Data management system and method for stoma nursing teaching

Abstract

The present application provides a data management system and method for stoma nursing teaching, which relates to the field of nursing teaching. It first stores the identity information of nursing students and teachers, as well as stoma nursing teaching video collections and virtual training tasks, then collects and transmits the training operation video of the target nursing student object during the virtual training process, and then performs a core information significant interactive analysis based on the training operation characteristics and the skeletal point position characteristics on the received training operation video to obtain the stoma nursing training feedback result of the target nursing student object, and finally generates the training operation suggestion based on the obtained stoma nursing training feedback result, and displays and stores it. In this way, not only can the teaching effect be improved and the cost be reduced, but also the risks in the practice of nursing students are avoided and the safety of teaching is enhanced. Nursing students can use virtual training to improve the proficiency and accuracy of stoma nursing operations, thereby better serving patients and improving patients' trust and satisfaction.

Description

Technical Field

[0001] The present application relates to the field of nursing teaching, and more specifically, to a data management system and method for stoma nursing teaching. Background Art

[0002] A stoma is an artificial opening created on the abdominal wall to treat intestinal or urinary system diseases. A section of the intestine or urethra is pulled out of the opening to discharge excrement. As an important part of medical nursing, stoma care involves complex nursing processes and key points of operation.

[0003] However, the traditional teaching model mainly relies on teachers' explanations and nursing students' actual operations, which has many shortcomings. First, the teaching cost is high, and patients are often reluctant to let nursing students or newly hired nurses perform the operations, resulting in limited practice opportunities for nursing students. Secondly, nursing students are resistant to contact with patients' feces or urine, which further limits their learning effect. In addition, due to factors such as nursing staff's unskilled operation, lack of knowledge, and nervousness and fear, it is easy to cause complications in patients and affect their quality of life. These problems not only increase the pain of patients, but also bring greater challenges to the medical team.

[0004] Therefore, an optimized data management solution for teaching stoma care is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a data management system and method for stoma care teaching.

[0006] According to one aspect of the present application, a data management system for stoma care teaching is provided, comprising: A management module, used to store user identity information of nursing students and teachers and store stoma nursing teaching video collections and virtual training tasks; A virtual training module is used to collect and transmit the training operation videos of the target nursing students during the virtual training process; The training analysis module includes: a key frame sampling unit, which is used to perform key frame sampling on the training operation video to obtain a time series of training operation key frames; a key feature extraction unit, which is used to extract key joint bone point position information from the time series of the training operation key frames to obtain a bone point time series feature vector; a key dual-stream feature extraction unit, which is used to perform dual-stream feature extraction on each training operation key frame to obtain a dual-stream feature map of the training operation key frame; a skeleton point-training operation interaction unit, which is used to perform skeleton point-training operation core information significant interaction on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector; an operation error type generation unit, which is used to pass the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector through a classifier-based stoma nursing training feedback generator to obtain a stoma nursing training feedback result; The teaching feedback module is used to generate practical training operation suggestions based on the stoma care practical training feedback results, and to display and store the practical training operation suggestions.

[0007] According to another aspect of the present application, a data management method for stoma care teaching is provided, comprising: Storing user identity information of nursing students and teachers, as well as stoma nursing teaching video collections and virtual training tasks; Collect and transmit the training operation videos of the target nursing students during the virtual training process; Receive the training operation video of the target nursing student object, and perform core information significant interaction analysis based on the training operation features and the skeleton point position features on the training operation video of the target nursing student object to generate stoma nursing training feedback results, including: sampling the key frames of the training operation video to obtain a time series of training operation key frames; extracting the key joint skeleton point position information of the time series of the training operation key frames to obtain a skeleton point time series feature vector; extracting dual-stream features of each training operation key frame to obtain a training operation key frame dual-stream feature map; performing skeleton point-training operation core information significant interaction on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector; passing the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector through a classifier-based stoma nursing training feedback generator to obtain a stoma nursing training feedback result; A practical training operation suggestion is generated based on the stoma care practical training feedback result, and the practical training operation suggestion is displayed and stored.

[0008] Compared with the prior art, the data management system and method for stoma nursing teaching provided by the present application first stores the identity information of nursing students and teachers, as well as stoma nursing teaching video collections and virtual training tasks, then collects and transmits the training operation video of the target nursing student object during the virtual training process, and then performs a significant interactive analysis of the core information based on the training operation characteristics and the skeletal point position characteristics on the received training operation video to obtain the stoma nursing training feedback result of the target nursing student object, and finally generates the training operation suggestion based on the obtained stoma nursing training feedback result, and displays and stores it. In this way, not only can the teaching effect be improved and the cost be reduced, but also the risks in the practice of nursing students are avoided and the safety of teaching is enhanced. Nursing students can use virtual training to improve the proficiency and accuracy of stoma nursing operations, thereby better serving patients and improving the trust and satisfaction of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0010] In the attached picture:

[0011] Figure 1 This is a system block diagram of a data management system for ostomy care teaching according to an embodiment of the present application.

[0012] Figure 2 It is a block diagram of the practical training analysis module in the data management system for ostomy care teaching according to an embodiment of the present application.

[0013] Figure 3 It is a block diagram of a skeleton point-practical training operation interaction unit in a data management system for ostomy care teaching according to an embodiment of the present application.

[0014] Figure 4 The present invention is a flowchart of a data management method for stoma care teaching according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] A stoma is an artificial opening created on the abdominal wall to treat intestinal or urinary system diseases. It allows a section of the intestine or urethra to be led out of the body to discharge excrement. As an important part of medical care, stoma care covers complex care processes and key points of operation.

[0017] However, the traditional teaching method mainly relies on teachers' lectures and nursing students' actual operation, which has many limitations. First, the teaching cost is high, and patients are usually reluctant to let nursing students or inexperienced nursing staff perform the operation, which limits the practice opportunities of nursing students. Secondly, nursing students often have psychological resistance when they come into contact with patients' feces or urine, which further affects the learning effect. In addition, due to lack of experience, insufficient knowledge reserves or nervousness during operation, nursing staff may cause complications in patients, thereby reducing their quality of life. These problems not only bring additional pain to patients, but also increase the difficulty and challenges of the work of the medical team.

[0018] Based on this, this application proposes a data management system for stoma nursing teaching, which improves the learning interest, confidence and practical ability of nursing students and improves the level of handling complications by simulating the real clinical environment and nursing process. Figure 1 1 is a system block diagram of a data management system for stoma care teaching according to an embodiment of the present application. Figure 1 As shown, in the data management system 100 for stoma nursing teaching, it includes: a management module 110, which is used to store user identity information of nursing students and teachers and store stoma nursing teaching video sets and virtual training tasks; a virtual training module 120, which is used to collect and transmit the training operation video of the target nursing student object during the virtual training process; a training analysis module 130, which is used to receive the training operation video of the target nursing student object, and perform core information significant interaction analysis of the training operation video of the target nursing student object based on the training operation characteristics and the skeleton point position characteristics to generate stoma nursing training feedback results; a teaching feedback module 140, which is used to generate training operation suggestions based on the stoma nursing training feedback results, and display and store the training operation suggestions.

