An intelligent analysis method for trauma orthopedic nursing data

Abstract

The present invention discloses an intelligent analysis method for trauma orthopedic nursing data, and relates to the field of medical data analysis. The intelligent analysis system for trauma orthopedic nursing data constructed by the method includes: a program management module, a data acquisition module, a data processing module, a nursing stage analysis module, a nursing data analysis module, and a nursing program optimization module. The present invention constructs a complete trauma orthopedic nursing program library, covering comprehensive nursing programs for different types of trauma, providing a rich data basis for subsequent analysis, and improving the comprehensiveness and applicability of nursing programs. Through the nursing stage analysis model, the patient's trauma type and nursing stage can be corrected and confirmed, which significantly improves the accuracy of the analysis.

Description

Technical Field

[0001] The present invention relates to the field of medical data analysis, and in particular to an intelligent analysis method for trauma orthopedic nursing data. Background Art

[0002] Orthopedic trauma care is an important part of the medical field, involving the diagnosis, treatment and rehabilitation of patients who have suffered various traumas (such as fractures, dislocations, soft tissue injuries, etc.). Traditional orthopedic trauma care mainly relies on the experience of doctors and clinical guidelines, but this approach has some limitations. First, there are differences in the experience and knowledge level of doctors, which may lead to subjectivity and inconsistency in the selection and implementation of care plans, thus affecting the treatment effect. Secondly, clinical guidelines are usually developed based on the results of group studies, which are difficult to fully meet the specific needs of individual patients. Each patient's injuries, physical conditions, complications, etc. are different, requiring personalized care plans.

[0003] With the rapid development of information technology, a large amount of trauma orthopedic nursing data has been accumulated, which contains rich medical information. How to effectively use this data to improve the quality and efficiency of trauma orthopedic nursing has become an urgent problem to be solved. In recent years, artificial intelligence technology, especially deep learning, has been increasingly widely used in the medical field, providing new possibilities for intelligent trauma orthopedic nursing. Deep learning models can learn complex patterns and rules from a large amount of medical data and are used to assist doctors in diagnosis, treatment planning and prognosis prediction. However, most of the existing deep learning-based trauma orthopedic nursing methods only focus on a single stage or a single data type, lack a comprehensive consideration of the entire patient care process, and fail to fully utilize the complementarity of different types of data. In addition, the existing methods do not adequately verify the reliability of the model output results, lack a mechanism to correct and confirm the model prediction results, and have certain risks. At the same time, the model has poor interpretability and is difficult to gain the trust and acceptance of clinicians.

[0004] Therefore, there is an urgent need for an intelligent analysis method for trauma orthopedic nursing data that can comprehensively utilize various types of patient data to provide personalized and comprehensive nursing plans with high accuracy and reliability to improve the quality and efficiency of trauma orthopedic nursing. Summary of the invention

[0005] A method for intelligent analysis of trauma orthopedic nursing data, comprising the following steps:

[0006] Step S1: constructing a trauma orthopedic nursing program library, in which trauma orthopedic nursing programs corresponding to different trauma types are stored. The trauma orthopedic nursing program consists of five major stages, including acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations;

[0007] Step S2: Obtaining nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient nursing, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, wherein the laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information;

[0008] Step S3: Processing the nursing data of trauma orthopedic patients, the processing operations include data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtaining characteristic parameters of the nursing data of trauma orthopedic patients, and constructing the characteristic parameters into patient nursing samples; the characteristic parameters are stored in matrices, including two matrices corresponding to Class 1 nursing data and Class 2 nursing data;

[0009] Step S4: the corresponding matrix of the first type of nursing data in the patient nursing sample is sent to the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; the accuracy of the nursing stage analysis model output result is obtained, if the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and the patient nursing stage in the first type of nursing data, whether there is supplementary information in the first type of nursing data is detected; if the supplementary information is empty, the accuracy of the model output result is compared with the first confirmation threshold, if it is greater than the first confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient, otherwise the supplementary information is obtained, and after reconstructing the patient nursing sample, the sample is sent to the nursing stage analysis model for calculation again; if the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold, if it is greater than the second confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient, otherwise the patient nursing sample is reconstructed, and the sample is sent to the nursing stage analysis model for calculation again, and when it is calculated again, the corrected confirmation threshold is used for detection; the reconstruction operation of the patient nursing sample includes weight adjustment;

[0010] Step S5: Reconstruct the patient care samples according to the correct trauma type and care stage of the patient, and send the reconstructed patient care samples to the nursing data analysis model for calculation, so as to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic care plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted through the group state to improve the accuracy of the model calculation and the training efficiency of the model;

