Intelligent management method and system for traditional Chinese medicine nursing data

Abstract

The invention relates to a traditional Chinese medicine nursing data intelligent management method and system, and the method comprises the steps: collecting traditional Chinese medicine nursing data, and converting the traditional Chinese medicine nursing data into structured semantic data and a standardized sign data set; constructing a dynamic knowledge graph according to the structured semantic data and the standardized physical sign data set; constructing a syndrome evolution model based on the dynamic knowledge graph, and generating an initial nursing scheme; conflict detection and intervention effect prediction are carried out on the initial nursing scheme, and a self-adaptive strategy feedback model is formed based on evaluation data after nursing implementation; generating nursing decision suggestions by utilizing case reasoning according to the self-adaptive strategy feedback model and the dynamic knowledge graph; in the execution process of the steps, federal learning is adopted to realize distributed processing of data, and a differential privacy mechanism is combined to guarantee dynamic desensitization and privacy security of the traditional Chinese medicine nursing data. The intelligent effect of traditional Chinese medicine data management can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data management, and in particular to a traditional Chinese medicine nursing data intelligent management method and system, electronic equipment and a non-transitory computer readable storage medium. BACKGROUND

[0002] Currently, the management of traditional Chinese medicine nursing data mainly relies on manual recording and experience accumulation. Nurses record the patient's signs, symptoms, syndrome types and nursing measures through paper medical records or basic electronic systems. Some hospitals have gradually introduced electronic medical record systems to digitize and statistically analyze traditional Chinese medicine nursing operations.

[0003] However, the existing methods generally have problems such as information fragmentation, low data utilization rate, and lack of intelligent assistance. On the one hand, traditional Chinese medicine nursing data has characteristics such as multi-source heterogeneity, strong subjectivity, and complex time sequence, and traditional management methods are difficult to achieve effective association and dynamic tracking between data. On the other hand, the current system focuses on static recording and lacks intelligent analysis and decision support for patient condition evolution trends, intervention measure effects and nursing path optimization, making it difficult to meet the clinical needs of fine and long-term management. SUMMARY

[0004] The present application provides a traditional Chinese medicine nursing data intelligent management method, system, electronic equipment and non-transitory computer readable storage medium that can improve the intelligent effect of traditional Chinese medicine data management.

[0005] The technical solution of the present application to solve the above technical problems is as follows: The present application provides a traditional Chinese medicine nursing data intelligent management method, which comprises: Collecting traditional Chinese medicine nursing data and converting the traditional Chinese medicine nursing data into structured semantic data and standardized sign data sets; According to the structured semantic data and the standardized sign data sets, a traditional Chinese medicine nursing knowledge graph is constructed, and a dynamic knowledge graph with spatio-temporal association is generated by combining fuzzy sets and graph neural networks; Based on the dynamic knowledge graph, a syndrome evolution model is constructed, and a current syndrome conclusion is obtained by syndrome reasoning based on the syndrome evolution model. An initial nursing plan is generated by calling a reinforcement learning model according to the current syndrome conclusion; The initial nursing plan is subjected to conflict detection and intervention effect prediction, and based on the evaluation data after nursing implementation, a self-adaptive strategy feedback model is formed by modeling the symptom and measure response relationship through contrast learning; According to the self-adaptive strategy feedback model and the dynamic knowledge graph, nursing decision suggestions are generated by case reasoning, visual operation guidance is performed through augmented reality, and nursing task allocation and execution are completed with the help of an intelligent scheduling system; In the process of executing the above steps, federated learning is used to realize distributed processing of data, and a differential privacy mechanism is combined to guarantee dynamic desensitization and privacy security of traditional Chinese medicine nursing data.

[0006] Optionally, the collection of the traditional Chinese medicine nursing data and the conversion of the traditional Chinese medicine nursing data into structured semantic data and a standardized sign data set comprise: Based on tongue image data of the traditional Chinese medicine nursing data, tongue texture feature values and coating coverage rate feature values are extracted through a convolutional neural network to generate standardized tongue diagnosis parameters; Based on nursing voice records of the traditional Chinese medicine nursing data, the voice data are transcribed into structured text through a voice recognition model fused with a traditional Chinese medicine terminology table to form a traditional Chinese medicine syndrome description data set.

[0007] Optionally, the construction of a traditional Chinese medicine nursing knowledge graph according to the structured semantic data and the standardized sign data set and the generation of a dynamic knowledge graph combined with a fuzzy set and a graph neural network comprise: Based on pulse waveform data of the standardized sign data set, an associated vector of pulse and syndrome is constructed by using a node embedding method in a graph neural network; And based on the associated vector and the structured semantic data, the probability distribution of each syndrome is calculated through a fuzzy membership function to construct a dynamic mapping network of syndrome and method as the dynamic knowledge graph.

[0008] Optionally, the construction of a syndrome evolution model based on the dynamic knowledge graph and the syndrome reasoning based on the syndrome evolution model to obtain a current syndrome conclusion comprise: Based on time sequence nursing records in the dynamic knowledge graph, a long short-term memory neural network is used to mine symptom evolution rules to generate a syndrome transition probability matrix; According to the syndrome transition probability matrix and the sign information of a current patient, a syndrome confidence distribution is output as the current syndrome conclusion through a Bayesian network reasoning module.

[0009] Optionally, the calling of a reinforcement learning model to generate an initial nursing scheme according to the current syndrome conclusion comprises: Based on the syndrome confidence distribution, the expected intervention effect of each nursing measure is calculated through a reward function in a reinforcement learning model to generate a joint intervention strategy of acupoint stimulation and drug compatibility as the initial nursing scheme.

[0010] Optionally, the conflict detection and intervention effect prediction of the initial nursing scheme comprise: Based on a historical nursing scheme database, a graph convolutional neural network is used to detect contraindication relationships between nursing measures to generate a warning list of nursing conflicts. adjust the initial nursing scheme according to the early warning list, and output an optimized nursing scheme after risk correction as the evaluation data.

