Deep learning-based traditional Chinese medicine massage scheme generation method and system

Abstract

The invention provides a deep learning-based traditional Chinese medicine massage scheme generation method and system, and the method comprises the steps: calling multi-dimensional diagnosis data of different patients from different medical databases based on a multi-dimensional data calling index, carrying out the processing of the multi-dimensional diagnosis data, and constructing a deep learning sample library; feature representation of deep learning samples in the deep learning sample library is extracted, learning and cycle optimization are carried out on the target network based on the feature representation, and a deep learning model is constructed; analyzing the body characterization data of the patient based on a deep learning model, generating a multi-dimensional massage scheme, and determining a personalized massage scheme of the patient based on a predicted massage effect of the multi-dimensional massage scheme; and receiving feedback data of the patient on the personalized massage scheme based on the management terminal, and dynamically optimizing the personalized massage scheme based on the feedback data. The massage scheme generation efficiency is improved, the matching between the obtained personalized massage scheme and the patient is ensured, and the massage effect on the patient is improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for generating a traditional Chinese medicine massage plan based on deep learning. Background Art

[0002] At present, deep learning, as an important branch of artificial intelligence, has powerful capabilities in data mining and pattern recognition, while Chinese massage, as an important part of traditional Chinese medicine, has significant effects in pain relief and body conditioning.

[0003] In recent years, the application of deep learning in the medical field has gradually increased, providing new possibilities for the intelligent generation of traditional Chinese medicine massage programs;

[0004] However, the formulation of traditional massage plans mostly relies on the physician's clinical experience and the patient's subjective description, lacking objective and accurate data support. In addition, the formulation process of massage plans is cumbersome and difficult to quickly respond to changes in patient needs. Therefore, how to achieve intelligent and personalized generation of massage plans has become a problem that needs to be solved urgently.

[0005] Therefore, in order to overcome the above-mentioned defects, the present invention provides a method and system for generating a Chinese medicine massage plan based on deep learning. Summary of the invention

[0006] The present invention provides a method and system for generating a Chinese medicine massage program based on deep learning, which is used to retrieve multi-dimensional diagnostic data corresponding to different patients from different medical databases, thereby ensuring the comprehensiveness of the acquired data, and processing the obtained multi-dimensional diagnostic data to achieve accurate and reliable construction of a deep learning sample library, thereby providing reliable data support for building a deep learning model. Secondly, the obtained deep learning samples are analyzed to achieve effective construction of a deep learning model according to the analysis results, thereby facilitating the generation of corresponding personalized massage programs for different patients according to the deep learning model, improving the efficiency and effect of massage program generation, simplifying the steps of massage program generation, and also facilitating improving the massage effect on patients. Finally, the personalized massage program is dynamically optimized through the patient's feedback data, thereby ensuring the matching of the ultimately obtained personalized massage program with the patient, thereby improving the massage effect on the patient.

[0007] The present invention provides a method for generating a Chinese medicine massage plan based on deep learning, comprising:

[0008] Step 1: Based on the multi-dimensional data retrieval index, multi-dimensional diagnostic data of different patients are retrieved from different medical databases, and the multi-dimensional diagnostic data are processed to build a deep learning sample library;

[0009] Step 2: Extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model;

[0010] Step 3: Analyze the patient's body representation data based on the deep learning model, generate a multi-dimensional massage plan, and determine the patient's personalized massage plan based on the predicted massage effect of the multi-dimensional massage plan;

[0011] Step 4: Receive the patient's feedback data on the personalized massage plan based on the management terminal, and dynamically optimize the personalized massage plan based on the feedback data.

[0012] Preferably, a method for generating a TCM massage program based on deep learning, in step 1, multi-dimensional diagnostic data of different patients are retrieved from different medical databases based on multi-dimensional data retrieval indicators, including:

[0013] Obtain multi-dimensional data retrieval indicators for the patient, and determine the corresponding medical database to be accessed based on each dimensional data retrieval indicator;

[0014] Extract the data retrieval restriction conditions corresponding to each dimensional data retrieval indicator, and generate a data retrieval request according to each dimensional data retrieval indicator;

[0015] Based on the data retrieval request, a corresponding medical database to be accessed is synchronously accessed, and based on the synchronous access result, data in the corresponding medical database to be accessed is retrieved according to the data retrieval restriction condition to obtain the diagnostic data to be retrieved;

[0016] The diagnostic data to be retrieved of the same patient in different to-be-accessed medical databases are retrieved in a distributed and synchronous manner, and multi-dimensional diagnostic data of different patients are obtained based on the distributed and synchronous retrieval results.

[0017] Preferably, a method for generating a traditional Chinese medicine massage plan based on deep learning, the multi-dimensional diagnostic data includes the patient's physical data, symptom data, medical history data, and constitution type data.

[0018] Preferably, a method for generating a TCM massage program based on deep learning, in step 1, multi-dimensional diagnostic data is processed to construct a deep learning sample library, including:

[0019] Acquire the obtained multi-dimensional diagnostic data and determine the cleaning items for the multi-dimensional diagnostic data;

[0020] Extract the data cleaning event set of each cleaning project and the cleaning index value corresponding to each cleaning event. At the same time, based on the preset cleaning requirements and data cleaning events, respectively, the execution order of the multi-dimensional diagnostic data is respectively determined, and the multi-dimensional diagnostic data is sequentially cleaned according to the cleaning index value based on the execution order to obtain effective multi-dimensional diagnostic data;

[0021] Extract the personalized data architecture of effective multi-dimensional diagnostic data, and reconstruct the data architecture based on the unified coding criteria. At the same time, determine the subordinate relationship between the multi-dimensional diagnostic data based on the medical knowledge graph, and perform multi-node association merging on the reconstructed multi-dimensional diagnostic data based on the subordinate relationship;

[0022] The multi-node association merging results are standardized, and the standardized multi-dimensional diagnostic data are stored to build a deep learning sample library.