[0019] In the embodiment of the present application, the management module 110 is used to store the user identity information of nursing students and teachers and to store stoma nursing teaching video collections and virtual training tasks. In particular, the virtual training tasks are stoma nursing processes and how to deal with stoma complications. It should be understood that accurately identifying the identities of nursing students and teachers is the basis for managing teaching activities. By storing the user identity information of nursing students and teachers, the system can distinguish different user roles, thereby providing a basis for authority allocation, teaching activity tracking, etc. In addition, the system also needs to store teaching resources such as stoma nursing teaching video collections and virtual training tasks to achieve centralized management of teaching resources. Specifically, the stoma nursing teaching video collection contains rich teaching content, and these resources are centrally stored in the system for teachers and nursing students to access and use. Teachers can select appropriate videos for teaching arrangements according to teaching needs, and nursing students can also learn and practice independently. The virtual training tasks include stoma nursing processes and the process of dealing with stoma complications, which can provide unified standards and specifications for teaching, help ensure that all nursing students can receive consistent teaching content, and avoid uneven teaching quality due to differences in teachers' teaching methods. It should be noted that the use of virtual teaching can break through the limitations of patient resources, time and space in traditional teaching. Nursing students can perform virtual training operations anytime and anywhere, and practice repeatedly until they are proficient. It can also avoid the psychological burden brought by direct contact with patient excrement and real complication scenarios, allowing nursing students to focus on learning and improving operational skills in a relaxed and stress-free environment and overcome psychological barriers. Most importantly, operational errors in a virtual environment will not cause harm to real patients. Nursing students can boldly try different operating methods, learn and grow from mistakes, and accumulate experience.

[0020] In the embodiment of the present application, the virtual training module 120 is used to collect and transmit the training operation video of the target nursing student during the virtual training process. Specifically, in the embodiment of the present application, the virtual training module is used to: simulate the stoma care process based on the virtual training task, and collect the training operation video of the target nursing student during the virtual training process; use the WebSocket or MQTT protocol to transmit the training operation video of the target nursing student.

[0021] Accordingly, considering that stoma care is a skill that requires extremely high practical operation, its operation process is complex and contains many details. By simulating real clinical scenarios, nursing students can repeatedly practice complex stoma care processes in a safe and controllable environment. Specifically, in an example of the present application, based on the virtual training task to simulate the stoma care process, a realistic clinical scene can be constructed using three-dimensional graphics real-time rendering technology, including elements such as the patient's abdominal model, stoma location, and required medical equipment, so that nursing students can perform nursing operations including cleaning wounds and changing stoma bags in this simulated clinical scene, so that nursing students can master nursing skills more proficiently and improve the accuracy and proficiency of nursing operations. At the same time, by recording the operation video of the nursing students, detailed training data can be provided for subsequent analysis and evaluation, which can help identify problems and shortcomings of the nursing students during the operation process.

[0022] Afterwards, the practical training operation video of the target nursing student object can be transmitted by using the WebSocket or MQTT protocol. It should be understood that WebSocket provides a full-duplex communication channel that allows real-time data exchange between the server and the client. The transmission of the practical training operation video of the target nursing student object requires as low a delay as possible to ensure that the nursing student's operation can be analyzed and fed back in time. The use of WebSocket for transmission can ensure the smoothness and real-time performance of data transmission. Unlike traditional HTTP connections, WebSocket connections remain persistently open after they are established until one of the parties actively closes them. This avoids the overhead and instability caused by frequent establishment and disconnection of connections, and is particularly suitable for long-term video transmission processes. It can ensure that video data is continuously and stably transmitted, and reduce video freezes or losses caused by connection interruptions. MQTT is a lightweight message transmission protocol that uses binary format for message encoding, and the protocol header information is very concise, and the bandwidth and resource consumption required for message transmission are relatively small. When transmitting the practical training operation video of the target nursing student object, especially in an environment with limited network bandwidth, MQTT can achieve efficient transmission of video data with low overhead and reduce the occupation of network resources. For mobile device applications, such as nursing students using tablets or smartphones to perform practical training and transmit videos, the low power consumption of MQTT is extremely advantageous. It can reduce the power consumption of the device, extend the battery life of the device, and ensure that a longer period of practical training and video transmission tasks can be completed on a single charge.

[0023] In an embodiment of the present application, the training analysis module 130 is used to receive the training operation video of the target nursing student object, and to generate stoma nursing training feedback results by performing significant interactive analysis of the core information based on the training operation characteristics and the skeletal point position characteristics on the training operation video of the target nursing student object. It should be understood that the training operation video can not only show the learning achievements of the nursing students, but also expose the problems in the teaching process. Therefore, based on the data analysis of actual operations, it can be found that there are deficiencies in which specific links of nursing, so as to formulate more targeted improvement suggestions. However, the traditional operation error judgment depends on the subjective judgment and experience accumulation of teachers, and teachers need to spend a lot of time watching and analyzing the operation video or on-site demonstration of each nursing student. This method is not only easy to miss key operation errors, but also leads to feedback lag, low efficiency, and it is difficult to provide effective improvement suggestions in time, which affects the learning effect and skill improvement of nursing students.

[0024] Based on this, the technical concept of this application is to use artificial intelligence-based video analysis and encoding methods to sample key frames of the practical operation video, and then perform skeleton point timing feature analysis and dual-stream feature extraction on each sampled practical operation key frame, so as to intelligently judge the type of practical operation error of the target nursing student based on the significant interaction between the skeleton point timing feature and the dual-stream feature of the practical operation key frame. This concept can capture the specific action details and posture changes of the nursing student in each operation step, and immediately generate detailed feedback results, while reducing the burden on teachers and improving overall teaching efficiency.