[0011] Optionally, the method further comprises: Based on the change curve of the patient's physiological parameters collected after the implementation of the initial nursing scheme, a feature space of symptom response is constructed using a contrast learning method, and a sensitivity index of nursing measures is generated based on the feature space as the evaluation data.

[0012] Optionally, the nursing decision suggestion is generated by case reasoning based on the adaptive strategy feedback model and the dynamic knowledge graph, comprising: Based on the symptom and measure response relationship represented in the adaptive strategy feedback model, a case set of preferred nursing is constructed by matching historical cases with a similarity higher than a preset value through a case reasoning engine. Combined with the patient's physical characteristics, an individualized nursing path is generated from the case set through a weighted fusion algorithm as the nursing decision suggestion.

[0013] Optionally, the visualization operation guidance through augmented reality comprises: According to the acupoint operation steps in the individualized nursing path, a three-dimensional meridian navigation path is generated through the spatial positioning module of the augmented reality device. Based on the deviation between the real-time hand movement trajectory of the nursing staff and the standard hand method, the display intensity and prompt mode of the navigation guidance are dynamically adjusted.

[0014] The present application also provides a traditional Chinese medicine nursing data intelligent management system, the system comprises: A data acquisition module is used to collect traditional Chinese medicine nursing data and convert the traditional Chinese medicine nursing data into structured semantic data and standardized physical data set. A graph generation module is used to construct a traditional Chinese medicine nursing knowledge graph according to the structured semantic data and the standardized physical data set, and generate a dynamic knowledge graph with spatio-temporal correlation by combining fuzzy sets and graph neural networks. An initial scheme module is used to construct a syndrome evolution model based on the dynamic knowledge graph, perform syndrome reasoning based on the syndrome evolution model to obtain a current syndrome conclusion, and call a reinforcement learning model to generate an initial nursing scheme according to the current syndrome conclusion. A model generation module is used to perform conflict detection and intervention effect prediction on the initial nursing scheme, and form an adaptive strategy feedback model by modeling the symptom and measure response relationship through contrast learning based on the evaluation data after nursing implementation. A nursing decision module is configured to generate a nursing decision suggestion by case-based reasoning based on the adaptive strategy feedback model and the dynamic knowledge graph, to visually guide an operation by augmented reality, and to complete nursing task allocation and execution by means of an intelligent scheduling system. An encryption management module is configured to implement distributed processing of data by federated learning during execution of the above modules, and to guarantee dynamic desensitization and privacy security of the traditional Chinese medicine nursing data by combining a differential privacy mechanism.

[0015] In addition, to achieve the above object, the present application further provides an electronic device, comprising: a memory for storing a computer software program; a processor for reading and executing the computer software program, thereby realizing the above-mentioned intelligent management method for traditional Chinese medicine nursing data.

[0016] In addition, to achieve the above object, the present application further provides a non-transitory computer readable storage medium, wherein the storage medium stores a computer software program, and the computer software program is executed by a processor to realize the above-mentioned intelligent management method for traditional Chinese medicine nursing data.

[0017] The present application has the following beneficial effects: (1) The present application realizes automatic analysis and standardized processing of voice, image and physiological parameters by combining multi-modal acquisition with natural language processing and deep learning, effectively solving the problems of non-structured, strong subjectivity and difficulty of utilization of traditional Chinese medicine nursing records.

[0018] (2) The present application constructs a multi-dimensional dynamic knowledge graph covering the correlation model of "syndrome - law - effect", introduces fuzzy set theory to deal with the uncertainty of traditional Chinese medicine concepts, and automatically mines the potential relationship between knowledge through graph neural network, realizing the structuring, computability and self-updating of knowledge.

[0019] (3) The present application models the evolution law of syndrome by combining recurrent neural network with Bayesian inference, adapts to the dynamic characteristics of the change of patient's syndrome type over time, and then outputs individualized nursing strategy based on reinforcement learning, effectively enhancing the timeliness and targeting of the scheme. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of the intelligent management method for traditional Chinese medicine nursing data provided by the present application; Figure 2 A structural schematic diagram of the intelligent management system for traditional Chinese medicine nursing data provided by the present application; Figure 3 A hardware structural schematic diagram of a possible electronic device provided by the present application; Figure 4 A hardware structural schematic diagram of a possible computer readable storage medium provided by the present application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the present application.

[0022] In the description of the present application, the terms "first", "second" are used only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0023] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or explanation". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that a person skilled in the art can realize the present application without using these specific details. In other examples, well-known structures and processes will not be described in detail to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the shown embodiments, but is consistent with the broadest scope in accordance with the principles and characteristics disclosed.

[0024] Please refer to Figure 1 , a flowchart of a traditional Chinese medicine nursing data intelligent management method of the present application is provided, including the following steps: Step S1: Collecting traditional Chinese medicine nursing data and converting the traditional Chinese medicine nursing data into structured semantic data and standardized sign data set.

[0025] Specifically, based on the traditional Chinese medicine nursing ontology library, the patient's voice, image and biological parameter data can be collected through a multi-modal acquisition terminal, the structured semantic data can be generated by analyzing the unstructured voice record using natural language processing technology, and the standardized sign data set can be generated by extracting features from the tongue diagnosis image and pulse diagnosis image using a deep learning model.

[0026] In some embodiments, step S1 can include: Based on the tongue image data of the traditional Chinese medicine nursing data, the tongue texture feature value and the coating coverage feature value are extracted through a convolutional neural network to generate standardized tongue diagnosis parameters; Based on the nursing voice records of traditional Chinese medicine nursing data, the voice data is transcribed into structured text through a speech recognition model integrated with the traditional Chinese medicine terminology vocabulary to form a traditional Chinese medicine symptom description dataset.