[0023] Preferably, a method for generating a TCM massage program based on deep learning stores standardized multi-dimensional diagnostic data, including:

[0024] Perform statistics on the standardized multi-dimensional diagnostic data to determine the byte volume of each dimension of diagnostic data;

[0025] Based on the byte volume, a distributed cache area of ​​equal space is created on the management terminal, and the diagnostic result data of each dimension is stored in the distributed cache area;

[0026] An association and merging record table is constructed based on the storage results, and the multi-node association and merging relationships between the multi-dimensional diagnostic data are synchronously recorded and stored.

[0027] Preferably, a method for generating a TCM massage program based on deep learning, in step 2, extracting feature representations of deep learning samples in a deep learning sample library, and learning and periodically optimizing a target network based on the feature representation to construct a deep learning model, including:

[0028] Acquire the deep learning sample library, and analyze the deep learning samples in the deep learning sample library to determine the massage subject of the deep learning samples;

[0029] Based on the massage subject, the deep learning samples are divided into objects, and the execution data corresponding to each massage subject is extracted based on the object division result;

[0030] Based on the server, a preset massage database is traversed to determine a set of massage techniques, and the execution data is split into action parameters and conditional parameters based on the semantic representation of the execution data;

[0031] Determine the massage action posture for each massage subject based on the semantic representation, split the massage action posture into action steps, and obtain the action amplitude characteristics of each action step based on the action parameters;

[0032] Based on the movement amplitude characteristics of each movement step and the set of massage techniques, the subordinate degree is matched to determine the target massage action corresponding to each massage subject, and the massage intensity and massage time of each target massage action are determined based on the conditional limit parameters;

[0033] At the same time, the body function data corresponding to the deep learning samples are extracted, and the different body function data, target massage movements, massage intensity and massage time of the same massage subject are mapped into categories;

[0034] Based on the category mapping results, the causal relationship between body function data, target massage action, massage intensity and massage time is determined, and the characteristic representation of different massage subjects is obtained based on the causal relationship;

[0035] The target network is selected based on the business execution goal, and the target network is learned and periodically optimized based on feature representation to obtain a deep learning model.

[0036] Preferably, a method for generating a TCM massage program based on deep learning, learning and periodically optimizing a target network based on feature representation, comprises:

[0037] Obtain the obtained feature representation and target network, and retrieve the corresponding model loss function from the server based on the service type of the target network;

[0038] Initialize the parameters of the target network, and perform a preset number of iterative training on the target network based on the feature representation;

[0039] Obtain the model parameters of the target network after each iterative training, and analyze the model parameters based on the model loss function to obtain the model loss value of each iterative training;

[0040] The model loss values ​​obtained by the preset number of iterative training are counted to obtain the corresponding value range distribution characteristics, and the iterative training result corresponding to the minimum model loss value is determined as the deep learning model based on the value range distribution characteristics;

[0041] At the same time, data in different medical databases are searched for novelty based on preset cycles, and when there are differential deep learning samples, the deep learning model is periodically optimized based on the feature representation of the differential deep learning samples.

[0042] Preferably, a method for generating a TCM massage program based on deep learning, in step 3, the patient's body representation data is analyzed based on a deep learning model to generate a multi-dimensional massage program, and the patient's personalized massage program is determined based on the predicted massage effect of the multi-dimensional massage program, including:

[0043] Obtain the patient's current body representation data, and input the body representation data into the deep learning model for processing, to obtain a multi-dimensional massage plan for different parts of the patient and the corresponding predicted massage effect;

[0044] Arrange the multi-dimensional massage plans for the same part in descending order of effect, and determine the massage plan corresponding to the first one as the personalized massage plan corresponding to the patient's current part based on the result of the order arrangement;

[0045] The communication protocol between the patient terminal and the management terminal is obtained, and the personalized massage plan is formatted and transmitted to the patient terminal based on the communication protocol, and the personalized massage plan is visualized on the patient terminal based on the transmission result.

[0046] Preferably, a method for generating a TCM massage program based on deep learning, in step 4, receiving feedback data of the patient on the personalized massage program based on the management terminal, and dynamically optimizing the personalized massage program based on the feedback data, including:

[0047] Obtaining patient feedback data on the personalized massage program, and using the feedback data as the first optimization parameter;

[0048] At the same time, the treatment process of the patient based on the personalized massage plan is monitored in real time, and the massage effect of the patient under the personalized massage plan is obtained based on the real-time monitoring results, and the massage effect is used as the second optimization parameter;

[0049] The optimization nodes and optimization parameters of the personalized massage program are determined based on the first optimization parameter and the second optimization parameter, and the personalized massage program is dynamically optimized based on the optimization nodes and the optimization parameters.