[0011] Step S6: According to the numbers of the subsequent three sub-stages and the sub-stages that have been implemented in the trauma orthopedic nursing process, match the trauma orthopedic nursing plans in the trauma orthopedic nursing plan library, and use the plan with the highest matching degree as the trauma orthopedic nursing plan required by the patient; if there are multiple plans with the highest matching degree, perform similarity calculation based on the patient's Class 1 nursing data and the standard patient data corresponding to the matched plan, and use the plan with the highest score as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

[0012] As a preferred technical solution of the present invention, the trauma orthopedic nursing solution includes:

[0013] The trauma orthopedic nursing plan is stored in the memory in the form of plan ID + trauma type mark + major stage mark + sub-stage mark code. The sub-stage mark code is compared with the nursing operation code table in the trauma orthopedic nursing plan library.

[0014] As a preferred technical solution of the present invention, the processing operation includes:

[0015] Data cleaning: clean the acquired trauma orthopedic patient care data, identify and remove duplicate records, incorrect entries and irrelevant data;

[0016] Feature labeling: According to the characteristics of the data, feature labeling is performed on the nursing data of trauma orthopedic patients. First, feature division is performed. For numerical data, feature labels are divided according to numerical intervals. For text data, description labels are set as feature labels for feature division. Each sub-item data in the nursing data of trauma orthopedic patients is replaced with feature labels. The final feature label is the feature parameter.

[0017] Weight construction reserves space for weight calculation for the characteristic parameters corresponding to Class 1 nursing data, and the initial weight of each sub-item data is set to 1;

[0018] Outlier detection uses the interquartile range (IQR) to identify and mark outliers of characteristic parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed.

[0019] Empty value marking, marking the control in the feature parameter, the marked empty value will not participate in the subsequent calculation, that is, the empty value part is ignored.

[0020] As a preferred technical solution of the present invention, reconstructing the patient care sample includes:

[0021] The reconstruction of the patient care sample includes the addition of feature parameters corresponding to the supplementary information and the adjustment of weights. The initial patient care sample does not contain supplementary information. If the accuracy of the output results of the analysis model in the care stage is not high enough, it is necessary to obtain supplementary information and add feature parameters corresponding to the supplementary information to the patient care sample;

[0022] To adjust the weight, through the trauma category judgment table, obtain the judgment data item corresponding to the modified trauma type output by the current model, multiply the weight of the characteristic parameter corresponding to the obtained judgment data item in the patient care sample by the adjustment coefficient, and the calculation result is the weight adjustment; the same method is used to adjust the weight for the nursing stage.

[0023] As a preferred technical solution of the present invention, the nursing data analysis model includes:

[0024] The input layer receives the reconstructed patient care samples, including category 1 and category 2 care data;

[0025] Convolutional layers, multiple convolutional layers are used to extract local features and patterns in patient care data;

[0026] Pooling layer, used to reduce feature dimensions and computational complexity and improve the robustness of the model;

[0027] The fully connected layer is used to integrate the features extracted by the convolutional layer and the pooling layer and map them to the numbers of the subsequent three sub-stages;

[0028] The output layer outputs the numbers of the three subsequent sub-stages of the trauma orthopedic care plan required by the patient;

[0029] The group trend optimization module embeds the group trend factor into the model to adjust the model hyperparameters;

[0030] Activation function: Softmax activation function is used in convolutional layers, fully connected layers, and output layers to introduce nonlinear relationships and improve the expressiveness of the model;

[0031] The construction of nursing data analysis model includes:

[0032] Obtain several patient care samples and corresponding trauma orthopedic care plans, and use the subsequent three sub-stages of the trauma orthopedic care plan corresponding to the patient's current care stage as sample labels; organize the samples into a stage prediction data set, and divide the stage prediction data set into a training set, a validation set, and a test set;

[0033] The initial nursing data analysis model is trained using the training set and validation set in the stage prediction data set, and the hyperparameters of the initial stage prediction model are adjusted by the group trend optimization algorithm to obtain the optimized nursing data analysis model;

[0034] The stage prediction test set is used to test the accuracy of the optimized stage prediction model. If the accuracy is within the set range, the optimized stage prediction model is used as the stage prediction model. Otherwise, the steps of adjusting the hyperparameters of the initial stage prediction model are repeated until the accuracy is less than the set range.