[0025] Specifically, Figure 2 FIG. 1 is a block diagram of a training analysis module in a data management system for stoma care teaching according to an embodiment of the present application. Figure 2As shown, the training analysis module 130 includes: a key frame sampling unit 131, which is used to perform key frame sampling on the training operation video to obtain a time series of training operation key frames; a key feature extraction unit 132, which is used to extract key joint skeleton point position information from the time series of the training operation key frames to obtain a skeleton point time series feature vector; a key dual-stream feature extraction unit 133, which is used to perform dual-stream feature extraction on each training operation key frame in the time series of the training operation key frames to obtain a dual-stream feature map of the training operation key frames; a skeleton point-training operation exchange unit 134, which is used to extract the key joint skeleton point position information from the time series of the training operation key frames to obtain a skeleton point time series feature vector; a key dual-stream feature extraction unit 133, which is used to extract the key joint skeleton point position information from the time series of the training operation key frames to obtain a skeleton point time series feature vector; a skeleton point-training operation exchange unit 134 ... The interaction unit 134 is used to perform a significant interaction of the skeleton point-training operation core information on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector; the operation error type generation unit 135 is used to pass the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector through a classifier-based stoma nursing training feedback generator to obtain the stoma nursing training feedback result, and the stoma nursing training feedback result is used to represent the training operation error type label of the target nursing student.

[0026] In the embodiment of the present application, the key frame sampling unit 131 is used to perform key frame sampling on the practical training operation video to obtain the time sequence of the key frames of the practical training operation. Accordingly, considering that the practical training operation video of the target nursing student object contains a large amount of data information, which contains the key moments or actions in the operation, but also has some background or irrelevant image frame information. Based on this, in the technical solution of the present application, the practical training operation video is sampled with key frames so that representative key frames can be selected from continuous video frames to obtain the time sequence of the key frames of the practical training operation. In particular, these key frames contain the most important information and key actions in the video, and can capture the key actions, step conversions and important events in the nursing student operation process, and the formed time sequence can present these key information in the order of the operation. In this way, the amount of data to be processed can be greatly reduced without losing the core information, providing a clear context for subsequent analysis, helping to accurately judge whether the nursing student's operation process is correct, whether the operation method is standardized, etc.

[0027] In the embodiment of the present application, the key feature extraction unit 132 is used to extract the key joint skeletal point position information of the time series of the training operation key frames to obtain the skeletal point timing feature vector. Specifically, in the embodiment of the present application, the key feature extraction unit is used to: extract the skeletal point position information of the key joints of the target nursing object from each training operation key frame in the time series of the training operation key frames to obtain the time series of the training operation skeletal point position; pass the time series of the training operation skeletal point position through a skeletal point position timing feature extractor based on a one-dimensional causal convolutional neural network model to obtain the skeletal point timing feature vector.

[0028] Accordingly, considering that in stoma care operation, the accuracy and standardization of the movements of key joints such as the hands and arms of the nursing students are crucial to the nursing effect. Compared with simply observing the video screen, the skeletal point position information is more accurate and objective, and is not disturbed by factors such as the complex background or light changes of the video screen, and can directly reflect the core characteristics of the nursing student's movements. Based on this, in the technical solution of the present application, the skeletal point position information of the key joints of the target nursing student object is extracted from each training operation key frame in the time series of the training operation key frame to obtain the time series of the training operation skeletal point position. That is, the skeletal point position information can reflect the body posture and movement trajectory of the nursing student, which is crucial for evaluating whether its operation is standardized and the type of error. By extracting the skeletal point position information of the key joints, the movement trajectory and position change of the key joints of the nursing student during the stoma care operation can be intuitively presented to accurately find problems in the operation process.

[0029] The following is a detailed description of a specific implementation process of "extracting the skeleton point position information of the key joints of the target nursing object from each training operation key frame in the time sequence of the training operation key frame to obtain the time sequence of the training operation skeleton point position": First of all, preprocessing the key frames of the training operation is an indispensable link. Since the original key frame image may have various problems that affect the accurate extraction of skeleton points, targeted processing is required. On the one hand, image enhancement is crucial. By adjusting the brightness, contrast and color balance of the key frame image, the image can be made clearer. For example, the histogram equalization method can enhance the contrast of the image, making the features of the nursing objects in the key frame more obvious, which is helpful for the subsequent recognition of human features. On the other hand, denoising is also a key step. The noise in the image may come from the quality of the shooting equipment, interference during the transmission process, etc. The use of algorithms such as Gaussian filtering or median filtering can effectively remove these noises, avoid their interference with skeleton point extraction, and improve the accuracy of subsequent processing.

[0030] After completing the keyframe preprocessing, the next step is to accurately detect and locate the target nursing objects in these processed keyframes. To achieve this goal, advanced target detection algorithms are needed. For example, algorithms such as the YOLO series and Faster R-CNN are fast and accurate, and can quickly identify the location of nursing objects in keyframes and generate bounding boxes containing nursing objects. Considering that nursing objects may be at different distances and angles in different keyframes, single-scale detection may not meet the needs, so a multi-scale detection method should be adopted. This method can ensure that nursing objects can be accurately detected in various complex situations, and determine the accurate range for subsequent skeleton point extraction.

[0031] After successfully locating the nursing object, it is necessary to select a suitable skeleton point extraction model and apply it to the located area. There are currently a variety of mature skeleton point extraction models to choose from, such as OpenPose and HRNet. These models have been trained and optimized with a large amount of data and can accurately identify multiple key joints of the human body, such as shoulders, elbows, wrists, hips, knees, and ankles. If the stoma care training scenario has specific requirements or the data has unique characteristics, the selected model can also be fine-tuned or retrained. Using a large amount of image data containing nursing operations for training can make the model better adapt to the stoma care training scenario and improve the accuracy of skeleton point extraction. Input the located nursing object area into the selected skeleton point extraction model, and the model will output the coordinate position of the skeleton point of each key joint in the image.

[0032] After obtaining the skeleton point position information of the key joints in each key frame, it is necessary to organize this information to construct the time series of the skeleton point position of the practical training operation. First, the skeleton point position information extracted from each key frame should be arranged in sequence according to the time sequence of the key frames of the practical training operation. This can ensure the continuity and accuracy of the time series and can truly reflect the changes in the movements of the nursing students during the entire practical training operation. Then, the organized skeleton point position information is stored. At the same time, the key frame timestamp corresponding to each skeleton point should be recorded, which is convenient for subsequent analysis and processing of the data, and can more clearly understand the process and rhythm of the nursing students' operation based on the time information.

[0033] Finally, in order to ensure the accuracy and reliability of the time series of the skeleton point positions obtained in the training operation, data verification and correction are required. For example, statistical methods are used to analyze the skeleton point position information. Through statistical analysis, outliers and discontinuities can be detected. For these abnormal data, they can be corrected or re-extracted in combination with the information of adjacent key frames, so that the time series data of the skeleton point positions of the entire training operation is more accurate and reliable.