[0050] The present invention provides a system for generating a traditional Chinese medicine massage program based on deep learning, comprising:

[0051] The learning sample library construction module is used to retrieve multi-dimensional diagnostic data of different patients from different medical databases based on multi-dimensional data retrieval indicators, and process the multi-dimensional diagnostic data to build a deep learning sample library;

[0052] The model building module is used to extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model;

[0053] A program determination module is used to analyze the patient's body representation data based on a deep learning model, generate a multi-dimensional massage program, and determine the patient's personalized massage program based on the predicted massage effect of the multi-dimensional massage program;

[0054] The program optimization module is used to receive the patient's feedback data on the personalized massage program based on the management terminal, and dynamically optimize the personalized massage program based on the feedback data.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. By retrieving multi-dimensional diagnostic data corresponding to different patients from different medical databases, the comprehensiveness of the acquired data is ensured, and the obtained multi-dimensional diagnostic data is processed to achieve accurate and reliable construction of the deep learning sample library, providing reliable data support for the construction of the deep learning model. Secondly, the obtained deep learning samples are analyzed to achieve effective construction of the deep learning model according to the analysis results, so as to facilitate the generation of corresponding personalized massage plans for different patients according to the deep learning model, improve the efficiency and effect of massage plan generation, simplify the steps of massage plan generation, and also facilitate improving the massage effect on patients. Finally, the personalized massage plan is dynamically optimized through the patient's feedback data, thereby ensuring the matching of the final personalized massage plan with the patient and improving the massage effect on the patient.

[0057] 2. By parsing the obtained deep learning sample library, the massage subject corresponding to the deep learning sample can be accurately and effectively determined. The deep learning samples are divided into objects according to the massage subject to obtain the execution data corresponding to each massage subject. Secondly, the execution data corresponding to each massage subject is parsed to accurately and effectively determine the movement amplitude characteristics of each massage subject and the massage intensity and massage time of each target massage action, so as to accurately and effectively determine the massage techniques of different massage subjects according to the movement amplitude characteristics, massage intensity and massage time. Finally, the causal relationship between the patient's physical function data, target massage action, massage intensity and massage time is determined, and the feature representation of different massage subjects is locked according to the causal relationship. The selected target network is learned and periodically optimized through the feature representation, so as to accurately and effectively construct the deep learning model, which provides convenience for formulating personalized massage plans for different patients.

[0058] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0059] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0061] Figure 1 It is a flow chart of a method for generating a Chinese medicine massage plan based on deep learning in an embodiment of the present invention;

[0062] Figure 2 This is a flowchart of step 1 in a method for generating a traditional Chinese medicine massage program based on deep learning in an embodiment of the present invention;

[0063] Figure 3 This is a structural diagram of a Chinese medicine massage program generation system based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0065] Embodiment 1:

[0066] This embodiment provides a method for generating a Chinese medicine massage program based on deep learning. Figure 1 As shown, including:

[0067] Step 1: Based on the multi-dimensional data retrieval index, multi-dimensional diagnostic data of different patients are retrieved from different medical databases, and the multi-dimensional diagnostic data are processed to build a deep learning sample library;

[0068] Step 2: Extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model;

[0069] Step 3: Analyze the patient's body representation data based on the deep learning model, generate a multi-dimensional massage plan, and determine the patient's personalized massage plan based on the predicted massage effect of the multi-dimensional massage plan;

[0070] Step 4: Receive the patient's feedback data on the personalized massage plan based on the management terminal, and dynamically optimize the personalized massage plan based on the feedback data.

[0071] In this embodiment, the multi-dimensional retrieval index is used to characterize the type of data that needs to be retrieved, including body data type, disease data type, and constitution data type.

[0072] In this embodiment, the multi-dimensional diagnostic data includes physical data (such as height, weight, body fat percentage, etc.), disease information (such as pain location, pain severity, etc.), medical history data, physical type, etc. The purpose is to obtain comprehensive data of the patient, so as to facilitate the accurate construction of a deep learning model.

[0073] In this embodiment, processing the multi-dimensional diagnostic data refers to cleaning, integrating and standardizing the multi-dimensional diagnostic data to facilitate the construction of a deep learning model.

[0074] In this embodiment, feature representation refers to the correlation between the patient's physical function and the massage technique, massage intensity and massage time represented by the deep learning sample, so as to facilitate targeted training of the target network and realize accurate and reliable construction of the deep learning model.

[0075] In this embodiment, the target network is a model framework selected from a model library according to the analysis business, and is a framework suitable for generating a massage plan.

[0076] In this embodiment, cycle optimization refers to optimizing the target network according to a preset cycle, with the aim of ensuring the real-time effectiveness of the final deep model.

[0077] In this embodiment, the body representation data refers to specific data corresponding to the patient's current physical condition, including pain location and body function data.

[0078] In this embodiment, the multi-dimensional massage plan refers to generating multiple different massage plans for each part, so as to facilitate the selection of the massage plan most suitable for the patient.

[0079] In this embodiment, the predicted massage effect is obtained by predicting the execution effect of the multi-dimensional massage plan through the constructed deep learning model, which is used to characterize the therapeutic effect that can be achieved after different massage plans are implemented on patients.

[0080] In this embodiment, the personalized massage plan refers to the massage plan that is finally obtained and is suitable for the patient.

[0081] In this embodiment, the feedback data is the feedback made by the patient on the implementation effect of the personalized massage plan, including deficiencies and the like.