[0035] As a preferred technical solution of the present invention, the group trend factor optimization algorithm includes:

[0036] Initialize the group, set the size of the particle group, the parameters corresponding to each particle represent the variables related to the stage prediction, and randomly initialize the parameter combination of the stage prediction model;

[0037] Calculate the fitness of the particle, which is based on the prediction error of the particle;

[0038] Get the reference direction vector of each particle: ; In the formula represents the historical optimal position of the group, represents the current position of the particle, and the reference direction vector represents the direction between the current position of the particle and the historical optimal solution of the group;

[0039] Calculate the cosine value of the angle between the particle's current velocity vector and the reference direction vector; count the number of particles whose cosine value of the angle is greater than the set threshold, and use the ratio as the group trend factor STF; set the limit value of the group trend factor, which is used to distinguish the role of the group trend factor. Specifically, when the group trend factor is greater than the limit value, the group is in a local development state and the inertia weight is reduced. When the group trend factor is less than the limit value, the group is in a global exploration state and the inertia weight is increased.

[0040] The inertia weight is dynamically adjusted according to the group trend factor, and the inertia weight is used to calculate the speed and position of the particle; for each particle, the current fitness of the particle is compared with the fitness of the historical optimal solution. If the current fitness is larger, it is updated to the optimal solution. The above process is iterated repeatedly to obtain the optimal parameter combination of the model and the prediction model of the optimization stage.

[0041] As a preferred technical solution of the present invention, similarity calculation includes:

[0042] The similarity calculation is performed through cosine similarity. The patient category 1 nursing data and standard patient data are converted into character strings. The similarity of the data is obtained by calculating the cosine similarity of the two character strings. A high cosine similarity means that the data are close.

[0043] An intelligent analysis system for trauma orthopedic nursing data, including the following modules:

[0044] A program management module is used to manage a trauma orthopedic nursing program library, which stores trauma orthopedic nursing programs corresponding to different trauma types. The trauma orthopedic nursing program consists of five major stages, including the acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage, and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations;

[0045] The data collection module is used to obtain nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient care, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, among which laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information;

[0046] A data processing module is used to process the nursing data of trauma orthopedic patients. The processing operations include data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtain the feature parameters of the nursing data of trauma orthopedic patients, and construct the feature parameters into patient nursing samples; the feature parameters are stored in matrices, including two matrices corresponding to Class 1 nursing data and Class 2 nursing data;

[0047] The nursing stage analysis module includes a nursing stage analysis unit and a nursing result correction unit. The nursing stage analysis unit is used to send the corresponding matrix of Class 1 nursing data in the patient nursing sample into the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; the nursing result correction unit is used to determine whether to correct the output result according to the accuracy of the output result of the nursing stage analysis model. If the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and patient nursing stage in Class 1 nursing data, it is detected whether there is supplementary information in Class 1 nursing data; if the supplementary information is empty, the accuracy of the model output result is compared with the first confirmation threshold. Comparison is performed. If the accuracy is greater than the first confirmation threshold, the output result of the current model is the correct type of injury and nursing stage of the patient. Otherwise, supplementary information is obtained, and after the patient nursing sample is reconstructed, the sample is sent to the nursing stage analysis model for recalculation. If the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold. If the accuracy is greater than the second confirmation threshold, the output result of the current model is the correct type of injury and nursing stage of the patient. Otherwise, the patient nursing sample is reconstructed, and the sample is sent to the nursing stage analysis model for recalculation. When the calculation is performed again, the modified confirmation threshold is used for detection. The reconstruction operation of the patient nursing sample includes weight adjustment.

[0048] The nursing data analysis module is used to reconstruct the patient nursing samples according to the correct trauma type and nursing stage of the patient, and send the reconstructed patient nursing samples to the nursing data analysis model for calculation to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic nursing plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted according to the group state to improve the accuracy of the model calculation and the training efficiency of the model;

[0049] The nursing plan optimization module is used to match the trauma orthopedic nursing plans in the trauma orthopedic nursing plan library according to the numbers of the subsequent three sub-stages and the sub-stages that have been implemented in the trauma orthopedic nursing process, and use the plan with the highest matching degree as the trauma orthopedic nursing plan required by the patient; if there are multiple plans with the highest matching degree, the similarity is calculated based on the patient's Class 1 nursing data and the standard patient data corresponding to the matched plan, and the plan with the highest score is used as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

[0050] The present invention has the following advantages:

[0051] The present invention constructs a complete trauma orthopedic nursing program library, covering comprehensive nursing programs for different types of trauma, providing a rich data basis for subsequent analysis, and improving the comprehensiveness and applicability of the nursing program; through the nursing stage analysis model, the patient's trauma type and nursing stage can be accurately identified and corrected to ensure the accuracy and effectiveness of the nursing program.

[0052] The present invention can flexibly handle different situations through multiple threshold comparisons and sample reconstruction mechanisms, thereby further ensuring the reliability of the analysis results.

