Teaching demonstration system for endocrinology department nursing

Abstract

The invention belongs to the technical field of intelligent teaching demonstration, and particularly relates to a teaching demonstration system for endocrinology department nursing, which comprises a data acquisition module, a data preprocessing module, a knowledge teaching demonstration module and a nursing operation demonstration module, according to the system, a convolutional neural network of a 3DU-Net architecture is adopted through a knowledge teaching display module, a semantic segmentation model is constructed to extract image features, a learner can recognize a complex structure of an endocrine organ, a multi-dimensional coordinate system is established based on the endocrine organ, and the geometric shape of the organ is displayed through point cloud processing, so that the learning efficiency is improved. A learner can understand the structure and the mutual relation of the organs from the spatial dimension; a nursing operation model is constructed through the nursing operation demonstration module to learn a space-time joint manifold, a nursing operation track is generated and optimized, the track is mapped to an organ 3D model to demonstrate nursing actions, hand action details and force control of nursing personnel are accurately simulated, and the proficiency and accuracy of nursing operation can be effectively improved.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent teaching demonstration, and specifically refers to a teaching demonstration system for endocrinology nursing. Background Art

[0002] The teaching demonstration system of endocrinology nursing can improve nursing staff's understanding of endocrinology knowledge and operational skills through data processing and model construction, and enhance the professionalism and accuracy of nursing work. However, the current teaching demonstration system of endocrinology nursing has the following shortcomings:

[0003] The teaching process of the teaching demonstration system cannot closely combine the morphological knowledge of endocrine organs with actual nursing operations. It is difficult for nursing staff to intuitively understand the impact of different organ morphological characteristics on the key points and difficulties of nursing operations. When performing nursing operations, they lack pertinence and cannot determine the best operation method based on the anatomical position and morphological characteristics of the organs, which reduces the effectiveness of nursing operations in clinical practice.

[0004] The organ display in the teaching demonstration system is not clear enough, and it is difficult for medical staff to clearly identify the subtle structure of endocrine organs. They are unable to deeply explore the intrinsic connection between nursing operations and organ structure, which limits the mastery of professional knowledge. Nurses lack sufficient virtual practice before actual operations, resulting in poor control of hand movement details and strength in clinical practice, low operation accuracy, and easy mistakes, which increases the risk of patient discomfort and complications. Summary of the invention

[0005] In view of the above problems, the present invention provides a teaching demonstration system for endocrinology nursing. Through the knowledge teaching display module, a convolutional neural network with a 3DU-Net architecture is adopted to construct a semantic segmentation model to extract image features, which is helpful for learners to identify the complex structure of endocrine organs, and a multi-dimensional coordinate system is established based on the endocrine organs. The geometric shape of the organs is displayed after point cloud processing, which is helpful for learners to understand the structure and relationship of the organs from a spatial dimension; through the nursing operation demonstration module, a nursing operation model is constructed to learn the spatiotemporal joint manifold, generate and optimize the nursing operation trajectory, and map it to the organ 3D model to demonstrate the nursing action, accurately simulating the hand movement details and strength control of the nursing staff, which can effectively improve the proficiency and accuracy of the nursing operation and reduce the risk of patient discomfort and complications caused by improper operation.

[0006] A teaching demonstration system for endocrinology nursing, including a data acquisition module, a data preprocessing module, a knowledge teaching display module and a nursing operation demonstration module;

[0007] The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the knowledge teaching display module, and the knowledge teaching display module is connected to the nursing operation demonstration module;

[0008] The data acquisition module acquires CT scan images of the human endocrine system, measures endocrine data of hormones and metabolites, and acquires nursing operation data;

[0009] The data preprocessing module performs preprocessing operations on the CT scan image and the endocrine data to obtain preprocessed images and preprocessed data;

[0010] The knowledge teaching and display module establishes a coordinate system according to the endocrine organ, uniformly samples to generate a point cloud, iteratively updates to obtain a stable point cloud, and displays the geometric shape of the endocrine organ;

[0011] The nursing operation demonstration module constructs a nursing operation model to learn the spatiotemporal joint manifold, optimizes the nursing operation trajectory, synthesizes the nursing operation trajectory with the preprocessed image and preprocessed data into a triple data set, constructs an organ 3D model, and dynamically demonstrates nursing actions by mapping the triple data set.