[0027] In a specific implementation, for the standardized processing of tongue diagnosis-related data, preferably, a convolutional neural network (CNN) model can be used to extract features from key areas in the image based on the collected tongue image data. Specifically, the CNN model is trained to identify visual features of the tongue, such as color, texture, and moisture, and focuses on extracting tongue texture feature values ​​and tongue coating coverage feature values ​​related to TCM syndrome differentiation. Tongue texture features may include but are not limited to morphological parameters such as granularity of the tongue surface, crack depth, and serrated edges; tongue coating coverage features can calculate the proportion of tongue coating in different areas (such as the tip, middle, and root of the tongue) based on pixel distribution area. The above-mentioned feature values ​​are quantified to form uniformly coded standardized tongue diagnosis parameters, which can be used as important physical sign indicators for input into subsequent syndrome differentiation models.

[0028] Furthermore, to process the speech data generated during the nursing process, it is preferred that a speech recognition model that incorporates a TCM vocabulary be constructed based on the voice recording data entered on-site by nursing staff. The speech recognition model can be based on a deep neural network structure (such as an end-to-end ASR model), introducing a terminology library, word vectors, and pronunciation dictionary specifically built for the TCM field, to achieve accurate recognition and transcription of professional terms such as TCM symptoms, syndromes, and signs. The transcribed results are subjected to structured semantic annotation to extract the corresponding symptom entities, attributes, and related contextual information, thereby generating a TCM symptom description dataset that conforms to TCM nursing expression standards. This dataset can be further used to construct semantic graph nodes, train dialectical rules, or recommend individualized nursing pathways.

[0029] Through the above method, the present invention can effectively convert the original tongue diagnosis images and voice records into standardized and structured nursing data input, providing high-quality data support for subsequent dynamic knowledge graph construction and intelligent dialectical reasoning.

[0030] Step S2: Based on the structured semantic data and standardized vital sign dataset, a knowledge graph of traditional Chinese medicine nursing is constructed, and a dynamic knowledge graph with spatiotemporal association is generated by combining fuzzy sets and graph neural networks.

[0031] Specifically, based on structured semantic data and standardized physical sign data sets, a multidimensional knowledge graph containing the "syndrome-method-effect" relationship can be constructed, combined with fuzzy set theory to represent the uncertainty of syndrome concepts, and a graph neural network algorithm can be used to mine the potential spatiotemporal associations between nodes to output a dynamic traditional Chinese medicine nursing knowledge graph.

[0032] In some embodiments, step S2 may include: Based on the pulse waveform data of the standardized physical sign dataset, the node embedding method in the graph neural network is used to construct the association vector between pulse condition and syndrome; Based on the association vector and structured semantic data, the probability distribution of each syndrome is calculated through the fuzzy membership function, and a dynamic syndrome-method mapping network is constructed as a dynamic knowledge graph.

[0033] In a specific implementation, in order to achieve effective modeling and dynamic association between TCM pulse and syndromes, preferably, an association representation for syndrome differentiation reasoning can be constructed based on the pulse waveform data in the standardized vital sign data set obtained in step S1.

[0034] Specifically, pulse waveform data can be derived from pulse signals acquired by pressure sensors, photoplethysmography (PPG), or other non-contact imaging devices. After preprocessing (including filtering, normalization, and feature alignment), it is input into a graph neural network (GNN) structure. In a GNN, nodes represent different types of pulse characteristics (such as floating, sinking, stringy, slippery, thin, and surging), and edges represent the temporal or coordinated changes between pulse patterns. Using node embedding technology, high-dimensional graph structure information can be compressed into a low-dimensional vector representation, thereby constructing a vectorized association between each pulse node and the associated syndrome, forming a pulse-symptom association vector set.

[0035] Furthermore, by fusing the aforementioned association vectors with structured semantic data (including patient complaints, symptom descriptions, and physical constitution information), a quantitative estimate of the credibility of candidate syndromes can be obtained. The present invention preferably employs a fuzzy membership function to model the uncertainty of syndromes. This function assigns a membership value between 0 and 1 to each syndrome, representing the degree of match between the syndrome and the current semantic state of the physical sign. This membership function can be calculated based on cosine similarity, Euclidean distance, or a correlation score based on a learning model between the pulse feature vector and the semantic feature vector.

[0036] Finally, based on the calculated distribution of syndrome membership, combined with the knowledge mapping relationship between "syndrome" and "method" (i.e., nursing intervention methods) in Traditional Chinese Medicine theory, a dynamic "syndrome-method" mapping network is constructed. As the core component of the dynamic knowledge graph of the present invention, this mapping network supports automatic updates, personalized adaptation, and graph structure reasoning driven by time series data, providing accurate and scalable knowledge support for subsequent syndrome differentiation decisions and recommended treatment plans.

[0037] By the above manner, the problem that traditional pulse conditions of traditional Chinese medicine are difficult to quantify and knowledge-driven analysis is effectively solved, and utilization efficiency of the physical sign information in the knowledge graph construction process and intelligent level of syndrome differentiation are improved.

[0038] Step S3: constructing a syndrome evolution model based on the dynamic knowledge graph, performing syndrome reasoning based on the syndrome evolution model to obtain a current syndrome differentiation conclusion, and calling a reinforcement learning model to generate an initial nursing scheme according to the current syndrome differentiation conclusion.

[0039] Specifically, the syndrome evolution model can be constructed based on the dynamic knowledge graph by using a recurrent neural network to identify the time sequence law of symptom changes, and a current syndrome differentiation conclusion can be generated by combining a Bayesian network reasoning, and an initial nursing scheme can be further output by calling a reinforcement learning model.

[0040] In some embodiments, step S3 can include: Based on the time sequence nursing records in the dynamic knowledge graph, a long short-term memory neural network is used to mine the symptom evolution law to generate a syndrome transfer probability matrix; According to the syndrome transfer probability matrix and the physical sign information of the current patient, a syndrome type confidence distribution is output by a Bayesian network reasoning module as a current syndrome differentiation conclusion.

[0041] In a specific implementation, to realize dynamic modeling and syndrome differentiation reasoning of the syndrome evolution process of a patient, preferably, the time sequence law of traditional Chinese medicine symptoms can be mined based on the time sequence nursing records in the constructed dynamic knowledge graph.