[0082] The working principle and beneficial effects of the above technical solution are: by retrieving multi-dimensional diagnostic data corresponding to different patients from different medical databases, the comprehensiveness of the acquired data is ensured, and the obtained multi-dimensional diagnostic data is processed to achieve accurate and reliable construction of the deep learning sample library, providing reliable data support for building a deep learning model. Secondly, the obtained deep learning samples are analyzed to achieve effective construction of the deep learning model according to the analysis results, so as to facilitate the generation of corresponding personalized massage plans for different patients according to the deep learning model, improve the efficiency and effect of massage plan generation, and simplify the steps of massage plan generation. Finally, the personalized massage plan is dynamically optimized through the patient's feedback data, thereby ensuring the matching of the final personalized massage plan with the patient and improving the massage effect on the patient.

[0083] Embodiment 2:

[0084] Based on Example 1, this example provides a method for generating a Chinese medicine massage program based on deep learning, such as Figure 2 As shown, in step 1, multi-dimensional diagnostic data of different patients are retrieved from different medical databases based on multi-dimensional data retrieval indicators, including:

[0085] Step 101: obtaining multi-dimensional data retrieval indicators for a patient, and determining a corresponding medical database to be accessed based on each dimensional data retrieval indicator;

[0086] Step 102: extracting the data retrieval limiting condition corresponding to each dimensional data retrieval indicator, and generating a data retrieval request according to each dimensional data retrieval indicator;

[0087] Step 103: synchronously access the corresponding medical database to be accessed based on the data retrieval request, and retrieve the data in the corresponding medical database to be accessed according to the data retrieval restriction condition based on the synchronous access result to obtain the diagnostic data to be retrieved;

[0088] Step 104: Distributed synchronous retrieval is performed on the diagnostic data to be retrieved of the same patient in different to-be-accessed medical databases, and multi-dimensional diagnostic data of different patients is obtained based on the distributed synchronous retrieval results.

[0089] In this embodiment, the medical database to be accessed refers to a database corresponding to each dimensional data retrieval indicator, for example, it may be a database corresponding to medical history.

[0090] In this embodiment, the data retrieval restriction condition refers to the requirement of each dimension data retrieval indicator to restrict the data retrieval process, including restrictions on the time and amount of data retrieval.

[0091] In this embodiment, the diagnostic data to be retrieved refers to the data that meets the retrieval requirements determined after searching the corresponding medical database to be accessed, that is, the data that needs to be retrieved in the end.

[0092] The working principle and beneficial effects of the above technical solution are: by analyzing the multi-dimensional data retrieval indicators, the medical database to be accessed and the data retrieval restriction conditions can be accurately and effectively determined. Finally, the data in the medical database to be accessed is retrieved according to the data retrieval restriction conditions, so as to achieve accurate and effective retrieval of the required data, thereby ensuring the comprehensiveness and reliability of the retrieval of multi-dimensional diagnostic data.

[0093] Embodiment 3:

[0094] Based on Example 2, this example provides a method for generating a Chinese medicine massage plan based on deep learning, and the multi-dimensional diagnostic data includes the patient's physical data, symptom data, medical history data, and physical constitution type data.

[0095] Embodiment 4:

[0096] Based on Example 1, this example provides a method for generating a TCM massage program based on deep learning. In step 1, multi-dimensional diagnostic data is processed to construct a deep learning sample library, including:

[0097] Acquire the obtained multi-dimensional diagnostic data and determine the cleaning items for the multi-dimensional diagnostic data;

[0098] Extract the data cleaning event set of each cleaning project and the cleaning index value corresponding to each cleaning event. At the same time, based on the preset cleaning requirements and data cleaning events, respectively, the execution order of the multi-dimensional diagnostic data is respectively determined, and the multi-dimensional diagnostic data is sequentially cleaned according to the cleaning index value based on the execution order to obtain effective multi-dimensional diagnostic data;

[0099] Extract the personalized data architecture of effective multi-dimensional diagnostic data, and reconstruct the data architecture based on the unified coding criteria. At the same time, determine the subordinate relationship between the multi-dimensional diagnostic data based on the medical knowledge graph, and perform multi-node association merging on the reconstructed multi-dimensional diagnostic data based on the subordinate relationship;

[0100] The multi-node association merging results are standardized, and the standardized multi-dimensional diagnostic data are stored to build a deep learning sample library.

[0101] In this embodiment, the cleaning project refers to the type of data cleaning performed on the multi-dimensional diagnostic data, including removal of duplicate data and outlier processing.

[0102] In this embodiment, the data cleaning event set includes specific cleaning targets and cleaning rules corresponding to each cleaning item.

[0103] In this embodiment, the cleaning index value refers to the specific cleaning degree represented by each cleaning item, that is, the accuracy.

[0104] In this embodiment, the preset cleaning requirements are set in advance, for example, the execution logic for different data cleaning events.

[0105] In this embodiment, the execution order refers to limiting the execution order of different data cleaning events, with the goal of ensuring the cleaning effect of multi-dimensional diagnostic data, for example, deduplicated data can be displayed, followed by abnormal value adjustment processing, etc.

[0106] In this embodiment, the effective multi-dimensional diagnostic data refers to the result obtained after cleaning the multi-dimensional diagnostic data.

[0107] In this embodiment, the personalized data architecture refers to the data structure between the effective multi-dimensional diagnostic data, that is, the composition of the effective multi-dimensional diagnostic data.

[0108] In this embodiment, the unified coding criteria are set in advance and are used to unify the formats of diagnostic data of different dimensions and other operations.

[0109] In this embodiment, multi-node association merging refers to merging and splicing data of the same type or with an association relationship.