[0012] The data preprocessing module obtains the process of preprocessing images and preprocessing data, including the following steps:

[0013] Step P1: Image denoising: Use Gaussian filtering method to remove noise interference from CT scan images;

[0014] Step P2: assigning image labels: using image segmentation algorithms to process CT scan images, identify endocrine organs, endocrine tissues and endocrine structures, and assign image labels respectively to obtain preprocessed images;

[0015] Step P3: Perform data cleaning: set the data range, clean the abnormal values ​​in the endocrine data that exceed the data range, normalize the endocrine data, and obtain preprocessed data.

[0016] The knowledge teaching display module displays the geometric shape of endocrine organs, including the following steps: knowledge graph, knowledge retrieval

[0017] Step S1: Processing image features: A convolutional neural network based on the 3DU-Net architecture is used to build a semantic segmentation model. The semantic segmentation model includes an encoding layer and a decoding layer. The encoding layer gradually extracts image features of the preprocessed image through convolution and pooling operations to obtain high-level semantic information. The decoding layer obtains fused image features through upsampling and convolution operations.

[0018] Step S2: Establishing a coordinate system: Establishing a coordinate system based on the endocrine organ, determining the parameters and range of the coordinate system, the coordinate system includes the depth coordinates in the vertical direction, the hierarchical coordinates from the surface to the inside, the angle coordinates around the endocrine organ, and the regional correlation coordinates for distinguishing the left and right lobes;

[0019] Step S3: Optimizing the mesh model: Based on the fused image features and the coordinate system, using a mesh generation algorithm, constructing an initial mesh model of the endocrine organ, optimizing the initial mesh model, and obtaining an optimized mesh model;

[0020] Step S4: iteratively updating the point cloud: uniformly sampling based on the optimized grid model to form a point cloud, selecting an initial reference point cloud, iteratively updating the remaining point clouds, and obtaining a stable point cloud;

[0021] Step S5: Displaying the organ shape: vectorizing and centering the stability point cloud to obtain a centering matrix, performing principal component analysis on the centering matrix, displaying the geometric shape of the endocrine organ, and calculating the principal component score;

[0022] Step S6: Calculate the reconstruction error: Calculate the reconstructed point cloud based on the principal component vector and the mean vector of the principal component score, calculate the Euclidean distance between the reconstructed point cloud and the stability point cloud, obtain the reconstruction error, and use the reconstruction error as an evaluation indicator to evaluate the stability point cloud.

[0023] Further, step S4 includes the following steps:

[0024] Step S41: Initially updating the point cloud: uniformly sampling on the coordinate system to obtain reference points, combining the reference points to form a point cloud, randomly selecting an initial reference point cloud, and the point clouds other than the initial reference point cloud in the point cloud are the remaining point clouds, iteratively updating the remaining point clouds, calculating the translation vectors and rotation matrices between the remaining point clouds, and performing translation and rotation operations on the remaining point clouds;

[0025] Step S42: Update the point cloud again: weighted average the coordinates of all point clouds to obtain an average point cloud, use the average point cloud as a new reference point cloud, and the point clouds other than the new reference point cloud in the point cloud are the remaining point clouds. Continue to calculate the translation vectors and rotation matrices between the remaining point clouds, perform translation and rotation operations on the remaining point clouds, and use the Laplacian smoothing algorithm to smooth the remaining point clouds.

[0026] The remaining point clouds are smoothed using the following formula:

[0027] ;

[0028] Among them, i and j represent the node index of the initial mesh model, represents the jth node of the initial mesh model, represents the new position of the i-th node after smoothing, represents the set of neighboring nodes of the ith node, Represents the number of nodes in the neighborhood node set;

[0029] Step S43: Evaluate point cloud index: set a stability index to measure the stability of the average point cloud, stop iterative updating when the stability index is reached, and obtain a stable point cloud.