[0042] Specifically, the time sequence nursing records include but are not limited to daily physical sign parameters, symptom complaints, nursing intervention contents and response results of the patient, and other time-labeled data. By inputting such multi-time point sequence information into a long short-term memory neural network (LSTM) model, the evolution trend and correlation structure of different symptoms over time can be learned and extracted by the modeling ability of the LSTM model on time sequence dependency.

[0043] The LSTM model effectively captures short-term and long-term information through a gating mechanism in the training process, and then generates a syndrome transfer probability matrix describing the transformation relationship between different syndromes. The probability matrix is used to quantify the probability of transformation of a certain syndrome to other syndromes under a specific time state, embodies the “syndrome changes with time” feature in traditional Chinese medicine syndrome differentiation process, and has dynamic and predictive properties.

[0044] Further, in combination with the current patient's physical sign information (including standardized tongue diagnosis parameters, pulse diagnosis waveform feature values, and structured semantic data), the same can be taken as a conditional input, and probability inference is performed through a Bayesian network inference module. The Bayesian network is a directed acyclic graph structure, the nodes represent syndrome variables and physical sign variables, and the edges represent the dependency relationship between variables. The likelihood of each syndrome under the current state is modeled by combining the prior probability and the conditional probability.

[0045] Through the inference of the above Bayesian network, a syndrome confidence distribution can be output, which represents the probability value of the current patient meeting multiple syndromes at the same time. This distribution serves as the current syndrome differentiation conclusion in the present application, providing a quantitative basis for subsequent intelligent generation of nursing plans, and also realizing the computability and transparency of the TCM syndrome differentiation process.

[0046] In the above manner, the present application can effectively identify the dynamic evolution path of the patient's symptoms, integrate the historical change trend and the current physical sign state, realize the time sequence modeling and intelligent decision inference of syndrome recognition, and improve the scientific nature and individualization level of TCM nursing process.

[0047] In some embodiments, based on the syndrome confidence distribution, the expected intervention effect of each nursing measure is calculated through a reward function in a reinforcement learning model, and a joint intervention strategy of acupoint stimulation and drug compatibility is generated as an initial nursing plan.

[0048] In a specific implementation, to realize intelligent generation of nursing intervention plans, preferably, based on the syndrome confidence distribution output by the aforementioned Bayesian network inference module, a reinforcement learning model is called to construct a nursing strategy optimization mechanism for multiple syndrome states.

[0049] Specifically, the syndrome confidence distribution reflects the matching probability of the current patient in multiple candidate syndromes, which can be regarded as the state input of nursing decision. On this basis, a reinforcement learning model is constructed with TCM nursing measures as the action space (Action Space) and the intervention effect as the reward function (Reward Function), which is used to evaluate the expected effect of different nursing intervention combinations under the current syndrome state.

[0050] Nursing measures include but are not limited to acupoint stimulation (such as acupuncture, finger pressure, moxibustion, etc.) and drug compatibility (such as external application and Chinese medicine decoction, etc.) and other TCM nursing intervention means. In the training stage, the model learns from the historical nursing path and nursing effect data to establish a mapping relationship between state-action-reward; in the application stage, the model selects the action combination with the maximum cumulative benefit under the given reward expectation according to the current syndrome state, i.e. generates a nursing strategy with the highest expected intervention effect.

[0051] The reward function can comprehensively consider the evaluation indexes such as symptom relief speed, physiological parameter improvement amplitude, patient subjective satisfaction, and quantize the effectiveness of each intervention measure or combined measure under different syndrome type states. In specific implementation, a deep reinforcement learning algorithm (such as DQN, DDPG, PPO, etc.) can be used for strategy optimization and output.

[0052] Finally, the system can output results according to the reinforcement learning model to generate an initial nursing scheme composed of acupoint stimulation measures and drug compatibility measures. The scheme has the advantages of strong pertinence, timely response, combination of theory and experience, and can provide individualized intervention path according to the individual characteristics of the patient in actual nursing, significantly improving the scientificity and efficiency of the TCM nursing scheme.

[0053] In the above manner, the present application realizes the whole process automation from data-driven syndrome differentiation to intelligent generation of nursing scheme, and provides a learnable, adjustable and self-adaptive intelligent decision support system for TCM nursing.

[0054] Step S4: Conflict detection and intervention effect prediction are performed on the initial nursing scheme, and based on the evaluation data after nursing implementation, a self-adaptive strategy feedback model is formed by modeling the symptom-measure response relationship through contrastive learning.

[0055] Specifically, the initial nursing scheme can be subjected to conflict detection and intervention effect prediction between measures, an optimized nursing scheme is generated, and based on the evaluation data after nursing implementation, a self-adaptive strategy feedback model for iterative adjustment is formed by modeling the symptom-measure response relationship through contrastive learning.

[0056] In some embodiments, step S4 can include: Based on the historical nursing scheme database, the contraindication relationship between nursing measures is detected by a graph convolutional neural network, and a warning list of nursing conflicts is generated; According to the warning list, the initial nursing scheme is adjusted, and the risk-corrected optimized nursing scheme is output as the evaluation data.

[0057] In specific implementation, in order to improve the safety and rationality of the nursing scheme, the joint intervention strategy initially generated can be further subjected to conflict detection and risk correction based on the historical nursing scheme database.

[0058] Specifically, the historical nursing scheme database contains a large number of real or simulated nursing cases, records the combination of nursing measures accepted by patients with different syndromes, the implementation time sequence, the curative effect feedback and adverse reaction information. Based on the database, a nursing measure graph structure is constructed, each nursing measure (including specific acupoint stimulation, drug compatibility, operation method, etc.) is taken as a node of the graph, and the co-occurrence relationship or logical order thereof in the same scheme is represented as an edge.

[0059] Based on the above graph structure, a graph convolutional neural network (GCN) is preferably used for relationship modeling and conflict pattern mining. The GCN model can capture high-order semantic dependencies between nursing measures, and identify measure combinations that may have contraindications, conflicts or interference effects under certain syndrome or constitution states.