[0110] The working principle and beneficial effects of the above technical solution are: by effectively cleaning the obtained multidimensional diagnostic data, the accuracy and effectiveness of the multidimensional diagnostic data finally obtained are ensured; secondly, the cleaned multidimensional diagnostic data are uniformly encoded, and the uniformly encoded multidimensional diagnostic data are subjected to subordinate association analysis according to the medical knowledge graph, so as to realize accurate and effective association and merging of data of the same category and post-association relationship; finally, the associated and merged multidimensional diagnostic data are standardized to realize accurate and reliable construction of a deep learning sample library, which provides reliable data support for building a deep learning model.

[0111] Embodiment 5:

[0112] On the basis of Example 4, this embodiment provides a method for generating a TCM massage plan based on deep learning, which stores standardized multi-dimensional diagnostic data, including:

[0113] Perform statistics on the standardized multi-dimensional diagnostic data to determine the byte volume of each dimension of diagnostic data;

[0114] Based on the byte volume, a distributed cache area of ​​equal space is created on the management terminal, and the diagnostic result data of each dimension is stored in the distributed cache area;

[0115] An association and merging record table is constructed based on the storage results, and the multi-node association and merging relationships between the multi-dimensional diagnostic data are synchronously recorded and stored.

[0116] In this embodiment, the byte amount refers to the specific number corresponding to each dimension of diagnostic data, so as to facilitate allocation of a storage area of ​​corresponding size for each dimension of diagnostic data.

[0117] In this embodiment, the distributed cache area refers to a specific data storage area corresponding to the diagnostic data of each dimension.

[0118] In this embodiment, the association and merging record table is used to record detailed information of the data with association and merging in the distributed cache area.

[0119] The working principle and beneficial effects of the above technical solution are: by statistically analyzing the standardized multi-dimensional diagnostic data, the byte amount of each dimension of diagnostic data can be accurately and effectively determined; secondly, a corresponding distributed cache area is constructed according to the byte amount to achieve effective storage of each dimension of diagnostic data; finally, an associated merge record table is constructed to achieve accurate and effective recording of associated merged data in the distributed cache area.

[0120] Embodiment 6:

[0121] Based on Example 1, this embodiment provides a method for generating a Chinese medicine massage program based on deep learning. In step 2, feature representations of deep learning samples in a deep learning sample library are extracted, and the target network is learned and periodically optimized based on the feature representation to construct a deep learning model, including:

[0122] Acquire the deep learning sample library, and analyze the deep learning samples in the deep learning sample library to determine the massage subject of the deep learning samples;

[0123] Based on the massage subject, the deep learning samples are divided into objects, and the execution data corresponding to each massage subject is extracted based on the object division result;

[0124] Based on the server, a preset massage database is traversed to determine a set of massage techniques, and the execution data is split into action parameters and conditional parameters based on the semantic representation of the execution data;

[0125] Determine the massage action posture for each massage subject based on the semantic representation, split the massage action posture into action steps, and obtain the action amplitude characteristics of each action step based on the action parameters;

[0126] Based on the movement amplitude characteristics of each movement step and the set of massage techniques, the subordinate degree is matched to determine the target massage action corresponding to each massage subject, and the massage intensity and massage time of each target massage action are determined based on the conditional limit parameters;

[0127] At the same time, the body function data corresponding to the deep learning samples are extracted, and the different body function data, target massage movements, massage intensity and massage time of the same massage subject are mapped into categories;

[0128] Based on the category mapping results, the causal relationship between body function data, target massage action, massage intensity and massage time is determined, and the characteristic representation of different massage subjects is obtained based on the causal relationship;

[0129] The target network is selected based on the business execution goal, and the target network is learned and periodically optimized based on feature representation to obtain a deep learning model.

[0130] In this embodiment, the massage subject refers to the massage part represented by the deep learning sample, such as the waist and neck.

[0131] In this embodiment, object segmentation refers to distinguishing the deep learning samples obtained according to the massage subjects, with the aim of determining the specific data corresponding to different massage subjects, thereby facilitating the determination of the characteristic representations of different massage subjects, wherein the execution data is the specific massage data corresponding to each massage subject obtained after object segmentation of the deep learning samples.

[0132] In this embodiment, the preset massage database is set in advance and is used to store a variety of different types of massage techniques.

[0133] In this embodiment, the semantic representation refers to the massage actions reflected in the execution data and the action strength or action amplitude value corresponding to each massage action.

[0134] In this embodiment, the action parameter refers to the massage action parameter contained in the execution data.

[0135] In this embodiment, the conditional limiting parameters refer to parameters such as the force range and massage time of the massage action defined in the execution data.

[0136] In this embodiment, the action step splitting refers to splitting the massage action of each massage subject according to the massage action posture, in order to determine the massage technique used by each massage subject according to the splitting result.

[0137] In this embodiment, the movement amplitude feature refers to the massage amplitude and movement range of each movement step.

[0138] In this embodiment, the target massage action refers to the massage technique used by the current massage subject matched from the massage technique set.

[0139] In this embodiment, the causal relationship refers to analyzing the body function data, target massage action, massage intensity and massage time, determining the fluctuation of the relative change trend among the four, and thus determining the causal relationship according to the fluctuation.

[0140] In this embodiment, the business execution target is known in advance, that is, the type of the current analysis business to be executed and the purpose to be achieved.

[0141] In this embodiment, the target network refers to a network framework suitable for building a deep learning model, such as a neural network.