[0030] Further, step S5 includes the following steps:

[0031] Step S51: vectorization processing: vectorize each stability point cloud to obtain the vector of each stability point cloud. Each stability point cloud contains n reference points, each reference point contains 4 coordinate dimensions, and the coordinate dimensions are (x, y, a, b). Each stability point cloud is abstracted into an N×4n point cloud matrix, where N represents the number of preprocessed images.

[0032] Step S52: Centralization processing: extract the coordinate dimension of the coordinate system, calculate the average value of all point clouds for each coordinate dimension, subtract the average value of the corresponding coordinate dimension from the vector of each stability point cloud, and obtain a centralization matrix;

[0033] Step S53: principal component analysis: perform principal component analysis on the centralization matrix, calculate the covariance matrix, and solve the eigenvalues ​​and eigenvectors of the covariance matrix;

[0034] The covariance matrix is ​​calculated using the following formula:

[0035] ;

[0036] in, represents the number of reference points, i represents the reference point index, j and k represent the coordinate dimension index, represents the elements in the covariance matrix, Represents the value of the j-th coordinate dimension of the i-th data point, Represents the value of the kth coordinate dimension of the ith data point, represents the average value of the j-th coordinate dimension, Represents the average value of the kth coordinate dimension;

[0037] Step S54: Sorting of eigenvectors: Sort the eigenvectors based on the size of the eigenvalues, select some principal components, display the geometric shapes of the endocrine organs, and calculate the principal component scores, which include the principal component vectors and the mean vectors.

[0038] The nursing operation demonstration module dynamically demonstrates the process of nursing actions by mapping the triplet data set, including the following steps:

[0039] Step D1: Constructing a nursing operation model: Construct a convolutional autoencoder model as a nursing operation model, including a time convolutional encoder and decoder, and set the time step and parameters of the nursing operation model;

[0040] Step D2: Generate initial latent code: Using the traditional inverse kinematics method, the nursing operation data is input into the temporal convolution encoder to generate the initial latent code. Based on the initial latent code, the connection between the nursing operation data and the preprocessed image and preprocessed data is preliminarily established to generate output features.

[0041] The formula used to generate output features is as follows:

[0042] ;

[0043] in, Indicates the layer index, Represents the position coordinates, Indicates The output features of the layer, Indicates The output feature map of the layer, Indicates The number of channels of the layer, Indicates The convolution kernel weight matrix of the layer, Indicates The bias vector of the layer, represents the activation function, i represents the index of traversing the previous layer of channels, a and b represent the index of traversing the position of the convolution kernel elements, Indicates The convolution kernel size of the layer, Indicates The step size of the layer;

[0044] Step D3: Optimize the initial latent code: Through the temporal convolution encoder, learn the spatiotemporal joint manifold between the nursing operation data and the CT scan image and endocrine data, input the initial latent code into the decoder, optimize the initial latent code according to the output features, and obtain the nursing operation trajectory;

[0045] The formula used is as follows:

[0046] ;

[0047] in, Represents the learned joint spatiotemporal manifold, Represents nursing operation data, represents the potential code of nursing operation data, Y represents endocrine data, Potential codes representing endocrine data, represents the reconstruction error, represents the correlation loss, represents the weight of the correlation loss;

[0048] Step D4: synthesizing a triplet data set: synthesizing the nursing operation trajectory with the preprocessed image and the preprocessed data to obtain a triplet data set;

[0049] Step D5: Dynamically demonstrate nursing actions: construct an organ 3D model, map the triplet dataset to the organ 3D model according to the time step, and dynamically demonstrate the nursing actions and corresponding reactions of the body parts at each time step.