[0060] The trained GCN model can evaluate the conflicts of the measure combinations included in the initial nursing plan, and output a set of warning lists of nursing conflicts, which clearly indicate the potential incompatible measures, corresponding conflict types (such as pharmacological antagonism, meridian interference, indication deviation, etc.) and risk level scores.

[0061] According to the warning list, the system performs logical verification and strategy adjustment on the initial nursing plan, preferably with the optimization goal of minimizing the loss of intervention effect, automatically replacing or recombining the conflicting intervention measures, and outputting the risk-corrected optimized nursing plan. This scheme not only retains the key effective components in the original scheme, but also avoids the combination conflicts that may cause adverse reactions or decreased efficacy.

[0062] The optimized nursing plan will be input as the basic data for subsequent nursing effect evaluation into the next stage of dynamic feedback and strategy iteration process.

[0063] In the above manner, the present application introduces a graph neural network driven conflict detection mechanism based on the generation of nursing plans, significantly improving the safety, rationality and practical value of traditional Chinese medicine nursing intervention plans, and reducing the risk of experience dependence and human error.

[0064] In some embodiments, based on the physiological parameter change curve of the patient collected after the implementation of the initial nursing plan, a contrast learning method can be used to construct a feature space of symptom response, and a sensitivity index of the nursing measure can be generated based on the feature space as evaluation data.

[0065] In specific implementations, to realize the quantitative evaluation and dynamic feedback of the actual effect of the optimized nursing plan, further based on the physiological parameter change information of the patient after the implementation of the nursing intervention, a symptom response model can be constructed and a sensitivity evaluation result of the nursing measure can be generated.

[0066] Specifically, after the implementation of the nursing plan is completed, the system continuously collects the physiological parameter data of the patient, including but not limited to pulse waveform, body temperature, blood pressure, skin conductance, respiratory rate, tongue evolution index and subjective symptom score, etc., and arranges them into a patient physiological parameter change curve in time sequence. The change curve and the nursing intervention measure establish a time alignment relationship, forming a "measure-response" corresponding sample pair.

[0067] Preferably, the above-mentioned sample pairs can be modeled using a contrastive learning method. The contrastive learning model trains an encoder network to map "positive sample pairs" (i.e., intervention-measure-response combinations that produce positive therapeutic effects after intervention) and "negative sample pairs" (i.e., combinations that show no improvement or worsening of symptoms after intervention) to a unified symptom-response feature space, and keeps response-similar measure combinations close together and non-responsive or ineffective combinations far apart in the space, thereby establishing an effective symptom change feature representation.

[0068] In the constructed feature space, the system can further cluster and calculate the distances of different nursing measures or measure combinations to evaluate their intervention effects on target symptoms under different constitutions, syndromes, and time points. By measuring the similarity of a certain measure to significantly improved response samples in the feature space, the effectiveness of the measure in a specific application scenario is quantified, thereby generating a sensitivity index for the measure.

[0069] The sensitivity index is used to measure the intervention ability and adaptability of a nursing measure on a target symptom, and a higher value represents a more explicit intervention effect and a more sensitive patient response, which can serve as an important basis for nursing path optimization, strategy screening, and individualized adjustment.

[0070] In the above manner, the present application realizes a real patient response data-driven nursing measure dynamic evaluation mechanism, improves the explainability and precision level of TCM nursing schemes, and constructs a closed-loop nursing management system of "collection-evaluation-feedback-optimization".

[0071] Step S5: According to the adaptive strategy feedback model and the dynamic knowledge graph, a nursing decision suggestion is generated using case-based reasoning, visual operation guidance is provided through augmented reality, and nursing task allocation and execution are completed with the help of an intelligent scheduling system.

[0072] Specifically, a nursing decision suggestion can be generated using case-based reasoning according to the adaptive strategy feedback model and the knowledge graph, the suggestion can be converted into visual operation guidance through augmented reality technology, and nursing tasks can be allocated to target nursing staff in combination with an intelligent scheduling system.

[0073] In some embodiments, step S5 can include: Based on the symptom-measure response relationship represented in the adaptive strategy feedback model, a case-based reasoning engine is used to match historical cases with a similarity higher than a preset value, and a case set of preferred nursing is constructed; In combination with patient constitution characteristics, a personalized nursing path is generated from the case set as a nursing decision suggestion through a weighted fusion algorithm.

[0074] In a specific implementation, to improve the individualization and empirical support capability of the nursing scheme, further based on the symptom and nursing measure response relationship represented in the adaptive strategy feedback model constructed in the foregoing, the generation and recommendation of the individualized nursing path can be realized through a case reasoning mechanism.

[0075] Specifically, the symptom-measure response relationship is obtained through contrastive learning and feature space modeling, reflecting the response sensitivity and intervention effect of a specific nursing measure under a specific symptom manifestation. Based on the known effective response combination in the model, the system constructs a case feature vector representation, covering information such as patient complaints, syndrome characteristics, sign parameters, nursing measures, and efficacy evaluation results.

[0076] In the implementation process, the system constructs a target case vector based on the comprehensive information of the current patient to be nursed, and performs high-dimensional feature matching in the historical case database through a case reasoning engine. Preferably, cosine similarity, Mahalanobis distance, or vector measurement methods based on graph matching can be used to identify the historical nursing samples closest to the current patient characteristics.

[0077] The system filters out a high-similarity case set based on a set similarity threshold (e.g., 85%), and constructs an optimal nursing case set. The case set contains multiple nursing paths that have achieved good intervention effects under similar syndromes, providing experience support and strategy reference for the current patient.

[0078] Further, to realize the individualization of the nursing path, the system also combines the constitution characteristics (such as cold, heat, deficiency, and excess, regional and seasonal adaptability, etc.) of the current patient with the constitution labels of the case set samples, and uses a weighted fusion algorithm to comprehensively evaluate the candidate nursing paths. This algorithm gives dynamic weight influence to the individual characteristics of the current patient while retaining the experience of similar cases, thereby generating an individualized nursing path.