[0142] The working principle and beneficial effects of the above technical solution are: by parsing the obtained deep learning sample library, the massage subject corresponding to the deep learning sample can be accurately and effectively determined, the deep learning sample is divided into objects according to the massage subject, and the execution data corresponding to each massage subject is obtained. Secondly, the execution data corresponding to each massage subject is parsed to achieve accurate and effective determination of the movement amplitude characteristics of each massage subject and the massage intensity and massage time of each target massage action, so as to achieve accurate and effective determination of the massage techniques of different massage subjects according to the movement amplitude characteristics, massage intensity and massage time. Finally, the causal relationship between the patient's physical function data, target massage action, massage intensity and massage time is determined, and the feature representation of different massage subjects is locked according to the causal relationship, and the selected target network is learned and periodically optimized through the feature representation, so as to achieve accurate and effective construction of the deep learning model, which provides convenience for formulating personalized massage plans for different patients.

[0143] Embodiment 7:

[0144] Based on Example 6, this embodiment provides a method for generating a TCM massage program based on deep learning, which performs learning and periodic optimization on a target network based on feature representation, including:

[0145] Obtain the obtained feature representation and target network, and retrieve the corresponding model loss function from the server based on the service type of the target network;

[0146] Initialize the parameters of the target network, and perform a preset number of iterative training on the target network based on the feature representation;

[0147] Obtain the model parameters of the target network after each iterative training, and analyze the model parameters based on the model loss function to obtain the model loss value of each iterative training;

[0148] The model loss values ​​obtained by the preset number of iterative training are counted to obtain the corresponding value range distribution characteristics, and the iterative training result corresponding to the minimum model loss value is determined as the deep learning model based on the value range distribution characteristics;

[0149] At the same time, data in different medical databases are searched for novelty based on preset cycles, and when there are differential deep learning samples, the deep learning model is periodically optimized based on the feature representation of the differential deep learning samples.

[0150] In this embodiment, the service type refers to the service analysis type that can be performed by the target network and the corresponding configuration parameters.

[0151] In this embodiment, the preset number of times is set in advance.

[0152] In this embodiment, the model loss value is used to characterize the current training effect on the target network. The smaller the model loss value, the better the training effect on the target network.

[0153] In this embodiment, the value range distribution feature refers to the value distribution obtained by statistically analyzing the model loss value obtained after each learning of the target network for a preset number of times.

[0154] In this embodiment, the preset period is set in advance.

[0155] In this embodiment, the difference deep learning sample refers to the emergence of new deep learning samples, that is, new data that is different from the data in the current deep learning sample library.

[0156] The working principle and beneficial effects of the above technical solution are: the target network is learned through the obtained feature representation, and the learning result with the best performance is selected from the learning results as the final deep learning model, thereby ensuring the reliability of the deep learning model finally obtained by the team. At the same time, data is searched and processed according to a preset period, so that the deep learning model can be effectively updated in time when new data appears, thereby ensuring the real-time effectiveness of the deep learning model.

[0157] Embodiment 8:

[0158] Based on Example 1, this example provides a method for generating a TCM massage program based on deep learning. In step 3, the patient's body representation data is analyzed based on a deep learning model to generate a multi-dimensional massage program, and a personalized massage program for the patient is determined based on the predicted massage effect of the multi-dimensional massage program, including:

[0159] Obtain the patient's current body representation data, and input the body representation data into the deep learning model for processing, to obtain a multi-dimensional massage plan for different parts of the patient and the corresponding predicted massage effect;

[0160] Arrange the multi-dimensional massage plans for the same part in descending order of effect, and determine the massage plan corresponding to the first one as the personalized massage plan corresponding to the patient's current part based on the result of the order arrangement;

[0161] The communication protocol between the patient terminal and the management terminal is obtained, and the personalized massage plan is formatted and transmitted to the patient terminal based on the communication protocol, and the personalized massage plan is visualized on the patient terminal based on the transmission result.

[0162] In this embodiment, the physical representation data refers to the patient's current physical condition data.

[0163] In this embodiment, the communication protocol is used to define requirements such as the communication method and communication format between the patient terminal and the management terminal.

[0164] The working principle and beneficial effects of the above technical solution are: by inputting the patient's current body representation data into the deep learning model for processing, a multi-dimensional massage plan for different parts of the patient and the corresponding predicted massage effects are determined; secondly, a personalized massage plan is selected from the multi-dimensional massage plan based on the predicted massage effect to ensure the applicability and reliability of the final personalized massage plan; finally, by obtaining the communication protocol between the patient terminal and the management terminal, the personalized massage plan is formatted and transmitted to the patient terminal according to the communication protocol, and is visualized on the patient terminal, so that the patient can accurately and effectively understand the personalized massage plan.

[0165] Embodiment 9:

[0166] Based on Example 1, this embodiment provides a method for generating a Chinese medicine massage program based on deep learning. In step 4, feedback data of the patient on the personalized massage program is received based on the management terminal, and the personalized massage program is dynamically optimized based on the feedback data, including:

[0167] Obtaining patient feedback data on the personalized massage program, and using the feedback data as the first optimization parameter;

[0168] At the same time, the treatment process of the patient based on the personalized massage plan is monitored in real time, and the massage effect of the patient under the personalized massage plan is obtained based on the real-time monitoring results, and the massage effect is used as the second optimization parameter;

[0169] The optimization nodes and optimization parameters of the personalized massage program are determined based on the first optimization parameter and the second optimization parameter, and the personalized massage program is dynamically optimized based on the optimization nodes and the optimization parameters.