[0034] It should be understood that the time series data of the position of the skeleton points in the practical training operation has dynamic and temporal characteristics, and it contains rich information about the changes of the skeleton points of the key joints of the nursing students during the stoma care operation process over time. The one-dimensional causal convolutional neural network model can effectively capture the local features and long-term dependencies in this time series data. Therefore, the present application effectively captures these time dependencies by passing the time series of the skeleton point positions of the practical training operation through the skeleton point position temporal feature extractor based on the one-dimensional causal convolutional neural network model, helps to understand the changes in the movements of the nursing students at different time points, and obtains the skeleton point temporal feature vector. In particular, the characteristics of causal convolution ensure that when analyzing the data at the current time point, only the data of the past and current moments are relied on, which conforms to the temporal order logic of the operation and avoids the interference of future information on the current analysis. Through this model, more representative and distinguishing features can be extracted from the complex skeleton point position time series. Compared with the traditional feature extraction method, the potential patterns and laws in the data can be more deeply excavated to more accurately reflect the operation level of the nursing students.

[0035] In an embodiment of the present application, the key dual-stream feature extraction unit 133 is used to perform dual-stream feature extraction on each training operation key frame in the time series of the training operation key frame to obtain a dual-stream feature map of the training operation key frame. Specifically, in an embodiment of the present application, the key dual-stream feature extraction unit is used to: use a dual-stream feature extractor based on EfficientNet and FlowNet to extract each training operation key frame in the time series of the training operation key frame to obtain a dual-stream feature map of the training operation key frame. Accordingly, considering that the stoma care operation contains rich visual information, it is difficult to fully capture the key information by relying solely on a single feature extraction method. Based on this, the present application performs dual-stream feature extraction on each training operation key frame in the time series of the training operation key frame to obtain a dual-stream feature map of the training operation key frame. In particular, the dual-stream feature extraction analyzes the key frames from two dimensions, space and time. Specifically, the spatial stream can extract static image features in the keyframes, such as the posture of the nursing student, the position of the stoma care tools, etc.; the time stream focuses on capturing the dynamic change features between adjacent keyframes, such as the sequence of the nursing student's actions, the change of the action speed, etc. The feature extraction of these two dimensions can more comprehensively and meticulously describe the behavior of the nursing student in the stoma care operation process, and provide a richer and more accurate data basis for subsequent analysis. In particular, in a specific example of the present application, a dual-stream feature extractor based on EfficientNet and FlowNet is used to extract each training operation keyframe in the time series of the training operation keyframe to obtain the dual-stream feature map of the training operation keyframe. It can be understood that EfficientNet shows excellent performance in tasks such as image classification. By optimizing the network structure and parameter configuration, it can extract rich image spatial features with less computing resource consumption. In the stoma care training key frame analysis, the posture of the nursing student during operation, the spatial relationship between the nursing tools and the stoma site, and other details can be accurately captured. FlowNet is specialized in processing optical flow information and is good at extracting motion features from continuous image frames, that is, the changes in the nursing students' operation actions in the time dimension, such as the speed, direction and coherence of the action. The combination of the two works simultaneously from the spatial and temporal dimensions to comprehensively and deeply mine the features in the key frames, avoiding the limitations of extracting information from a single model, and thus more accurately understanding the nursing students' operation behaviors.

[0036] In the embodiment of the present application, the skeleton point-training operation interaction unit 134 is used to perform a skeleton point-training operation core information significant interaction on the skeleton point temporal feature vector and the training operation key frame dual stream feature map to obtain a skeleton point-training operation dual stream semantic cross-modal significant interaction encoding vector. Specifically, Figure 3FIG. 1 is a block diagram of a skeleton point-practice operation interaction unit in a data management system for stoma care teaching according to an embodiment of the present application. Figure 3 As shown, the skeleton point-practical training operation interaction unit 134 includes: a skeleton point temporal core clue extraction subunit 1341, which is used to perform point convolution core clue extraction on the skeleton point temporal feature vector to obtain a skeleton point temporal core clue encoding vector; a training operation key frame core clue extraction subunit 1342, which is used to extract the training operation key frame dual-stream feature core clue encoding vector from the training operation key frame dual-stream feature map; a skeleton point-practical training operation core clue encoding subunit 1343, which is used to construct a skeleton point-practical training operation core clue weaving template matrix between the skeleton point temporal core clue encoding vector and the training operation key frame dual-stream feature core clue encoding vector; a training operation key frame feature disassembly subunit 1344, which is used to perform feature extraction along the channel dimension on the training operation key frame dual-stream feature map. Disassemble to obtain a set of local two-stream feature matrices of training operation key frames; the skeleton point-training operation cross-modal encoding subunit 1345 is used to use the skeleton point temporal feature vector as the query vector, each training operation key frame local two-stream feature matrix in the set of training operation key frame local two-stream feature matrices as the key matrix and the skeleton point-training operation core clue weaving template matrix as the prior information constraint matrix, and perform cross-modal template constraint heterogeneous transformation on it to obtain a set of skeleton point-training operation cross-modal fine-grained interaction coding vectors; the skeleton point-training operation significant interaction feature generation subunit 1346 is used to calculate the positional mean vector of the set of skeleton point-training operation cross-modal fine-grained interaction coding vectors to obtain the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector.

[0037] It should be understood that the skeleton point time series feature vector mainly reflects the change of the position of the skeleton point of the key joint of the nursing student over time, focusing on the temporal dynamics of the action; the dual-stream feature map of the key frame of the practical operation comprehensively captures the static and dynamic characteristics of the operation scene from the spatial and temporal dimensions, including rich information on the environment, tool use and action changes. Therefore, in order to integrate the advantages of the two, explore the potential connection between them, and comprehensively and deeply reveal the inherent laws and essential characteristics of the nursing student's operation behavior to avoid information omission, this application performs a significant interaction of the skeleton point-practical operation core information on the skeleton point time series feature vector and the dual-stream feature map of the key frame of the practical operation to obtain the skeleton point-practical operation dual-stream semantic cross-modal significant interaction encoding vector. Specifically, this method takes the two as core elements and constructs a special interaction mode and fusion mechanism. With the help of attention mechanism, appropriate conversion algorithm and targeted constraint learning means, the semantic association and complementary characteristics of these two different modal features are deeply excavated and integrated, so as to improve the accuracy and effectiveness of the evaluation of the practical operation of stoma nursing for nursing students. For example, in stoma care, it can not only evaluate whether the hand movements when changing the ostomy bag are standard, but also comprehensively consider the coordination with the stoma position and the use of care tools in the entire operation scenario, so as to give a more objective and accurate operation evaluation.

[0038] Specifically, it is first necessary to perform point convolution core clue extraction on the skeleton point temporal feature vector to obtain the skeleton point temporal core clue encoding vector. The above process can be expressed by the formula:

[0039] in, is the temporal feature vector of the skeleton point, is the point convolutional coding, is the bias vector, is the activation function, It is the skeleton point temporal core cue encoding vector.