[0050] The beneficial effects achieved by the present invention are as follows:

[0051] (1) The knowledge teaching and demonstration module of the system uses a convolutional neural network with a 3DU-Net architecture to build a semantic segmentation model, which can deeply extract the features of preprocessed images and obtain high-level semantic information, which helps to more clearly identify the subtle structures of endocrine organs during the learning process and deepen the understanding of the physiological and pathological basis of endocrine organs;

[0052] (2) The nursing operation demonstration module of the system constructs a nursing operation model to learn the spatiotemporal joint manifold, uses traditional inverse kinematics methods to generate initial latent codes and optimize the nursing operation trajectory, accurately simulates the movement details and force control of the nurse's hands during injection, and enables the nurse to conduct sufficient virtual practice before the actual operation, so as to perform nursing operations more accurately in clinical practice and reduce the risk of patient discomfort and complications caused by improper operation;

[0053] (3) The knowledge teaching demonstration module of the system combines the coordinate system established based on the endocrine organs with the ternary data set synthesized by the nursing operation demonstration module, which closely links abstract knowledge with actual nursing operations. During the learning process, nurses can intuitively see how the morphological characteristics of different endocrine organs affect the key points and difficulties of nursing operations. For example, the optimal location and angle of insulin injection can be determined based on the anatomical position and morphological characteristics of the pancreas, thereby improving the pertinence and effectiveness of nursing operations in clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A module schematic diagram of a teaching demonstration system for endocrinology nursing proposed by the present invention;

[0055] Figure 2 This is a schematic diagram of the processing flow of the knowledge teaching display module proposed in the present invention. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0057] Example 1: Reference Figure 1,This embodiment provides a teaching demonstration system for endocrinology nursing, including a data acquisition module, a data preprocessing module, a knowledge teaching display module and a nursing operation demonstration module;

[0058] The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the knowledge teaching display module, and the knowledge teaching display module is connected to the nursing operation demonstration module;

[0059] The data acquisition module acquires CT scan images of the human endocrine system, measures endocrine data of hormones and metabolites, and acquires nursing operation data;

[0060] The data preprocessing module performs preprocessing operations on the CT scan image and the endocrine data to obtain preprocessed images and preprocessed data;

[0061] The knowledge teaching and display module establishes a coordinate system according to the endocrine organ, uniformly samples to generate a point cloud, iteratively updates to obtain a stable point cloud, and displays the geometric shape of the secretory organ;

[0062] The nursing operation demonstration module constructs a nursing operation model to learn the spatiotemporal joint manifold, optimizes the nursing operation trajectory, synthesizes the nursing operation trajectory with the preprocessed image and preprocessed data into a triple data set, constructs an organ 3D model, and dynamically demonstrates nursing actions by mapping the triple data set.

[0063] Example 2: Reference Figure 2 This embodiment is based on embodiment 1, and the process of obtaining the preprocessed image and preprocessed data by the data preprocessing module includes the following steps:

[0064] Step P1: Image denoising: Use Gaussian filtering method to remove noise interference from CT scan images;

[0065] Step P2: assigning image labels: using image segmentation algorithms to process CT scan images, identify endocrine organs, endocrine tissues and endocrine structures, and assign image labels respectively to obtain preprocessed images;

[0066] In this example, thyroid follicles, thyroid stroma, and vascular structures are identified;

[0067] Step P3: Perform data cleaning: set the data range, clean the abnormal values ​​in the endocrine data that exceed the data range, normalize the endocrine data, and obtain preprocessed data.

[0068] Example 3: This example is based on Example 2. The knowledge teaching display module displays the geometric shape of endocrine organs, including the following steps: knowledge graph, knowledge retrieval

[0069] Step S1: Processing image features: A convolutional neural network based on the 3DU-Net architecture is used to build a semantic segmentation model. The semantic segmentation model includes an encoding layer and a decoding layer. The encoding layer gradually extracts image features of the preprocessed image through convolution and pooling operations to obtain high-level semantic information. The decoding layer fuses image features through upsampling and convolution operations.