[0170] In this embodiment, the first optimization parameter refers to using the patient's feedback data as a specific basis parameter for optimizing the personalized massage program.

[0171] In this embodiment, the second optimization parameter refers to a specific basis parameter for optimizing the personalized massage plan by taking the massage effect of the patient under the personalized massage plan.

[0172] In this embodiment, the optimization node refers to the specific content of optimizing the personalized massage plan.

[0173] In this embodiment, the optimization parameter refers to the specific degree of optimization of the specific content that needs to be optimized in the personalized massage program.

[0174] The working principle and beneficial effect of the above technical solution are: by using the patient's feedback data and the massage effect as the first optimization parameter and the second optimization parameter respectively, the optimization nodes and optimization parameters of the personalized massage plan can be accurately and effectively determined according to the first optimization parameter and the second optimization parameter, and the personalized massage plan can be accurately and effectively dynamically optimized according to the optimization nodes and the optimization parameters, thereby ensuring the reliability of the final personalized massage plan.

[0175] Embodiment 10:

[0176] This embodiment provides a system for generating a Chinese medicine massage program based on deep learning. Figure 3 As shown, including:

[0177] The learning sample library construction module is used to retrieve multi-dimensional diagnostic data of different patients from different medical databases based on multi-dimensional data retrieval indicators, and process the multi-dimensional diagnostic data to build a deep learning sample library;

[0178] The model building module is used to extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model;

[0179] A program determination module is used to analyze the patient's body representation data based on a deep learning model, generate a multi-dimensional massage program, and determine the patient's personalized massage program based on the predicted massage effect of the multi-dimensional massage program;

[0180] The program optimization module is used to receive the patient's feedback data on the personalized massage program based on the management terminal, and dynamically optimize the personalized massage program based on the feedback data.

[0181] The working principle and beneficial effects of the above technical solution are: by retrieving multi-dimensional diagnostic data corresponding to different patients from different medical databases, the comprehensiveness of the acquired data is ensured, and the obtained multi-dimensional diagnostic data is processed to achieve accurate and reliable construction of the deep learning sample library, providing reliable data support for building a deep learning model. Secondly, the obtained deep learning samples are analyzed to achieve effective construction of the deep learning model according to the analysis results, so as to facilitate the generation of corresponding personalized massage plans for different patients according to the deep learning model, improve the efficiency and effect of massage plan generation, simplify the steps of massage plan generation, and also facilitate improving the massage effect on patients. Finally, the personalized massage plan is dynamically optimized through the patient's feedback data, thereby ensuring the matching of the final personalized massage plan with the patient and improving the massage effect on the patient.

[0182] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for generating a Chinese medicine massage program based on deep learning, characterized in that: include: Step 1: Based on the multi-dimensional data retrieval index, multi-dimensional diagnostic data of different patients are retrieved from different medical databases, and the multi-dimensional diagnostic data are processed to build a deep learning sample library; Step 2: Extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model; Step 3: Analyze the patient's body representation data based on the deep learning model, generate a multi-dimensional massage plan, and determine the patient's personalized massage plan based on the predicted massage effect of the multi-dimensional massage plan; Step 4: Receive the patient's feedback data on the personalized massage plan based on the management terminal, and dynamically optimize the personalized massage plan based on the feedback data.

2. A method for generating a Chinese medicine massage program based on deep learning according to claim 1, characterized in that: In step 1, multi-dimensional diagnostic data of different patients are retrieved from different medical databases based on multi-dimensional data retrieval indicators, including: Obtain multi-dimensional data retrieval indicators for the patient, and determine the corresponding medical database to be accessed based on each dimensional data retrieval indicator; Extract the data retrieval restriction conditions corresponding to each dimensional data retrieval indicator, and generate a data retrieval request according to each dimensional data retrieval indicator; Based on the data retrieval request, a corresponding medical database to be accessed is synchronously accessed, and based on the synchronous access result, data in the corresponding medical database to be accessed is retrieved according to the data retrieval restriction condition to obtain the diagnostic data to be retrieved; The diagnostic data to be retrieved of the same patient in different to-be-accessed medical databases are retrieved in a distributed and synchronous manner, and multi-dimensional diagnostic data of different patients are obtained based on the distributed and synchronous retrieval results.

3. A method for generating a Chinese medicine massage program based on deep learning according to claim 2, characterized in that: Multi-dimensional diagnostic data includes the patient's physical data, symptom data, medical history data, and physical type data.

4. The method for generating a Chinese medicine massage program based on deep learning according to claim 1, characterized in that: In step 1, the multi-dimensional diagnostic data is processed to build a deep learning sample library, including: Acquire the obtained multi-dimensional diagnostic data and determine the cleaning items for the multi-dimensional diagnostic data; Extract the data cleaning event set of each cleaning project and the cleaning index value corresponding to each cleaning event. At the same time, based on the preset cleaning requirements and data cleaning events, respectively, the execution order of the multi-dimensional diagnostic data is respectively determined, and the multi-dimensional diagnostic data is sequentially cleaned according to the cleaning index value based on the execution order to obtain effective multi-dimensional diagnostic data; Extract the personalized data architecture of effective multi-dimensional diagnostic data, and reconstruct the data architecture based on the unified coding criteria. At the same time, determine the subordinate relationship between the multi-dimensional diagnostic data based on the medical knowledge graph, and perform multi-node association merging on the reconstructed multi-dimensional diagnostic data based on the subordinate relationship; The multi-node association merging results are standardized, and the standardized multi-dimensional diagnostic data are stored to build a deep learning sample library.