[0053] The present invention constructs a nursing data analysis model using a deep convolutional neural network and introduces group trend factors for optimization, thereby effectively improving the accuracy and training efficiency of the model; through matching and similarity calculations, the most suitable personalized nursing plan can be recommended to patients, significantly improving the pertinence and effectiveness of the nursing plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 The present invention is a schematic diagram of the structure of an intelligent analysis system for trauma orthopedic nursing data adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0056] Embodiment 1, a method for intelligent analysis of trauma orthopedic nursing data, comprising the following steps:

[0057] Step S1: constructing a trauma orthopedic nursing program library, in which trauma orthopedic nursing programs corresponding to different trauma types are stored. The trauma orthopedic nursing program consists of five major stages, including acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations;

[0058] The Orthopaedic Trauma Care Program includes:

[0059] The trauma orthopedic nursing plan is stored in the memory in the form of plan ID + trauma type mark + major stage mark + sub-stage mark code. The sub-stage mark code is compared with the nursing operation code table in the trauma orthopedic nursing plan library.

[0060] Step S2: Obtaining nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient nursing, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, wherein the laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information;

[0061] Step S3: Processing the nursing data of trauma orthopedic patients, the processing operations include data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtaining characteristic parameters of the nursing data of trauma orthopedic patients, and constructing the characteristic parameters into patient nursing samples; the characteristic parameters are stored in matrices, including two matrices corresponding to Class 1 nursing data and Class 2 nursing data;

[0062] Processing operations include:

[0063] Data cleaning: clean the acquired trauma orthopedic patient care data, identify and remove duplicate records, incorrect entries and irrelevant data;

[0064] Feature labeling: According to the characteristics of the data, feature labeling is performed on the nursing data of trauma orthopedic patients. First, feature division is performed. For numerical data, feature labels are divided according to numerical intervals. For text data, description labels are set as feature labels for feature division. Each sub-item data in the nursing data of trauma orthopedic patients is replaced with feature labels. The final feature label is the feature parameter.

[0065] Weight construction reserves space for weight calculation for the characteristic parameters corresponding to Class 1 nursing data, and the initial weight of each sub-item data is set to 1;

[0066] Outlier detection uses the interquartile range (IQR) to identify and mark outliers of characteristic parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed.

[0067] Empty value marking, marking the control in the feature parameter, the marked empty value will not participate in the subsequent calculation, that is, the empty value part is ignored.

[0068] Step S4: the corresponding matrix of the first type of nursing data in the patient nursing sample is sent to the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; the accuracy of the nursing stage analysis model output result is obtained, if the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and the patient nursing stage in the first type of nursing data, whether there is supplementary information in the first type of nursing data is detected; if the supplementary information is empty, the accuracy of the model output result is compared with the first confirmation threshold, if it is greater than the first confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient, otherwise the supplementary information is obtained, and after reconstructing the patient nursing sample, the sample is sent to the nursing stage analysis model for calculation again; if the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold, if it is greater than the second confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient, otherwise the patient nursing sample is reconstructed, and the sample is sent to the nursing stage analysis model for calculation again, and when it is calculated again, the corrected confirmation threshold is used for detection; the reconstruction operation of the patient nursing sample includes weight adjustment;

[0069] Sample restructured patient care includes:

[0070] The reconstruction of the patient care sample includes the addition of feature parameters corresponding to the supplementary information and the adjustment of weights. The initial patient care sample does not contain supplementary information. If the accuracy of the output results of the analysis model in the care stage is not high enough, it is necessary to obtain supplementary information and add feature parameters corresponding to the supplementary information to the patient care sample;

[0071] To adjust the weight, through the trauma category judgment table, obtain the judgment data item corresponding to the modified trauma type output by the current model, multiply the weight of the characteristic parameter corresponding to the obtained judgment data item in the patient care sample by the adjustment coefficient, and the calculation result is the weight adjustment; the same method is used to adjust the weight for the nursing stage.

[0072] The care phase analysis model includes:

[0073] The data preprocessing layer is used to preprocess the input Class 1 nursing data and convert the data into a format acceptable to the model;

[0074] The stage classification layer is used to input data to classify the patient's injury type and care stage, which is composed of a pre-trained classification model;

[0075] The result output layer is used to obtain the classification results of trauma types and nursing stages and the accuracy of model output, and output them.