[0079] The individualized nursing path can be output as the final nursing decision suggestion to the nursing execution module, and can support the implementation of nursing personnel through augmented reality visualization, intelligent terminal push, etc.

[0080] In the above manner, the present application not only realizes the intelligent inheritance and migration of TCM nursing experience, but also improves the accuracy and adaptability of the nursing path by integrating individual characteristics, providing an efficient and reliable auxiliary decision-making mechanism for the intelligent generation of TCM nursing schemes.

[0081] In some embodiments, step S5 can further include: According to the acupoint operation steps in the individualized nursing path, a three-dimensional meridian navigation path is generated through the spatial positioning module of the augmented reality device; Based on the deviation between the real-time hand movement trajectory of the nursing personnel and the standard hand technique, the display intensity and prompting mode of the navigation guidance are dynamically adjusted.

[0082] In a specific implementation, to improve the accuracy and standardized execution effect of acupoint operation in the personalized nursing path, preferably, the system further integrates an augmented reality device for providing visual meridian navigation and operation guidance.

[0083] Specifically, based on the acupoint stimulation steps in the generated personalized nursing path, the corresponding acupoint spatial position information, meridian attribution relationship and operation requirements are extracted. The system calls the spatial positioning module configured by the augmented reality (AR) device, uses a depth camera, an infrared sensor or a wearable positioning device to obtain the body posture model on the surface of the nursing object (patient) in real time.

[0084] Based on the registration relationship between the above human model and the standard acupoint library, the system can superimpose a three-dimensional meridian navigation path on the surface of the nursing object, which combines the requirements of the meridian direction, the depth of the acupoint level and the operation direction, dynamically generates the acupoint stimulation sequence and path trajectory in the corresponding nursing scheme. The navigation path can be presented as a semi-transparent floating line, a flashing indicator point or a pseudo-meridian vein graph to realize visual operation guidance.

[0085] During the nursing operation process, the system further obtains the real-time motion trajectory of the hands of the nursing staff, which can be collected by gesture recognition sensors, inertial measurement units (IMU) or optical tracking modules in the AR device. The hand trajectory is compared with the preset standard operation method (including stimulation position, pressing angle, force direction and frequency, etc.) in real time.

[0086] Preferably, the system calculates the deviation level according to the spatial deviation value and the time dynamic change between the two, and dynamically adjusts the display intensity and prompt mode of the navigation guidance accordingly. For example, when the nursing staff's operation deviation is small, only the faint navigation path is used for light guidance; when the deviation increases, the operation correction prompt is enhanced through the ways of enhanced display brightness, path color change, voice prompt or vibration feedback, etc., to ensure the accuracy and safety of the nursing operation.

[0087] In the above manner, the present application realizes an intelligent visual guidance mechanism based on augmented reality in the nursing path execution process, breaks through the bottleneck problems in traditional Chinese medicine nursing, such as the dependence on experience for acupoint positioning and the difficulty in quantifying the execution quality of manual operation, significantly improves the standardization degree and individual adaptation ability of Chinese medicine nursing scheme, and provides effective technical support for complex nursing operation.

[0088] Step S6: During the execution of steps S1 to S5, federated learning is used to realize distributed processing of data, and a differential privacy mechanism is combined to guarantee the dynamic desensitization and privacy security of the traditional Chinese medicine nursing data.

[0089] Specifically, during the execution of S1 to S5, sensitive data can be distributed modeled and locally updated based on the federated learning mechanism, while the differential privacy mechanism is used to inject noise and desensitize the data processing results at each stage, ensuring data privacy and ethical compliance.

[0090] In specific implementations, to ensure user privacy protection and ethical compliance throughout the data processing and intelligent analysis process, preferably, during the execution of steps S1 to S5, the system introduces a federated learning mechanism and a differential privacy mechanism to build a multi-level privacy security system.

[0091] Specifically, in the processes of nursing data collection, semantic analysis, knowledge graph construction, syndrome reasoning, scheme optimization and execution involved in steps S1 (multi-source data standardization processing) to S5 (cooperative decision and execution), the data processed may involve sensitive information such as patient personal identity, physical characteristics, symptom description, and nursing response. To achieve intelligent modeling and continuous optimization without revealing the original data, the system preferably uses a federated learning (Federated Learning) mechanism.

[0092] The federated learning mechanism is configured to collaboratively train TCM nursing related models (such as semantic analysis models, graph update modules, strategy generation engines, etc.) across multiple center nodes (such as different medical institutions and local nursing terminals). The core is that each participating node only trains local data and transmits model parameters (such as gradient, weight update information) rather than original data to the central coordination server. This approach effectively avoids privacy risks during centralized storage and transmission of patient data, enabling distributed modeling and localized model updating.

[0093] During the above federated training or intermediate result sharing process, to further enhance data protection, the system introduces a differential privacy (Differential Privacy) mechanism. This mechanism injects random noise conforming to Laplace distribution or Gaussian distribution into the parameters on each node or in the data processing results (such as structured text, graph node relationships, strategy weights, etc.), ensuring that attackers cannot infer any specific user's data from the global model or statistical output, meeting quantifiable privacy protection standards (such as $\epsilon$-differential privacy constraints).

[0094] In addition, the system can dynamically adjust the noise injection intensity and desensitization strategies according to the data sensitivity level and usage scenarios, such as generalization processing, hash encoding or pseudo-anonymization operations on geographic location information and individual characteristic variables, to further reduce the risk of data re-identification.

[0095] Through the combination of the above federal learning and differential privacy double mechanism, the application realizes efficient utilization and intelligent analysis of traditional Chinese nursing data, while effectively protecting the privacy rights of patients and the requirements of medical ethics, ensuring the compliance, security and credibility of the system in the whole life cycle of data.

[0096] Please refer to Figure 2 , Figure 2 The structure diagram of a traditional Chinese nursing data intelligent management system provided by the application is shown.