[0070] Step S2: Establishing a coordinate system: Establishing a coordinate system based on the endocrine organ, determining the parameters and range of the coordinate system, the coordinate system includes the depth coordinates in the vertical direction, the hierarchical coordinates from the surface to the inside, the angle coordinates around the endocrine organ, and the regional correlation coordinates for distinguishing the left and right lobes;

[0071] Step S3: Optimizing the mesh model: Based on the fused image features and the coordinate system, using a mesh generation algorithm, constructing an initial mesh model of the endocrine organ, optimizing the initial mesh model, and obtaining an optimized mesh model;

[0072] Step S4: iteratively updating the point cloud: uniformly sampling based on the optimized grid model to form a point cloud, selecting an initial reference point cloud, iteratively updating the remaining point clouds, and obtaining a stable point cloud;

[0073] Step S4 includes the following steps:

[0074] Step S41: Initially updating the point cloud: uniformly sampling on the coordinate system to obtain reference points, combining the reference points to form a point cloud, randomly selecting an initial reference point cloud, and the point clouds other than the initial reference point cloud in the point cloud are the remaining point clouds, iteratively updating the remaining point clouds, calculating the translation vectors and rotation matrices between the remaining point clouds, and performing translation and rotation operations on the remaining point clouds;

[0075] Step S42: Update the point cloud again: weighted average the coordinates of all point clouds to obtain an average point cloud, use the average point cloud as a new reference point cloud, and the point clouds other than the new reference point cloud in the point cloud are the remaining point clouds. Continue to calculate the translation vectors and rotation matrices between the remaining point clouds, perform translation and rotation operations on the remaining point clouds, and use the Laplacian smoothing algorithm to smooth the remaining point clouds.

[0076] The remaining point clouds are smoothed using the following formula:

[0077] ;

[0078] Among them, i and j represent the node index of the initial mesh model, represents the jth node of the initial mesh model, represents the new position of the i-th node after smoothing, represents the set of neighboring nodes of the ith node, Represents the number of nodes in the neighborhood node set;

[0079] Step S43: Evaluate the point cloud index: set a stability index to measure the stability of the average point cloud, stop iterative updating when the stability index is reached, and obtain a stable point cloud;

[0080] Step S5: Displaying the organ shape: vectorizing and centering the stability point cloud to obtain a centering matrix, performing principal component analysis on the centering matrix, displaying the geometric shape of the endocrine organ, and calculating the principal component score;

[0081] Step S5 includes the following steps:

[0082] Step S51: vectorization processing: vectorize each stability point cloud to obtain the vector of each stability point cloud. Each stability point cloud contains n reference points, each reference point contains 4 coordinate dimensions, and the coordinate dimensions are (x, y, a, b). Each stability point cloud is abstracted into an N×4n point cloud matrix, where N represents the number of preprocessed images.

[0083] Step S52: Centralization processing: extract the coordinate dimension of the coordinate system, calculate the average value of all point clouds for each coordinate dimension, subtract the average value of the corresponding coordinate dimension from the vector of each stability point cloud, and obtain a centralization matrix;

[0084] Step S53: principal component analysis: perform principal component analysis on the centralization matrix, calculate the covariance matrix, and solve the eigenvalues ​​and eigenvectors of the covariance matrix;

[0085] The covariance matrix is ​​calculated using the following formula:

[0086] ;

[0087] in, represents the number of reference points, i represents the reference point index, j and k represent the coordinate dimension index, represents the elements in the covariance matrix, Represents the value of the j-th coordinate dimension of the i-th data point, Represents the value of the kth coordinate dimension of the ith data point, represents the average value of the j-th coordinate dimension, Represents the average value of the kth coordinate dimension;

[0088] Step S54: Sorting eigenvectors: Sorting eigenvectors based on eigenvalues, selecting some principal components, displaying the geometric shapes of endocrine organs, and calculating principal component scores, which include principal component vectors and mean vectors;

[0089] Step S6: Calculate the reconstruction error: Calculate the reconstructed point cloud based on the principal component vector and the mean vector of the principal component score, calculate the Euclidean distance between the reconstructed point cloud and the stability point cloud, obtain the reconstruction error, and use the reconstruction error as an evaluation indicator to evaluate the stability point cloud.