[0076] Step S5: Reconstruct the patient care samples according to the correct trauma type and care stage of the patient, and send the reconstructed patient care samples to the nursing data analysis model for calculation, so as to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic care plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted through the group state to improve the accuracy of the model calculation and the training efficiency of the model;

[0077] Nursing data analysis models include:

[0078] The input layer receives the reconstructed patient care samples, including category 1 and category 2 care data;

[0079] Convolutional layers, multiple convolutional layers are used to extract local features and patterns in patient care data;

[0080] Pooling layer, used to reduce feature dimensions and computational complexity and improve the robustness of the model;

[0081] The fully connected layer is used to integrate the features extracted by the convolutional layer and the pooling layer and map them to the numbers of the subsequent three sub-stages;

[0082] The output layer outputs the numbers of the three subsequent sub-stages of the trauma orthopedic care plan required by the patient;

[0083] The group trend optimization module embeds the group trend factor into the model to adjust the model hyperparameters;

[0084] Activation function: Softmax activation function is used in convolutional layers, fully connected layers, and output layers to introduce nonlinear relationships and improve the expressiveness of the model;

[0085] The construction of nursing data analysis model includes:

[0086] Obtain several patient care samples and corresponding trauma orthopedic care plans, and use the subsequent three sub-stages of the trauma orthopedic care plan corresponding to the patient's current care stage as sample labels; organize the samples into a stage prediction data set, and divide the stage prediction data set into a training set, a validation set, and a test set;

[0087] The initial nursing data analysis model is trained using the training set and validation set in the stage prediction data set, and the hyperparameters of the initial stage prediction model are adjusted by the group trend optimization algorithm to obtain the optimized nursing data analysis model;

[0088] The stage prediction test set is used to test the accuracy of the optimized stage prediction model. If the accuracy is within the set range, the optimized stage prediction model is used as the stage prediction model. Otherwise, the steps of adjusting the hyperparameters of the initial stage prediction model are repeated until the accuracy is less than the set range.

[0089] The group trend factor optimization algorithm includes:

[0090] Initialize the group, set the size of the particle group, the parameters corresponding to each particle represent the variables related to the stage prediction, and randomly initialize the parameter combination of the stage prediction model;

[0091] Calculate the fitness of the particle, which is based on the prediction error of the particle;

[0092] Get the reference direction vector of each particle: ; In the formula represents the historical optimal position of the group, represents the current position of the particle, and the reference direction vector represents the direction between the current position of the particle and the historical optimal solution of the group;

[0093] Calculate the cosine value of the angle between the particle's current velocity vector and the reference direction vector; count the number of particles whose cosine value of the angle is greater than the set threshold, and use the ratio as the group trend factor STF; set the limit value of the group trend factor, which is used to distinguish the role of the group trend factor. Specifically, when the group trend factor is greater than the limit value, the group is in a local development state and the inertia weight is reduced. When the group trend factor is less than the limit value, the group is in a global exploration state and the inertia weight is increased.

[0094] The inertia weight is dynamically adjusted according to the group trend factor, and the inertia weight is used to calculate the speed and position of the particle; for each particle, the current fitness of the particle is compared with the fitness of the historical optimal solution. If the current fitness is larger, it is updated to the optimal solution. The above process is iterated repeatedly to obtain the optimal parameter combination of the model and the prediction model of the optimization stage.

[0095] Step S6: According to the numbers of the subsequent three sub-stages and the sub-stages that have been implemented in the trauma orthopedic nursing process, match the trauma orthopedic nursing plans in the trauma orthopedic nursing plan library, and use the plan with the highest matching degree as the trauma orthopedic nursing plan required by the patient; if there are multiple plans with the highest matching degree, perform similarity calculation based on the patient's Class 1 nursing data and the standard patient data corresponding to the matched plan, and use the plan with the highest score as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

[0096] Similarity calculation includes:

[0097] The similarity calculation is performed through cosine similarity. The patient category 1 nursing data and standard patient data are converted into character strings. The similarity of the data is obtained by calculating the cosine similarity of the two character strings. A high cosine similarity means that the data are close.

[0098] Example 2, a trauma orthopedic nursing data intelligent analysis system, see Figure 1 As shown, it includes the following modules:

[0099] A program management module is used to manage a trauma orthopedic nursing program library, which stores trauma orthopedic nursing programs corresponding to different trauma types. The trauma orthopedic nursing program consists of five major stages, including the acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage, and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations;

[0100] The data collection module is used to obtain nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient care, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, among which laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information;

[0101] A data processing module is used to process the nursing data of trauma orthopedic patients. The processing operations include data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtain the feature parameters of the nursing data of trauma orthopedic patients, and construct the feature parameters into patient nursing samples; the feature parameters are stored in matrices, including two matrices corresponding to Class 1 nursing data and Class 2 nursing data;