[0097] As Figure 2 shown, the traditional Chinese nursing data intelligent management system provided by the embodiment of the application comprises: A data acquisition module 301 is configured to acquire traditional Chinese nursing data and convert the traditional Chinese nursing data into structured semantic data and standardized physical sign data sets. A graph generation module 302 is configured to construct a traditional Chinese nursing knowledge graph according to the structured semantic data and the standardized physical sign data sets, and generate a dynamic knowledge graph with spatiotemporal correlation in combination with a fuzzy set and a graph neural network. An initial scheme module 303 is configured to construct a syndrome evolution model based on the dynamic knowledge graph, perform syndrome reasoning based on the syndrome evolution model to obtain a current syndrome conclusion, and call a reinforcement learning model to generate an initial nursing scheme according to the current syndrome conclusion. A model generation module 304 is configured to perform conflict detection and intervention effect prediction on the initial nursing scheme, and form an adaptive strategy feedback model by modeling the symptom and measure response relationship through comparative learning based on the evaluation data after nursing implementation. A nursing decision module 305 is configured to generate a nursing decision suggestion by using case reasoning according to the adaptive strategy feedback model and the dynamic knowledge graph, perform visual operation guidance through augmented reality, and complete nursing task allocation and execution with the help of an intelligent scheduling system. An encryption management module 306 is configured to realize distributed processing of data by using federal learning in the process of execution of the above modules, and protect the dynamic desensitization and privacy security of the traditional Chinese nursing data in combination with a differential privacy mechanism.

[0098] Please refer to Figure 3 , Figure 3 The embodiment of the electronic device provided by the embodiment of the application is shown. As Figure 3 shown, the electronic device 400 comprises a memory 410, a processor 420, and a computer program 411 stored in the memory 410 and capable of running on the processor 420, and the processor 420 implements the following steps when executing the computer program 411: S1: acquiring traditional Chinese nursing data and converting the traditional Chinese nursing data into structured semantic data and standardized physical sign data sets; S2: Construct a TCM nursing knowledge graph according to the structured semantic data and the standardized symptom data set, and generate a dynamic knowledge graph associated with time and space by combining a fuzzy set and a graph neural network; S3: Construct a syndrome evolution model based on the dynamic knowledge graph, and perform syndrome reasoning based on the syndrome evolution model to obtain a current syndrome conclusion, and call a reinforcement learning model to generate an initial nursing scheme according to the current syndrome conclusion; S4: Perform conflict detection and intervention effect prediction on the initial nursing scheme, and form an adaptive strategy feedback model by modeling the symptom and measure response relationship through comparative learning based on the evaluation data after nursing implementation; S5: Generate a nursing decision suggestion by case reasoning according to the adaptive strategy feedback model and the dynamic knowledge graph, perform visual operation guidance through augmented reality, and complete nursing task allocation and execution with the help of an intelligent scheduling system; S6: In the process of executing steps S1 to S5, federated learning is used to realize distributed processing of data, and a differential privacy mechanism is used to ensure dynamic desensitization and privacy security of TCM nursing data.

[0099] Please refer to Figure 4 , Figure 4 An embodiment of a computer readable storage medium provided by the embodiment of the application is shown in the figure. Figure 4 As shown in the figure, the embodiment provides a computer readable storage medium 500, which stores a computer program 411, and the computer program 411 is executed by a processor to implement the following steps: S1: Collect TCM nursing data, and convert the TCM nursing data into structured semantic data and a standardized symptom data set; S2: Construct a TCM nursing knowledge graph according to the structured semantic data and the standardized symptom data set, and generate a dynamic knowledge graph associated with time and space by combining a fuzzy set and a graph neural network; S3: Construct a syndrome evolution model based on the dynamic knowledge graph, and perform syndrome reasoning based on the syndrome evolution model to obtain a current syndrome conclusion, and call a reinforcement learning model to generate an initial nursing scheme according to the current syndrome conclusion; S4: Perform conflict detection and intervention effect prediction on the initial nursing scheme, and form an adaptive strategy feedback model by modeling the symptom and measure response relationship through comparative learning based on the evaluation data after nursing implementation; S5: Generate a nursing decision suggestion by case reasoning according to the adaptive strategy feedback model and the dynamic knowledge graph, perform visual operation guidance through augmented reality, and complete nursing task allocation and execution with the help of an intelligent scheduling system; S6: In the process of executing the steps S1 to S5, the distributed processing of data is achieved by using federated learning, and the dynamic desensitization and privacy security of traditional Chinese medicine nursing data are ensured by combining with the differential privacy mechanism.

[0100] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0101] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented 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.

[0102] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts 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, a special-purpose computer, an embedded computer, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 a system for implementing the functions specified in one or more flows and / or blocks.

[0103] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 a system for implementing the functions specified in one or more flows and / or blocks.

[0104] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 a system for implementing the functions specified in one or more flows and / or blocks.

[0105] While the preferred embodiments of the application have been described, additional variations and modifications can be made to these embodiments by those skilled in the art once they have the benefit of the present disclosure without departing from the spirit and scope of the application. Accordingly, it is intended that the appended claims include all such modifications and variations as fall within the scope of the present application.

[0106] It is apparent that those skilled in the art can make various changes and modifications to the application without departing from the spirit and scope of the application. It is therefore intended that the present application cover all such changes and modifications that are within its scope.

Claims

1. A method for intelligent management of traditional Chinese medicine nursing data, characterized in that: include: S1: Collecting TCM nursing data and converting the TCM nursing data into structured semantic data and standardized vital sign datasets; S2: Based on the structured semantic data and the standardized physical sign dataset, a TCM nursing knowledge graph is constructed, and a dynamic knowledge graph with spatiotemporal association is generated by combining fuzzy sets and graph neural networks; S3: constructing a syndrome evolution model based on the dynamic knowledge graph, performing syndrome reasoning based on the syndrome evolution model to obtain a current syndrome differentiation conclusion, and calling a reinforcement learning model to generate an initial nursing plan according to the current syndrome differentiation conclusion; S4: Detect conflicts and predict intervention effects of the initial nursing plan, and based on the evaluation data after nursing implementation, model the relationship between symptoms and measures through comparative learning to form an adaptive strategy feedback model; S5: Based on the adaptive strategy feedback model and the dynamic knowledge graph, case-based reasoning is used to generate nursing decision suggestions, visual operation guidance is provided through augmented reality, and nursing task allocation and execution are completed with the help of an intelligent scheduling system; S6: During the execution of steps S1 to S5, federated learning is used to implement distributed processing of data, and a differential privacy mechanism is combined to ensure dynamic desensitization and privacy security of traditional Chinese medicine nursing data.