[0090] Example 4: This example is based on Example 2. The knowledge teaching display module displays the geometric shape of endocrine organs, including the following steps: knowledge graph, knowledge retrieval

[0091] Step S1: Processing image features: A convolutional neural network based on the 3DU-Net architecture is used to build a semantic segmentation model. The semantic segmentation model includes an encoding layer and a decoding layer. The encoding layer gradually extracts image features of the preprocessed image through convolution and pooling operations to obtain high-level semantic information. The decoding layer fuses image features through upsampling and convolution operations.

[0092] Step S2: Establishing a coordinate system: establishing a coordinate system based on the endocrine organs, and determining the parameters and range of the coordinate system;

[0093] Step S3: Optimizing the mesh model: Based on the image features and the coordinate system, using a mesh generation algorithm, constructing an initial mesh model of the endocrine organ, optimizing the initial mesh model, and obtaining an optimized mesh model;

[0094] Step S4: iteratively updating the point cloud: uniformly sampling based on the optimized grid model to form a point cloud, selecting an initial reference point cloud, iteratively updating the remaining point clouds, and obtaining a stable point cloud;

[0095] Step S5: Displaying the organ shape: vectorizing and centering the stability point cloud to obtain a centering matrix, performing principal component analysis on the centering matrix, displaying the geometric shape of the endocrine organ, and calculating the principal component score;

[0096] Step S6: Calculate the reconstruction error: Calculate the reconstructed point cloud based on the principal component vector and the mean vector of the principal component score, calculate the Euclidean distance between the reconstructed point cloud and the stability point cloud, obtain the reconstruction error, and use the reconstruction error as an evaluation indicator to evaluate the stability point cloud.

[0097] Embodiment 5: This embodiment is based on embodiment 3, and the nursing operation demonstration module dynamically demonstrates the process of nursing actions by mapping the triplet data set, including the following steps:

[0098] Step D1: Constructing a nursing operation model: Construct a convolutional autoencoder model as a nursing operation model, including a time convolutional encoder and decoder, and set the time step and parameters of the nursing operation model;

[0099] Step D2: Generate initial latent code: Using the traditional inverse kinematics method, the nursing operation data is input into the temporal convolution encoder to generate the initial latent code. Based on the initial latent code, the connection between the nursing operation data and the preprocessed image and preprocessed data is preliminarily established to generate output features.

[0100] The formula used to generate output features is as follows:

[0101] ;

[0102] in, Indicates the layer index, Represents the position coordinates, Indicates The output features of the layer, Indicates The output feature map of the layer, Indicates The number of channels of the layer, Indicates The convolution kernel weight matrix of the layer, Indicates The bias vector of the layer, represents the activation function, i represents the index of traversing the previous layer of channels, a and b represent the index of traversing the position of the convolution kernel elements, Indicates The convolution kernel size of the layer, Indicates The step size of the layer;

[0103] Step D3: Optimize the initial latent code: Through the temporal convolution encoder, learn the spatiotemporal joint manifold between the nursing operation data and the CT scan image and endocrine data, input the initial latent code into the decoder, optimize the initial latent code according to the output features, and obtain the nursing operation trajectory;

[0104] The formula used is as follows:

[0105] ;

[0106] in, Represents the learned joint spatiotemporal manifold, Represents nursing operation data, represents the potential code of nursing operation data, Y represents endocrine data, Potential codes representing endocrine data, represents the reconstruction error, represents the correlation loss, represents the weight of the correlation loss;

[0107] Step D4: synthesizing a triplet data set: synthesizing the nursing operation trajectory with the preprocessed image and the preprocessed data to obtain a triplet data set;

[0108] Step D5: Dynamically demonstrate nursing actions: construct an organ 3D model, map the triplet data set to the organ 3D model according to the time step, and dynamically demonstrate the nursing actions and corresponding reactions of body parts at each time step;