[0102] The nursing stage analysis module includes a nursing stage analysis unit and a nursing result correction unit. The nursing stage analysis unit is used to send the corresponding matrix of Class 1 nursing data in the patient nursing sample into the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; the nursing result correction unit is used to determine whether to correct the output result according to the accuracy of the output result of the nursing stage analysis model. If the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and patient nursing stage in Class 1 nursing data, it is detected whether there is supplementary information in Class 1 nursing data; if the supplementary information is empty, the accuracy of the model output result is compared with the first confirmation threshold. Comparison is performed. If the accuracy is greater than the first confirmation threshold, the output result of the current model is the correct type of injury and nursing stage of the patient. Otherwise, supplementary information is obtained, and after the patient nursing sample is reconstructed, the sample is sent to the nursing stage analysis model for recalculation. If the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold. If the accuracy is greater than the second confirmation threshold, the output result of the current model is the correct type of injury and nursing stage of the patient. Otherwise, the patient nursing sample is reconstructed, and the sample is sent to the nursing stage analysis model for recalculation. When the calculation is performed again, the modified confirmation threshold is used for detection. The reconstruction operation of the patient nursing sample includes weight adjustment.

[0103] The nursing data analysis module is used to reconstruct the patient nursing samples according to the correct trauma type and nursing stage of the patient, and send the reconstructed patient nursing samples to the nursing data analysis model for calculation to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic nursing plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted according to the group state to improve the accuracy of the model calculation and the training efficiency of the model;

[0104] The nursing plan optimization module is used to match the trauma orthopedic nursing plans in the trauma orthopedic nursing plan library according to the numbers of the subsequent three sub-stages and the sub-stages that have been implemented in the trauma orthopedic nursing process, and use the plan with the highest matching degree as the trauma orthopedic nursing plan required by the patient; if there are multiple plans with the highest matching degree, the similarity is calculated based on the patient's Class 1 nursing data and the standard patient data corresponding to the matched plan, and the plan with the highest score is used as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

[0105] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for intelligent analysis of trauma orthopedic nursing data, characterized in that: The following steps are involved: Step S1: constructing a trauma orthopedic nursing program library, in which trauma orthopedic nursing programs corresponding to different trauma types are stored. The trauma orthopedic nursing program consists of five major stages, including acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations; Step S2: Obtaining nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient nursing, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, wherein the laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information; Step S3: Processing the nursing data of trauma orthopedic patients, including data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtaining feature parameters of the nursing data of trauma orthopedic patients, and constructing the feature parameters into patient care samples; The characteristic parameters are stored in matrices, including two matrices corresponding to category 1 nursing data and category 2 nursing data; Step S4: The corresponding matrix of the nursing data of category 1 in the patient nursing sample is sent to the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; Obtain the accuracy of the output result of the nursing stage analysis model. If the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and patient nursing stage in the first type of nursing data, check whether there is supplementary information in the first type of nursing data. If the supplementary information is empty, compare the accuracy of the model output result with the first confirmation threshold. If it is greater than the first confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient. Otherwise, obtain the supplementary information, reconstruct the patient nursing sample, and then send the sample to the nursing stage analysis model for recalculation. If the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold. If it is greater than the second confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient. Otherwise, the patient care sample is reconstructed and the sample is sent to the nursing stage analysis model for recalculation. When the calculation is performed again, the modified confirmation threshold is used for detection; the reconstruction operation of the patient care sample includes weight adjustment; Step S5: Reconstruct the patient care samples according to the correct trauma type and care stage of the patient, and send the reconstructed patient care samples to the nursing data analysis model for calculation, so as to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic care plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted through the group state to improve the accuracy of the model calculation and the training efficiency of the model; Step S6: According to the numbers of the subsequent three sub-stages and the sub-stages of the trauma orthopedic nursing process that have been implemented, the trauma orthopedic nursing plans are matched in the trauma orthopedic nursing plan library, and the plan with the highest matching degree is used as the trauma orthopedic nursing plan required by the patient; If there are multiple plans with the highest matching degree, similarity calculation is performed based on the patient's Category 1 nursing data and the standard patient data corresponding to the matching plan, and the plan with the highest score is used as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

2. The method for intelligent analysis of trauma orthopedic nursing data according to claim 1, characterized in that: The Orthopaedic Trauma Care Program includes: The trauma orthopedic nursing plan is stored in the memory in the form of plan ID + trauma type mark + major stage mark + sub-stage mark code. The sub-stage mark code is compared with the nursing operation code table in the trauma orthopedic nursing plan library.