2. The intelligent management method for traditional Chinese medicine nursing data according to claim 1, characterized in that: The collecting of TCM nursing data and converting the TCM nursing data into structured semantic data and standardized vital sign data sets include: Based on the tongue image data of the traditional Chinese medicine nursing data, the tongue texture feature value and the tongue coating coverage feature value are extracted through a convolutional neural network to generate standardized tongue diagnosis parameters; Based on the nursing voice records of the TCM nursing data, the voice data is transcribed into structured text through a speech recognition model integrated with a TCM terminology vocabulary to form a TCM symptom description dataset.

3. The intelligent management method for traditional Chinese medicine nursing data according to claim 2, characterized in that: The method of constructing a TCM nursing knowledge graph based on the structured semantic data and the standardized physical sign dataset, and generating a spatiotemporal dynamic knowledge graph by combining fuzzy sets and graph neural networks, includes: Based on the pulse waveform data of the standardized vital sign data set, a node embedding method in a graph neural network is used to construct an association vector between pulse condition and syndrome; Based on the association vector and the structured semantic data, the probability distribution of each syndrome is calculated through a fuzzy membership function, and a dynamic syndrome-method mapping network is constructed as the dynamic knowledge graph.

4. The intelligent management method for traditional Chinese medicine nursing data according to claim 3 is characterized in that: The constructing of a syndrome evolution model based on the dynamic knowledge graph, and performing syndrome reasoning based on the syndrome evolution model to obtain a current syndrome differentiation conclusion, include: Based on the time-series nursing records in the dynamic knowledge graph, a long short-term memory neural network is used to mine the symptom evolution pattern and generate a syndrome transition probability matrix; According to the syndrome transition probability matrix and the current patient's physical sign information, the syndrome type confidence distribution is output as the current syndrome differentiation conclusion through the Bayesian network reasoning module.

5. The intelligent management method for traditional Chinese medicine nursing data according to claim 4 is characterized in that: The step of calling the reinforcement learning model to generate an initial nursing plan according to the current syndrome differentiation conclusion includes: Based on the syndrome type confidence distribution, the expected intervention effect of each nursing measure is calculated through the reward function in the reinforcement learning model, and a combined intervention strategy of acupoint stimulation and drug combination is generated as the initial nursing plan.

6. The intelligent management method for traditional Chinese medicine nursing data according to claim 5, characterized in that: The conflict detection and intervention effect prediction of the initial nursing plan includes: Based on the historical nursing plan database, a graph convolutional neural network is used to detect taboo relationships between nursing measures and generate an early warning list of nursing conflicts. The initial nursing plan is adjusted according to the early warning list, and the risk-corrected optimized nursing plan is output as the evaluation data.

7. The intelligent management method for traditional Chinese medicine nursing data according to claim 6, characterized in that: The method further comprises: Based on the patient's physiological parameter change curve collected after the implementation of the initial nursing plan, a comparative learning method is used to construct a feature space of symptom response, and a sensitivity index of the nursing measure is generated based on the feature space as the evaluation data.

8. The intelligent management method for traditional Chinese medicine nursing data according to claim 7, characterized in that: Generating nursing decision suggestions using case-based reasoning based on the adaptive strategy feedback model and the dynamic knowledge graph includes: Based on the symptom-measure response relationship represented in the adaptive strategy feedback model, historical cases with similarity higher than a preset value are matched by a case-based reasoning engine to construct a case set of optimal care; Combined with the patient's physical characteristics, a personalized nursing path is generated from the case set through a weighted fusion algorithm as the nursing decision recommendation.

9. The intelligent management method for traditional Chinese medicine nursing data according to claim 8, characterized in that: The visual operation guidance through augmented reality includes: According to the acupoint manipulation steps in the personalized nursing pathway, a three-dimensional meridian navigation pathway is generated through a spatial positioning module of an augmented reality device; Based on the deviation between the real-time hand movement trajectory of the nurse and the standard technique, the display intensity and prompt method of the navigation guidance are dynamically adjusted.

10. An intelligent management system for traditional Chinese medicine nursing data, characterized in that: The system comprises: A data collection module is used to collect TCM nursing data and convert the TCM nursing data into structured semantic data and standardized vital sign data sets; A graph generation module is used to construct a TCM nursing knowledge graph based on the structured semantic data and the standardized physical sign dataset, and to generate a spatiotemporal dynamic knowledge graph by combining fuzzy sets and graph neural networks; An initial plan module is used to construct a syndrome evolution model based on the dynamic knowledge graph, perform syndrome reasoning based on the syndrome evolution model to obtain a current syndrome differentiation conclusion, and call a reinforcement learning model to generate an initial nursing plan based on the current syndrome differentiation conclusion; A model generation module is used to detect conflicts and predict intervention effects of the initial nursing plan, and to form an adaptive strategy feedback model by comparative learning and modeling the relationship between symptoms and measures based on the evaluation data after nursing implementation; A nursing decision-making module is used to generate nursing decision suggestions based on the adaptive strategy feedback model and the dynamic knowledge graph using case-based reasoning, provide visual operation guidance through augmented reality, and complete nursing task allocation and execution with the help of an intelligent scheduling system; The encryption management module is used to implement distributed data processing using federated learning during the execution of the above modules, and combines the differential privacy mechanism to ensure the dynamic desensitization and privacy security of traditional Chinese medicine nursing data.

Images