[0109] In this embodiment, the dynamic demonstration nursing action is as follows: the nursing staff first takes out the insulin pen from the insulin pen box and carefully checks the remaining amount of the pen cartridge and the condition of the needle; when preparing for the injection, the nursing staff gently pinches the skin on the patient's abdomen with one hand to form a small skin fold, and holds the insulin pen firmly with the other hand, and slowly inserts the needle into the skin fold at a 45-degree angle;

[0110] The corresponding reactions of the body parts are: the patient's abdominal muscles will tighten unconsciously. During the injection, the patient may feel a slight tingling sensation. The body will remain relatively still and the attention will be focused on the injection site. After the injection is completed, the nurse will pull out the needle, and a small needle hole will be left on the patient's abdominal skin. There may be a very small amount of bleeding spots, and then the body will gradually relax. It is necessary to pay attention to whether there is any discomfort reaction at the injection site afterwards.

[0111] Embodiment 6: This embodiment is based on embodiment 4, and the nursing operation demonstration module dynamically demonstrates the process of nursing actions by mapping the triplet data set, including the following steps:

[0112] Step D1: Constructing a nursing operation model: Construct a convolutional autoencoder model as a nursing operation model, including an encoder and a decoder, and set the nursing operation model parameters;

[0113] Step D2: Generate initial latent code: Input the nursing operation data into the encoder to generate initial latent code, and based on the initial latent code, preliminarily establish the connection between the nursing operation data and the preprocessed image and preprocessed data to generate output features;

[0114] Step D3: Optimizing the initial latent code: Through the temporal convolution encoder, the connection between the nursing operation data and the CT scan image and endocrine data is learned, the initial latent code is input into the decoder, and the initial latent code is optimized according to the output features to obtain the nursing operation trajectory;

[0115] Step D4: synthesizing a triplet data set: synthesizing the nursing operation trajectory with the preprocessed image and the preprocessed data to obtain a triplet data set;

[0116] Step D5: Dynamically demonstrate nursing actions: construct an organ 3D model, map the triplet dataset to the organ 3D model according to the time step, and dynamically demonstrate the nursing actions and corresponding reactions of the body parts at each time step.

[0117] The present invention and its implementation methods are described above. Such description is not restrictive. If a person skilled in the art is inspired by it and creatively designs structures and implementation methods similar to the technical solution without departing from the purpose of the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A teaching demonstration system for endocrinology nursing, comprising a data acquisition module and a data preprocessing module, wherein the data acquisition module acquires CT scan images of the endocrine system, measures endocrine data of hormones and metabolites, and acquires nursing operation data; and the data preprocessing module performs preprocessing operations on the CT scan images and endocrine data to obtain preprocessed images and preprocessed data; and characterized in that: The system also includes a knowledge teaching display module and a nursing operation demonstration module; The knowledge teaching and display module establishes a coordinate system according to the endocrine organ, uniformly samples to generate a point cloud, iteratively updates to obtain a stable point cloud, and displays the geometric shape of the secretory organ; The nursing operation demonstration module constructs a nursing operation model to learn the spatiotemporal joint manifold, optimizes the nursing operation trajectory, synthesizes the nursing operation trajectory with the preprocessed image and preprocessed data into a triple data set, constructs an organ 3D model, and dynamically demonstrates nursing actions by mapping the triple data set.

2. A teaching demonstration system for endocrinology nursing according to claim 1, characterized in that: The knowledge teaching demonstration module represents the process of the geometric shape of the endocrine organ, including the following steps: Step S1: Processing image features: A convolutional neural network based on the 3DU-Net architecture is used to build a semantic segmentation model. The semantic segmentation model includes an encoding layer and a decoding layer. The encoding layer gradually extracts image features of the preprocessed image through convolution and pooling operations, and the decoding layer obtains fused image features through upsampling and convolution operations. Step S2: Establishing a coordinate system: establishing a coordinate system based on the endocrine organs, and determining the parameters and range of the coordinate system; Step S3: Optimizing the mesh model: Based on the fused image features and the coordinate system, using a mesh generation algorithm, constructing an initial mesh model of the endocrine organ, optimizing the initial mesh model, and obtaining an optimized mesh model; Step S4: iteratively updating the point cloud: uniformly sampling based on the optimized grid model to form a point cloud, selecting an initial reference point cloud, iteratively updating the remaining point clouds, and obtaining a stable point cloud; Step S5: Displaying the organ shape: vectorizing and centering the stability point cloud to obtain a centering matrix, performing principal component analysis on the centering matrix, displaying the geometric shape of the endocrine organ, and calculating the principal component score; Step S6: Calculate the reconstruction error: Calculate the reconstructed point cloud based on the principal component vector and the mean vector of the principal component score, calculate the Euclidean distance between the reconstructed point cloud and the stability point cloud, obtain the reconstruction error, and use the reconstruction error as an evaluation indicator to evaluate the stability point cloud.