3. The method for intelligent analysis of trauma orthopedic nursing data according to claim 1, characterized in that: Processing operations include: Data cleaning: clean the acquired trauma orthopedic patient care data, identify and remove duplicate records, incorrect entries and irrelevant data; Feature labeling: According to the characteristics of the data, feature labeling is performed on the nursing data of trauma orthopedic patients. First, feature division is performed. For numerical data, feature labels are divided according to numerical intervals. For text data, description labels are set as feature labels for feature division. Each sub-item data in the nursing data of trauma orthopedic patients is replaced with feature labels. The final feature label is the feature parameter. Weight construction reserves space for weight calculation for the characteristic parameters corresponding to Class 1 nursing data, and the initial weight of each sub-item data is set to 1; Outlier detection uses the interquartile range (IQR) to identify and mark outliers of characteristic parameters, confirms outliers, and corrects or removes outliers based on the confirmation results. The confirmation process includes classifying outliers according to the divided interquartile range intervals, and the classification results are absolute anomalies and relative anomalies. If the classification result is relative anomalies, the outliers are corrected by taking the average value of the data obtained at the previous moment and the next moment, otherwise, the outliers are removed. Empty value marking, marking the control in the feature parameter, the marked empty value will not participate in the subsequent calculation, that is, the empty value part is ignored.

4. The method for intelligent analysis of trauma orthopedic nursing data according to claim 1, characterized in that: Sample restructured patient care includes: The reconstruction of the patient care sample includes the addition of feature parameters corresponding to the supplementary information and the adjustment of weights. The initial patient care sample does not contain supplementary information. If the accuracy of the output results of the analysis model in the care stage is not high enough, it is necessary to obtain supplementary information and add feature parameters corresponding to the supplementary information to the patient care sample; To adjust the weight, through the trauma category judgment table, obtain the judgment data item corresponding to the modified trauma type output by the current model, multiply the weight of the characteristic parameter corresponding to the obtained judgment data item in the patient care sample by the adjustment coefficient, and the calculation result is the weight adjustment; the same method is used to adjust the weight for the nursing stage.

5. The method for intelligent analysis of trauma orthopedic nursing data according to claim 1, characterized in that: Nursing data analysis models include: The input layer receives the reconstructed patient care samples, including category 1 and category 2 care data; Convolutional layers, multiple convolutional layers are used to extract local features and patterns in patient care data; Pooling layer, used to reduce feature dimensions and computational complexity and improve the robustness of the model; The fully connected layer is used to integrate the features extracted by the convolutional layer and the pooling layer and map them to the numbers of the subsequent three sub-stages; The output layer outputs the numbers of the three subsequent sub-stages of the trauma orthopedic care plan required by the patient; The group trend optimization module embeds the group trend factor into the model to adjust the model hyperparameters; Activation function: Softmax activation function is used in convolutional layers, fully connected layers, and output layers to introduce nonlinear relationships and improve the expressiveness of the model; The construction of nursing data analysis model includes: Obtain several patient care samples and corresponding trauma orthopedic care plans, and use the subsequent three sub-stages of the trauma orthopedic care plan corresponding to the patient's current care stage as sample labels; organize the samples into a stage prediction data set, and divide the stage prediction data set into a training set, a validation set, and a test set; The initial nursing data analysis model is trained using the training set and validation set in the stage prediction data set, and the hyperparameters of the initial stage prediction model are adjusted by the group trend optimization algorithm to obtain the optimized nursing data analysis model; The stage prediction test set is used to test the accuracy of the optimized stage prediction model. If the accuracy is within the set range, the optimized stage prediction model is used as the stage prediction model. Otherwise, the steps of adjusting the hyperparameters of the initial stage prediction model are repeated until the accuracy is less than the set range.

6. The method for intelligent analysis of trauma orthopedic nursing data according to claim 5, characterized in that: The group trend factor optimization algorithm includes: Initialize the group, set the size of the particle group, the parameters corresponding to each particle represent the variables related to the stage prediction, and randomly initialize the parameter combination of the stage prediction model; Calculate the fitness of the particle, which is based on the prediction error of the particle; Get the reference direction vector of each particle: ; In the formula represents the historical optimal position of the group, represents the current position of the particle, and the reference direction vector represents the direction between the current position of the particle and the historical optimal solution of the group; Calculate the cosine value of the angle between the particle's current velocity vector and the reference direction vector; count the number of particles whose cosine value of the angle is greater than the set threshold, and use the ratio as the group trend factor STF; set the limit value of the group trend factor, which is used to distinguish the role of the group trend factor. Specifically, when the group trend factor is greater than the limit value, the group is in a local development state and the inertia weight is reduced. When the group trend factor is less than the limit value, the group is in a global exploration state and the inertia weight is increased. The inertia weight is dynamically adjusted according to the group trend factor, and the inertia weight is used to calculate the speed and position of the particle; for each particle, the current fitness of the particle is compared with the fitness of the historical optimal solution. If the current fitness is larger, it is updated to the optimal solution. The above process is iterated repeatedly to obtain the optimal parameter combination of the model and the prediction model of the optimization stage.