5. The method for generating a Chinese medicine massage program based on deep learning according to claim 4, characterized in that: Store standardized multi-dimensional diagnostic data, including: Perform statistics on the standardized multi-dimensional diagnostic data to determine the byte volume of each dimension of diagnostic data; Based on the byte volume, a distributed cache area of ​​equal space is created on the management terminal, and the diagnostic result data of each dimension is stored in the distributed cache area; An association and merging record table is constructed based on the storage results, and the multi-node association and merging relationships between the multi-dimensional diagnostic data are synchronously recorded and stored.

6. The method for generating a Chinese medicine massage program based on deep learning according to claim 1, characterized in that: In step 2, feature representations of deep learning samples in the deep learning sample library are extracted, and the target network is learned and periodically optimized based on the feature representations to build a deep learning model, including: Acquire the deep learning sample library, and analyze the deep learning samples in the deep learning sample library to determine the massage subject of the deep learning samples; Based on the massage subject, the deep learning samples are divided into objects, and the execution data corresponding to each massage subject is extracted based on the object division result; Based on the server, a preset massage database is traversed to determine a set of massage techniques, and the execution data is split into action parameters and conditional parameters based on the semantic representation of the execution data; Determine the massage action posture for each massage subject based on the semantic representation, split the massage action posture into action steps, and obtain the action amplitude characteristics of each action step based on the action parameters; Based on the movement amplitude characteristics of each movement step and the set of massage techniques, the subordinate degree is matched to determine the target massage action corresponding to each massage subject, and the massage intensity and massage time of each target massage action are determined based on the conditional limit parameters; At the same time, the body function data corresponding to the deep learning samples are extracted, and the different body function data, target massage movements, massage intensity and massage time of the same massage subject are mapped into categories; Based on the category mapping results, the causal relationship between body function data, target massage action, massage intensity and massage time is determined, and the characteristic representation of different massage subjects is obtained based on the causal relationship; The target network is selected based on the business execution goal, and the target network is learned and periodically optimized based on feature representation to obtain a deep learning model.

7. A method for generating a Chinese medicine massage program based on deep learning according to claim 6, characterized in that: Learning and periodic optimization of the target network based on feature representation, including: Obtain the obtained feature representation and target network, and retrieve the corresponding model loss function from the server based on the service type of the target network; Initialize the parameters of the target network, and perform a preset number of iterative training on the target network based on the feature representation; Obtain the model parameters of the target network after each iterative training, and analyze the model parameters based on the model loss function to obtain the model loss value of each iterative training; The model loss values ​​obtained by the preset number of iterative training are counted to obtain the corresponding value range distribution characteristics, and the iterative training result corresponding to the minimum model loss value is determined as the deep learning model based on the value range distribution characteristics; At the same time, data in different medical databases are searched for novelty based on preset cycles, and when there are differential deep learning samples, the deep learning model is periodically optimized based on the feature representation of the differential deep learning samples.

8. The method for generating a Chinese medicine massage program based on deep learning according to claim 1, characterized in that: In step 3, the patient's body representation data is analyzed based on the deep learning model to generate a multi-dimensional massage plan, and the patient's personalized massage plan is determined based on the predicted massage effect of the multi-dimensional massage plan, including: Obtain the patient's current body representation data, and input the body representation data into the deep learning model for processing, to obtain a multi-dimensional massage plan for different parts of the patient and the corresponding predicted massage effect; Arrange the multi-dimensional massage plans for the same part in descending order of effect, and determine the massage plan corresponding to the first one as the personalized massage plan corresponding to the patient's current part based on the result of the order arrangement; The communication protocol between the patient terminal and the management terminal is obtained, and the personalized massage plan is formatted and transmitted to the patient terminal based on the communication protocol, and the personalized massage plan is visualized on the patient terminal based on the transmission result.

9. The method for generating a Chinese medicine massage program based on deep learning according to claim 1, characterized in that: In step 4, the patient's feedback data on the personalized massage program is received based on the management terminal, and the personalized massage program is dynamically optimized based on the feedback data, including: Obtaining patient feedback data on the personalized massage program, and using the feedback data as the first optimization parameter; At the same time, the treatment process of the patient based on the personalized massage plan is monitored in real time, and the massage effect of the patient under the personalized massage plan is obtained based on the real-time monitoring results, and the massage effect is used as the second optimization parameter; The optimization nodes and optimization parameters of the personalized massage program are determined based on the first optimization parameter and the second optimization parameter, and the personalized massage program is dynamically optimized based on the optimization nodes and the optimization parameters.

10. A Chinese medicine massage program generation system based on deep learning, characterized in that: include: The learning sample library construction module is used to retrieve multi-dimensional diagnostic data of different patients from different medical databases based on multi-dimensional data retrieval indicators, and process the multi-dimensional diagnostic data to build a deep learning sample library; The model building module is used to extract the feature representation of deep learning samples in the deep learning sample library, and learn and periodically optimize the target network based on the feature representation to build a deep learning model; A program determination module is used to analyze the patient's body representation data based on a deep learning model, generate a multi-dimensional massage program, and determine the patient's personalized massage program based on the predicted massage effect of the multi-dimensional massage program; The program optimization module is used to receive the patient's feedback data on the personalized massage program based on the management terminal, and dynamically optimize the personalized massage program based on the feedback data.

Images