3. A teaching demonstration system for endocrinology nursing according to claim 2, characterized in that: Step S4 includes the following steps: Step S41: Initially updating the point cloud: uniformly sampling on the coordinate system to obtain reference points, combining the reference points to form a point cloud, randomly selecting an initial reference point cloud, and the point clouds other than the initial reference point cloud in the point cloud are the remaining point clouds, iteratively updating the remaining point clouds, calculating the translation vectors and rotation matrices between the remaining point clouds, and performing translation and rotation operations on the remaining point clouds; Step S42: Update the point cloud again: weighted average the coordinates of all point clouds to obtain an average point cloud, use the average point cloud as a new reference point cloud, and the point clouds other than the new reference point cloud in the point cloud are the remaining point clouds. Continue to calculate the translation vectors and rotation matrices between the remaining point clouds, perform translation and rotation operations on the remaining point clouds, and use the Laplacian smoothing algorithm to smooth the remaining point clouds. Step S43: Evaluate point cloud index: set a stability index to measure the stability of the average point cloud, stop iterative updating when the stability index is reached, and obtain a stable point cloud.

4. A teaching demonstration system for endocrinology nursing according to claim 3, characterized in that: Step S5 includes the following steps: Step S51: vectorization processing: vectorization processing is performed on each stability point cloud to obtain a vector of each stability point cloud; Step S52: Centralization processing: extract the coordinate dimension of the coordinate system, calculate the average value of all point clouds for each coordinate dimension, subtract the average value of the corresponding coordinate dimension from the vector of each stability point cloud, and obtain a centralization matrix; Step S53: principal component analysis: perform principal component analysis on the centralization matrix, calculate the covariance matrix, and solve the eigenvalues ​​and eigenvectors of the covariance matrix; Step S54: Sorting of eigenvectors: Sorting the eigenvectors based on the size of the eigenvalues, selecting some principal components to represent the geometric shapes of the endocrine organs, and calculating the principal component scores.

5. A teaching demonstration system for endocrinology nursing according to claim 2, characterized in that: The nursing operation demonstration module dynamically displays the nursing action process by mapping the triplet data set, including the following steps: Step D1: Constructing a nursing operation model: Construct a convolutional autoencoder model as a nursing operation model, including a time convolutional encoder and decoder, and set the time step and parameters of the nursing operation model; Step D2: Generate initial latent code: Using the traditional inverse kinematics method, the nursing operation data is input into the temporal convolution encoder to generate the initial latent code. Based on the initial latent code, the connection between the nursing operation data and the preprocessed image and preprocessed data is preliminarily established to generate output features; Step D3: Optimize the initial latent code: Through the temporal convolution encoder, learn the spatiotemporal joint manifold between the nursing operation data and the CT scan image and endocrine data, input the initial latent code into the decoder, optimize the initial latent code according to the output features, and obtain the nursing operation trajectory; Step D4: synthesizing a triplet data set: synthesizing the nursing operation trajectory with the preprocessed image and the preprocessed data to obtain a triplet data set; Step D5: Dynamically demonstrate nursing actions: construct an organ 3D model, map the triplet dataset to the organ 3D model according to the time step, and dynamically demonstrate the nursing actions and corresponding reactions of body parts at each time step.

Images