[0040] It should be understood that the skeleton point time series feature vector contains a lot of information about the changes in the position of the key joint skeleton points over time during the stoma care training operation of the nursing students. This information may be complex and scattered. Through the point convolution core clue extraction, this information can be refined and integrated. That is, the generated skeleton point time series core clue encoding vector is an abstract and highly condensed representation of the original features. It removes unnecessary details and retains only the core information that best represents the nursing students' operation characteristics. In particular, the stoma care operation of the nursing students involves complex movements of multiple joints at different times, and these movements are presented as high-dimensional information in the skeleton point time series feature vector. The point convolution operation can ensure that these high-dimensional information is not lost in the process of extracting core clues. For example, when changing the stoma bag operation, the information of the coordinated movements of the joints of the hand and the arms, shoulders, etc. is still fully reflected in the skeleton point time series core clue encoding vector after point convolution processing. This ensures that the correctness and standardization of the operation can be comprehensively analyzed from multiple dimensions during the subsequent evaluation, avoiding evaluation bias caused by information loss.

[0041] Next, the training operation key frame dual stream feature core clue encoding vector is extracted from the training operation key frame dual stream feature map. The above process can be expressed by the formula:

[0042] in, is the dual-stream feature map of the training operation key frame, For meshing operations , , ,and They are the first, second, and third local dual-stream feature maps of the training operation key frame. and The local two-stream feature map of the training operation key frame, is a convolutional code with a convolution kernel of 3×3. yes function, It is the first in the set of local two-stream enhanced feature maps of the training operation key frame. The local two-stream enhanced feature map of the training operation key frame, is a global pooling operation, It is the first in the set of local two-stream enhanced feature vectors of the training operation key frame. The local two-stream enhanced feature vector of the training operation key frame, and are the maximum and minimum values, respectively. , and is the modulation parameter, and They are The mean and variance of The first set of local dual-stream significant coefficients of the training operation key frame The local two-stream significant coefficients of the training operation key frames, Indicated by natural constant The exponential function value with base , is the number of vectors in the set of local dual-stream enhanced feature vectors of the training operation key frame, It is the core clue encoding vector of the dual-stream feature of the training operation key frame.

[0043] It should be understood that the dual-stream feature map of the key frame of the practical operation contains rich spatial information, and the regional attention mechanism is like a "focusing magnifying glass" that can identify and highlight the significant areas in the feature map. In the stoma care operation scenario, these significant areas may be the stoma site, the contact point between the nursing tools and the stoma, and the key action area of ​​the nursing students' hands. Through the re-weighting of attention, the information of these key areas is given priority. When extracting the core clue encoding vector of the dual-stream feature of the key frame of the practical operation, it can ensure that the spatial information that plays a key role in the stoma care operation is not ignored, so that the evaluation can focus more on important operation links. The processing method of this step pays attention to both spatial information and semantic features. Specifically, in stoma care, spatial information (such as the position of nursing tools around the stoma, the relative position of nursing students' limbs and the patient's body) is equally important as semantic features (such as the nursing intention and standard process represented by the operation action). The regional attention mechanism enables spatial information to be highlighted. At the same time, based on the local receptive field characteristics of the convolution operation, when processing the significant areas, it can also mine and retain semantic features closely related to these spatial information. For example, when a nursing student changes an ostomy bag, they can not only retain the spatial information of the position of the bag and the stoma, but also understand the semantics of "sealing to prevent leakage" represented by this operation, ensuring a comprehensive understanding and evaluation of the operation behavior from multiple perspectives.

[0044] Specifically, in an embodiment of the present application, the skeleton point-training operation core clue encoding subunit includes: a core association matrix construction secondary subunit, which is used to use a feature mapping function to process the skeleton point temporal core clue encoding vector and the training operation key frame dual-stream feature core clue encoding vector to obtain the skeleton point-training operation core association matrix; a core association matrix global constraint secondary subunit, which is used to perform feature processing on the skeleton point-training operation core association matrix based on global semantic constraints to obtain the skeleton point-training operation core clue weaving template matrix.

[0045] Next, the skeleton point temporal core clue encoding vector and the training operation key frame dual stream feature core clue encoding vector are processed using a feature mapping function to obtain a skeleton point-training operation core association matrix. The above process can be expressed as:

[0046] in, is the skeleton point temporal core cue encoding vector, is the core clue encoding vector of the dual-stream feature of the training operation key frame, is the transpose operation, and is a feature mapping function, such as a linear mapping or a nonlinear kernel function, It is the skeleton point-practical operation core association matrix.

[0047] It should be understood that the skeleton point temporal core clue encoding vector mainly reflects the time series characteristics of the key joint movements during the nursing student's operation, while the training operation key frame dual-stream feature core clue encoding vector focuses on the spatial and semantic features in the operation scene, and the two belong to information of different modalities. The constructed skeleton point-practical operation core association matrix can closely link the information of these two different modalities, which essentially reflects the global semantic constraints between different modal information. For example, in stoma care operations, global semantic constraints are reflected in the standardization and coherence of the entire operation process. For example, from preparing nursing tools, cleaning stomas, to changing stoma bags and fixing them, each link has certain norms and logical orders in time and space. The skeleton point-practical operation core association matrix can link the semantics of these different links in terms of action timing and operation scenes, evaluate whether the nursing student's operation conforms to the overall nursing specifications, and ensure the integrity and correctness of the operation. In addition to reflecting global semantic constraints, the matrix can also capture fine-grained relationships between local salient regions. In stoma care operations, local significant areas include the skin around the stoma, the contact points between the care tools and the stoma, and the key movement parts of the nursing students' hands. The matrix can accurately measure the subtle correlation between the action timing and spatial characteristics of these local areas. For example, when cleaning the skin around the stoma, the change in the strength and angle of the nursing students' hand movements over time and the corresponding relationship with different areas of the skin around the stoma can be clearly presented through the calculation of the matrix, thereby helping the system to more carefully evaluate whether the nursing students' operations are standardized and improve the accuracy of the evaluation.

[0048] More specifically, in an embodiment of the present application, the core association matrix global constraint secondary subunit is used to: utilize the residual mechanism and the feature mapping function to perform preliminary structured semantic alignment on the skeleton point temporal core clue encoding vector and the training operation key frame dual-stream feature core clue encoding vector to obtain the skeleton point-training operation semantic alignment vector and the training operation-skeleton point semantic alignment vector; calculate the skeleton point-training operation explicit global semantic correlation matrix between the skeleton point-training operation semantic alignment vector and the training operation-skeleton point semantic alignment vector; based on the skeleton point-training operation explicit global semantic correlation matrix, perform significant fine-grained compensation on the skeleton point-training operation core association matrix to obtain the skeleton point-training operation core clue weaving template matrix. This process can be expressed by the formula:

[0049] in, and They are the skeleton point-training operation semantic alignment vector and the training operation-skeleton point semantic alignment vector. To calculate the square of the vector's two-norm, is the logarithmic function value with base 2, is matrix multiplication, is the skeleton point-training operation explicit global semantic relevance matrix, It is the skeleton point-practice operation core association matrix, It is the skeleton point-practical training operation core clue weaving template matrix.