7. The method for intelligent analysis of trauma orthopedic nursing data according to claim 1, characterized in that: Similarity calculation includes: The similarity calculation is performed through cosine similarity. The patient category 1 nursing data and standard patient data are converted into character strings. The similarity of the data is obtained by calculating the cosine similarity of the two character strings. A high cosine similarity means that the data are close.

8. An intelligent analysis system for trauma orthopedic nursing data, characterized in that: The method for intelligent analysis of trauma orthopedic nursing data according to any one of claims 1 to 7 is applied, and the intelligent analysis system for trauma orthopedic nursing data includes the following modules: A program management module is used to manage a trauma orthopedic nursing program library, which stores trauma orthopedic nursing programs corresponding to different trauma types. The trauma orthopedic nursing program consists of five major stages, including the acute stage, trauma recovery stage, functional recovery stage, rehabilitation stage, and follow-up stage. The major stages of the trauma orthopedic nursing program are divided based on the nursing stages of trauma orthopedic patients, and the major stages of the trauma orthopedic nursing program are divided into multiple sub-stages based on nursing operations; The data collection module is used to obtain nursing data of trauma orthopedic patients, including Class 1 nursing data and Class 2 nursing data. Class 1 nursing data includes basic information of patients, types of patient trauma, stages of patient care, wound surface information, vital signs information, complication information, laboratory evaluation information and imaging evaluation information, among which laboratory evaluation information and imaging evaluation information are supplementary information; Class 2 nursing data includes implemented nursing record information, medication record information, patient psychological information and patient follow-up information; The data processing module is used to process the nursing data of trauma orthopedic patients. The processing operations include data cleaning, feature labeling, weight construction, outlier detection and null value labeling, obtain the feature parameters of the nursing data of trauma orthopedic patients, and construct the feature parameters into patient care samples; The characteristic parameters are stored in matrices, including two matrices corresponding to category 1 nursing data and category 2 nursing data; The nursing stage analysis module includes a nursing stage analysis unit and a nursing result correction unit. The nursing stage analysis unit is used to send the corresponding matrix of Class 1 nursing data in the patient nursing sample into the nursing stage analysis model for calculation to obtain the corrected trauma type and the corrected patient nursing stage; the nursing result correction unit is used to determine whether to correct the output result according to the accuracy of the output result of the nursing stage analysis model. If the corrected trauma type or the corrected patient nursing stage is inconsistent with the patient trauma type and patient nursing stage in Class 1 nursing data, detect whether there is supplementary information in Class 1 nursing data; if the supplementary information is empty, compare the accuracy of the model output result with the first confirmation threshold. If it is greater than the first confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient. Otherwise, obtain the supplementary information, reconstruct the patient nursing sample, and then send the sample to the nursing stage analysis model for calculation again; If the supplementary information is not empty, the accuracy of the model output result is compared with the second confirmation threshold. If it is greater than the second confirmation threshold, the current model output result is the correct trauma type and nursing stage of the patient. Otherwise, the patient care sample is reconstructed and the sample is sent to the nursing stage analysis model for recalculation. When the calculation is performed again, the modified confirmation threshold is used for detection; the reconstruction operation of the patient care sample includes weight adjustment; The nursing data analysis module is used to reconstruct the patient nursing samples according to the correct trauma type and nursing stage of the patient, and send the reconstructed patient nursing samples to the nursing data analysis model for calculation to obtain the numbers of the subsequent three sub-stages corresponding to the trauma orthopedic nursing plan required by the patient; the nursing data analysis model is constructed based on a deep convolutional neural network, and a group trend factor is introduced to optimize the model, and the inertia weight is dynamically adjusted according to the group state to improve the accuracy of the model calculation and the training efficiency of the model; The nursing plan optimization module is used to match the trauma orthopedic nursing plans in the trauma orthopedic nursing plan library according to the numbers of the subsequent three sub-stages and the sub-stages that have been implemented in the trauma orthopedic nursing process, and use the plan with the highest matching degree as the trauma orthopedic nursing plan required by the patient; If there are multiple plans with the highest matching degree, similarity calculation is performed based on the patient's Category 1 nursing data and the standard patient data corresponding to the matching plan, and the plan with the highest score is used as the trauma orthopedic nursing plan required by the patient; the matching method and similarity calculation are both performed through cosine similarity.

Images