Figure 3 Take a processor 10 as an example.

[0127] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0128] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0129] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0130] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0131] The computer device also includes an input device 30 and an output device 40. The processor 10, memory 20, input device 30, and output device 40 can be connected via a bus or other means. Figure 3 Taking the example of a connection between China and Israel via a bus.

[0132] Input device 30 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the computer device, such as a touchscreen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 40 may include display devices, auxiliary lighting devices (e.g., LEDs), and haptic feedback devices (e.g., vibration motors). The aforementioned display devices include, but are not limited to, liquid crystal displays, light-emitting diodes, displays, and plasma displays. In some alternative embodiments, the display device may be a touchscreen.

[0133] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and subsequently stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the methods shown in the above embodiments are implemented.

[0134] This application provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method of any embodiment of this application.

[0135] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

[0136] The systems, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0137] For ease of description, the above devices are described separately by function as various units. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware.

[0138] Those skilled in the art will understand that embodiments of this application can be provided as a system or computer program product. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0139] This application is described with reference to flowchart illustrations and / or block diagrams of systems and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0140] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0141] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0142] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the method embodiments are basically similar to the system embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0143] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0144] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A planning and evaluation system for traditional Chinese medicine diagnosis and treatment, characterized in that, The system includes: The data acquisition module is used to acquire multimodal monitoring data of the target object in the current diagnosis and treatment stage; wherein, the multimodal monitoring data includes dynamic physiological parameters and TCM four diagnostic parameters; the dynamic physiological parameters refer to time-series data that can reflect the target object's physiological basis; the TCM four diagnostic parameters refer to digital data that can characterize the target object's TCM physical signs; The effect prediction module is used to perform effect simulation prediction based on the multimodal monitoring data and the current intervention plan of the target object, so as to obtain the plan adjustment results; The feedback evaluation module is used to simulate and evaluate the current intervention plan based on the adjustment results of the plan, and obtain feedback evaluation results; wherein, the feedback evaluation results are used to provide the decision-making end with quantitative reference for diagnosis and treatment adjustment, so as to generate the subsequent intervention plan for the target object in the next stage of diagnosis and treatment.

2. The system according to claim 1, characterized in that, The data acquisition module is also used for: The dynamic physiological parameters are continuously collected using wearable devices; The parameters of the four diagnostic methods of traditional Chinese medicine are continuously collected using a dedicated data acquisition device; wherein, the parameters of the four diagnostic methods of traditional Chinese medicine include observation, auscultation and olfaction, inquiry and palpation. Historical parameter data is obtained by performing structured processing on the historical text data of the target object. The dynamic physiological parameters, the four diagnostic parameters of traditional Chinese medicine, and the historical parameter data are integrated to obtain the multimodal monitoring data.

3. The system according to claim 1, characterized in that, The effect prediction module is also used for: Feature extraction and fusion processing are performed based on the multimodal monitoring data and the current intervention plan to obtain target profile data; wherein, the target profile data is used to reflect the current state of the target object in the current diagnosis and treatment stage; Based on the preset knowledge graph and the target profile data, matching and reasoning are performed to obtain the state evaluation result; The effect prediction model is used to optimize the solution based on the current intervention plan to obtain the plan adjustment result; wherein, the effect prediction model is a model constructed based on the current intervention plan data, under preset constraints, and with the state assessment result and the target profile data as input.

4. The system according to claim 3, characterized in that, The effect prediction module also includes a data fusion unit, used for: Based on the multimodal monitoring data, multidimensional feature extraction is performed to obtain a dynamic feature vector; Based on the current intervention plan, structural analysis and vectorization are performed to obtain the plan feature vector; wherein, the plan feature vector is used to reflect the treatment intention for the target object; Attention fusion processing is performed based on the dynamic feature vector and the planned feature vector to obtain target profile data.

5. The system according to claim 4, characterized in that, The data fusion unit further includes a data preprocessing subunit, used for: Temporal convolution extraction is performed based on the dynamic physiological parameters to obtain temporal feature data; Image and / or signal extraction processing is performed based on the parameters of the four diagnostic methods of traditional Chinese medicine to obtain quantitative feature data; The time-series feature data and the quantized feature data are concatenated to obtain the dynamic feature vector.

6. The system according to claim 3, characterized in that, The effect prediction module further includes a state reasoning unit, used for: Based on the preset knowledge graph and the target profile data, state similarity is calculated to obtain state similarity results; wherein, the state similarity results are used to reflect the degree of similarity between at least one candidate state in the preset knowledge graph and the current state; A multi-head attention model is used to perform state weight allocation and analysis based on the target profile data and the state similarity results to obtain the state evaluation results.

7. The system according to claim 3, characterized in that, The effect prediction module also includes an optimization solution unit, used for: The search space is adjusted based on the parameters defined in the current intervention plan, and a predictive objective function is constructed based on the state assessment results. The Bayesian optimization algorithm is used to iteratively optimize the prediction objective function based on the preset constraints in the parameter adjustment search space to obtain the optimized objective function. If the optimized objective function satisfies the preset iteration conditions, the updated intervention plan corresponding to the optimized objective function will be used as the scheme adjustment result.

8. A computer device, characterized in that, Includes the planning and evaluation system for TCM diagnosis and treatment as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, implements the functions of the planning and evaluation system for TCM diagnosis and treatment as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the functions of the planning and evaluation system for TCM diagnosis and treatment as described in any one of claims 1 to 7.

Images