[0050] In particular, for the skeleton point temporal feature vector used for the core clue representation of different modality features And the training operation key frame dual stream feature core clue encoding vector , through its residual mechanism based on heterogeneous modality mapping, it performs preliminary structured semantic alignment of common explicit associations in the shared representation space to obtain the skeleton point-trained operational semantic alignment vector And practical operation-skeletal point semantic alignment vector Then, based on the global semantic spatialization representation of the graph structure under explicit modeling of mapping preference, the residual mechanism is further used to capture the "missing" information in the alignment process under different mapping modalities, and adaptive compensation is performed through explicit association to further capture the fine-grained significant difference information between modalities in the initial alignment process. In this way, the contradiction between the alignment complexity of the intrinsic driving mapping of cross-modal differences and the intuitiveness of the explicit association of the graph structure can be optimized, and the illusion of global semantic constraints between modalities under the condition of misalignment of fine-grained relationships between local salient regions can be reduced.

[0051] Then, the dual-stream feature map of the training operation key frame is decomposed along the channel dimension to obtain a set of local dual-stream feature matrices of the training operation key frame. The above process can be expressed by the formula:

[0052] in, is the dual-stream feature map of the training operation key frame, is the feature decomposition operation along the channel dimension, , , and It is the first, second, and third in the set of local two-stream feature matrices of the training operation key frame. and The local two-stream feature matrix of the training operation key frame.

[0053] It should be understood that the dual-stream feature map of the key frame of the practical operation integrates the rich spatial and temporal information of the stoma care operation scene, covering multiple aspects such as the environment, nursing student movements, and the use of nursing tools. The features are disassembled along the channel dimension, so that the model can analyze the operation features more finely. Taking the stoma cleaning operation as an example, the features related to the change of the cleaning action strength, the features of the relative position of the cleaning tool and the stoma, and the features of the movement trajectory of the nursing student's arm can be extracted separately. This disassembly allows the model to have a deeper understanding of the details of each local feature, avoiding the neglect of key information due to the mixing of overall information, so as to accurately grasp the subtleties of the nursing student's operation. The generated set of local dual-stream feature matrices of the key frames of the practical operation can also provide an ideal unitized modeling input for subsequent cross-modal interactive operations, ensuring that the content between complex modalities can be captured more finely. For example, when the nursing student is pasting the stoma bag, the bone point temporal features of the hand joint action can be accurately associated with the local dual-stream feature matrix related to the hand action posture, which is conducive to improving the quality of cross-modal information fusion and enabling the model to more accurately analyze the nursing student's operation behavior.

[0054] Next, the skeleton point temporal feature vector is used as the query vector, each training operation key frame local two-stream feature matrix in the set of training operation key frame local two-stream feature matrices is used as the key matrix, and the skeleton point-training operation core clue weaving template matrix is ​​used as the prior information constraint matrix, and a cross-modal template constraint heterogeneous transformation is performed on it to obtain a set of skeleton point-training operation cross-modal fine-grained interaction encoding vectors. The above process can be expressed by the formula:

[0055] in, is matrix multiplication, It is the skeleton point-practice operation core clue weaving template matrix, It is the set of local two-stream feature matrices of the training operation key frame. The local two-stream feature matrix of the training operation key frame, yes The scale is the matrix width multiplied by the matrix height, is the normalization function, is the first in the set of skeleton point-training operation cross-modal fine-grained interaction encoding vectors Skeleton point-trained operation cross-modal fine-grained interaction encoding vector.

[0056] It should be understood that the skeleton point time series feature vector is used as the query vector to obtain the response from the local two-stream feature matrix set of the training operation key frame. This process realizes the deep fusion of two different modal information. For example, in the stoma care operation, the skeleton point time series feature vector reflects the movement changes of key joints such as the hands of the nursing students during the operation, while the local two-stream feature matrix set of the training operation key frame contains the spatial and motion information of different local areas in the operation scene. In this way, the action time series information of the nursing students is closely combined with the spatial information of the operation scene, which can comprehensively and meticulously present the synergistic relationship of various aspects of information in the operation process, avoiding the one-sidedness of information caused by single modal analysis. At the same time, different from the native Transformer, the introduction of the skeleton point-practical operation core clue weaving template matrix as a priori information constraint matrix can set norms and standards for the interaction of different modal information, and ensure the standardization of the evaluation in the evaluation of the stoma care process. That is, the skeleton point-practical operation cross-modal fine-grained interaction encoding vector set obtained by cross-modal template constraint heterogeneous transformation has stronger representation ability. Each skeleton point-practice operation cross-modal fine-grained interaction encoding vector integrates information from both modalities, including not only the temporal changes of hand movements, but also the spatial and semantic features of specific areas in the operation scene. This enables the essential characteristics of nursing students' operation behaviors to be reflected more deeply and accurately in subsequent operation evaluations. Compared with single modality information, it can provide a richer and more comprehensive basis for evaluation, thus improving the value and reliability of the evaluation results.

[0057] Finally, the position mean vector of the set of the skeleton point-training operation cross-modal fine-grained interaction encoding vectors is calculated to obtain the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector. The above process can be expressed by the formula:

[0058] in, is the first in the set of skeleton point-training operation cross-modal fine-grained interaction encoding vectors Skeleton point-training operation cross-modal fine-grained interaction encoding vector, is the number of vectors in the set of the skeleton point-training operation cross-modal fine-grained interaction encoding vectors, It is the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector.

[0059] It should be understood that the set of skeleton point-practical operation cross-modal fine-grained interaction encoding vectors contains a large amount of data, and directly processing this data will bring a high computational burden to the system. By calculating the position mean vector, many fine-grained interaction encoding vectors are condensed, which greatly reduces the amount of data. It should be noted that although data compression is performed, the method of calculating the position mean ensures that the main semantic information is retained. Each skeleton point-practical operation cross-modal fine-grained interaction encoding vector reflects different aspects of ostomy care operations, such as the association between nursing students' hand movements and specific areas of the operation scene. When calculating the mean vector, these key information are integrated together to form a comprehensive and more representative encoding vector. That is, the skeleton point-practical operation dual-stream semantic cross-modal significant interaction encoding vector finally generated integrates a variety of operation information and reflects the performance of nursing students in ostomy care operations as a whole.

[0060] In an embodiment of the present application, the operation error type generating unit 135 is used to pass the skeleton point-practical operation dual stream semantic cross-modal significant interaction encoding vector through a classifier-based stoma nursing training feedback generator to obtain the stoma nursing training feedback result, and the stoma nursing training feedback result is used to represent the training operation error type label of the target nursing student object. That is, the skeleton point-practical operation dual stream semantic cross-modal significant interaction encoding vector obtained by semantically significant interaction with the skeleton point time series feature vector and the training operation key frame dual stream feature map is classified and processed to intelligently judge the training operation error type of the target nursing student object. It should be understood that the classifier has powerful pattern recognition and classification capabilities, and can parse and classify such complex encoding vectors. Through pre-training, the classifier can learn the corresponding relationship between different encoding vector features and various training operation error types, thereby converting the encoding vector into an easy-to-understand training operation error type label. In this way, the obtained stoma nursing training feedback results are presented as operation error type labels, which can clearly and explicitly point out the types of problems that the target nursing students have in the practical training operations, and provide highly targeted teaching guidance for the follow-up. In particular, in a specific embodiment of the present application, the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector is passed through a classifier-based stoma nursing training feedback generator to obtain the stoma nursing training feedback result, and the stoma nursing training feedback result is used to represent the training operation error type label of the target nursing student object, including: using the fully connected layer of the classifier to fully connect the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector to obtain the skeleton point-training operation dual-stream semantic cross-modal significant interaction fully connected encoding feature vector; the skeleton point-training operation dual-stream semantic cross-modal significant interaction fully connected encoding feature vector is input into the Softmax classification function of the classifier to obtain the stoma nursing training feedback result.

[0061] In summary, the training analysis module 130 is clearly explained, which uses artificial intelligence-based video analysis and encoding methods to perform key frame sampling on the training operation video of the target nursing student object, and then performs skeleton point timing feature analysis and dual-stream feature extraction on each sampled training operation key frame, so as to intelligently judge the training operation error type of the target nursing student object based on the significant interaction between the skeleton point timing feature and the dual-stream feature of the training operation key frame. In this way, the specific action details and posture changes of the nursing student in each operation step can be captured, and detailed feedback results can be generated immediately, which is conducive to reducing the burden on teachers and improving overall teaching efficiency.

[0062] In an embodiment of the present application, the teaching feedback module 140 is used to generate training operation suggestions based on the stoma care training feedback results, and display and store the training operation suggestions. Specifically, in an embodiment of the present application, the teaching feedback module is used to: generate the training operation suggestions based on the stoma care training feedback results, and retrieve the stoma care teaching video corresponding to the stoma care training feedback results from the stoma care teaching video set; display the training operation suggestions and the stoma care teaching video on the virtual training interface of the target nursing student, and store the stoma care training feedback results and the training operation suggestions in a database.

[0063] It should be understandable that the feedback results of stoma nursing training clearly pointed out the problems that nursing students had in the operation process, but with only problem feedback, nursing students may not know how to improve. Therefore, it is necessary to generate actionable suggestions based on these results to provide nursing students with a clear direction for improvement. At the same time, textual suggestions may not be intuitive enough, and the teaching videos in the stoma nursing teaching video set can show the correct operation methods and processes in an intuitive and dynamic way. Retrieving the teaching videos corresponding to the feedback results can enable nursing students to understand and learn the correct operations more clearly, combine theory with practice, and deepen their understanding and mastery of knowledge.

[0064] The following is a detailed description of a specific implementation process of "generating practical training operation suggestions based on the feedback results of the stoma nursing practical training": The system first receives feedback on stoma care training from the training analysis module. These feedback results are presented in the form of labels of operation errors, which provide directions for the subsequent generation of suggestions. Next, the system will conduct in-depth analysis of these labels, extract key information, and accurately determine the specific category to which the error belongs, such as whether it is an operation error, an operation sequence error, or improper use of nursing tools.

[0065] In order to give targeted, accurate and effective suggestions, a large and comprehensive stoma care knowledge graph is built inside the system. This knowledge graph covers a wide range of content, including stoma care operation specifications, complete nursing processes, various common error cases and corresponding solutions. Once the system parses the error type, it will quickly match and search in the knowledge graph based on the error type. For example, once it is detected that a nursing student makes an operational error in changing the stoma bag, the system will quickly retrieve the standard process for changing the stoma bag, various error types that may occur in this link, and the corresponding correct operating methods and improvement suggestions.

[0066] After retrieving relevant knowledge, the system will use machine learning algorithms and preset rules to generate practical training operation suggestions. The machine learning algorithm continuously optimizes its own model by learning from a large number of correct and incorrect cases of stoma care, so that it can provide detailed and standardized operation step suggestions for different error types to help nursing students correct errors. For example, for operating method errors, the system can generate correction suggestions accurate to each action detail based on the specific manifestation of the error and the experience of handling a large number of similar cases, guiding nursing students to gradually improve.

[0067] Considering the differences in learning progress, knowledge mastery and operating habits among different nursing students, in order to make the suggestions more targeted and effective, the system will combine the historical training data of nursing students to optimize the generated suggestions in an individualized manner. The system will analyze the past learning trajectories of nursing students, such as which knowledge points are well mastered, which operations are prone to errors, and what form of guidance they are more inclined to accept. If the system finds that a nursing student has a high acceptance of graphical guidance content in past training, then when generating suggestions, it will give priority to providing content with operation diagrams or animation demonstration links to better meet their learning needs.

[0068] In order to allow nursing students to understand and accept suggestions more intuitively and comprehensively, the system will integrate the generated text suggestions with relevant images, videos and other materials. For some operating method errors that are difficult to accurately describe in words, simple text suggestions may not allow nursing students to clearly understand. At this time, the system will match the corresponding teaching video clips or decomposed action pictures and present them to nursing students together with the text suggestions. In this way, nursing students can receive information through multiple senses and understand the correct operating methods more quickly and accurately.

[0069] Before outputting suggestions, the system will conduct a rigorous evaluation and verification of the generated suggestions. On the one hand, the system will use virtual simulation technology to simulate the process of nursing students following the suggestions, and verify the feasibility and effectiveness of the suggestions by observing the results of the simulated operations. If it is found during the simulation process that the suggestions may cause new problems or cannot effectively correct the errors, the system will adjust and optimize the suggestions. On the other hand, the system will refer to the implementation effect data of suggestions under similar error types in history, and conduct a rationality evaluation of the newly generated suggestions to ensure that the suggestions can effectively help nursing students improve their operations.

[0070] In particular, when nursing students receive and review suggestions, the system is not passive, but collects feedback from nursing students in real time. For example, nursing students have questions about the suggestions, whether they understand them, etc. Based on these feedbacks, the system can dynamically adjust the content of the suggestions and further optimize the expression and presentation of the suggestions. If nursing students have questions about a suggestion, the system will further explain it, or provide more relevant cases and materials to improve the practicality and operability of the suggestions.

[0071] It should be understood that by displaying practical training operation suggestions on the virtual training interface of the target nursing students, the nursing students can understand their shortcomings. At the same time, combined with the stoma nursing teaching video displayed on the virtual training interface, it can help nursing students understand the correct operation method more intuitively. This close combination of theory and practice helps to strengthen the nursing students' understanding and memory of relevant operations, thereby strengthening the learning effect of relevant operation skills. And by storing the stoma nursing training feedback results and practical training operation suggestions in the database, it can provide rich data support for teaching research and teaching improvement. Specifically, teachers can understand the common problems and learning difficulties of nursing students by analyzing these data, so as to adjust teaching strategies and content and improve teaching quality. In addition, the stored practical training feedback results and operation suggestions also enable nursing students to review at any time in the subsequent learning process. They can understand their learning trajectory, summarize lessons learned, discover their progress and shortcomings by viewing historical records, and conduct targeted review and intensive training to promote the improvement of autonomous learning ability.

[0072] In summary, a data management system 100 for stoma nursing teaching based on an embodiment of the present application is described, which first stores the identity information of nursing students and teachers, as well as stoma nursing teaching video collections and virtual training tasks, then collects and transmits the training operation video of the target nursing student object during the virtual training process, and then performs a core information significant interactive analysis based on the training operation characteristics and the skeletal point position characteristics on the received training operation video to obtain the stoma nursing training feedback result of the target nursing student object, and finally generates the training operation suggestion based on the obtained stoma nursing training feedback result, and displays and stores it. In this way, not only can the teaching effect be improved and the cost be reduced, but also the risks in the practice of nursing students are avoided and the safety of teaching is enhanced. Nursing students can use virtual training to improve the proficiency and accuracy of stoma nursing operations, thereby better serving patients and improving patients' trust and satisfaction.

[0073] Figure 4 FIG. 1 is a flow chart of a data management method for stoma care teaching according to an embodiment of the present application. Figure 4 As shown, in the data management method for stoma nursing teaching, it includes: S110, storing user identity information of nursing students and teachers and storing stoma nursing teaching video collections and virtual training tasks; S120, collecting and transmitting the training operation video of the target nursing student object during the virtual training process; S130, receiving the training operation video of the target nursing student object, and performing core information significant interaction analysis of the training operation video of the target nursing student object based on the training operation characteristics and the skeleton point position characteristics to generate stoma nursing training feedback results; S140, generating training operation suggestions based on the stoma nursing training feedback results, and displaying and storing the training operation suggestions.

[0074] In an embodiment of the present application, a training operation video of the target nursing student is received, and a core information significant interaction analysis based on the training operation features and the skeleton point position features is performed on the training operation video of the target nursing student to generate an ostomy nursing training feedback result, including: key frame sampling of the training operation video to obtain a time series of training operation key frames; key joint skeleton point position information extraction of the time series of the training operation key frames to obtain a skeleton point time series feature vector; dual-stream feature extraction of each training operation key frame to obtain a dual-stream feature map of the training operation key frame; skeleton point-training operation core information significant interaction is performed on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector; the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector is passed through a classifier-based ostomy nursing training feedback generator to obtain an ostomy nursing training feedback result.

[0075] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned data management method for stoma care teaching have been referred to above. Figures 1 to 3 The data management system for ostomy care teaching has been described in detail, and therefore, its repeated description will be omitted.

[0076] In summary, a data management method for stoma nursing teaching based on an embodiment of the present application is described, which first stores the identity information of nursing students and teachers, as well as stoma nursing teaching video sets and virtual training tasks, then collects and transmits the training operation video of the target nursing student object during the virtual training process, and then performs a core information significant interactive analysis based on the training operation characteristics and the skeletal point position characteristics on the received training operation video to obtain the stoma nursing training feedback result of the target nursing student object, and finally generates the training operation suggestion based on the obtained stoma nursing training feedback result, and displays and stores it. In this way, not only can the teaching effect be improved and the cost be reduced, but also the risks in the practice of nursing students are avoided and the safety of teaching is enhanced. Nursing students can use virtual training to improve the proficiency and accuracy of stoma nursing operations, thereby better serving patients and improving the trust and satisfaction of patients.

Claims

1. A data management system for stoma nursing teaching, characterized in that: include: A management module, used to store user identity information of nursing students and teachers and store stoma nursing teaching video collections and virtual training tasks; A virtual training module is used to collect and transmit the training operation videos of the target nursing students during the virtual training process; The training analysis module includes: a key frame sampling unit, which is used to perform key frame sampling on the training operation video to obtain a time series of training operation key frames; a key feature extraction unit, which is used to extract key joint bone point position information from the time series of the training operation key frames to obtain a bone point time series feature vector; a key dual-stream feature extraction unit, which is used to perform dual-stream feature extraction on each training operation key frame to obtain a dual-stream feature map of the training operation key frame; a skeleton point-training operation interaction unit, which is used to perform skeleton point-training operation core information significant interaction on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector; an operation error type generation unit, which is used to pass the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector through a classifier-based stoma nursing training feedback generator to obtain a stoma nursing training feedback result; The teaching feedback module is used to generate practical training operation suggestions based on the stoma care practical training feedback results, and to display and store the practical training operation suggestions.

2. The data management system for stoma care teaching according to claim 1, characterized in that: The virtual training module is used to: Simulating the stoma care process based on the virtual training task, and collecting the training operation video of the target nursing students during the virtual training process; The practical training operation video of the target nursing student is transmitted using the WebSocket or MQTT protocol.

3. The data management system for stoma care teaching according to claim 2, characterized in that: The stoma nursing training feedback result is used to indicate a training operation error type label of the target nursing student.

4. The data management system for stoma care teaching according to claim 3, characterized in that: The key feature extraction unit is used to: Extracting the skeleton point position information of the key joints of the target nursing object from each training operation key frame in the time sequence of the training operation key frame to obtain the time sequence of the training operation skeleton point position; The time series of the skeleton point positions of the training operation is passed through a skeleton point position time series feature extractor based on a one-dimensional causal convolutional neural network model to obtain the skeleton point time series feature vector.

5. The data management system for stoma care teaching according to claim 4, characterized in that: The key dual-stream feature extraction unit is used to: use a dual-stream feature extractor based on EfficientNet and FlowNet to extract each training operation key frame in the time series of the training operation key frame to obtain the dual-stream feature map of the training operation key frame.

6. The data management system for stoma care teaching according to claim 5, characterized in that: The skeleton point-practice training operation interaction unit includes: A skeleton point temporal core clue extraction subunit, used for performing point convolution core clue extraction on the skeleton point temporal feature vector to obtain a skeleton point temporal core clue encoding vector; A training operation key frame core clue extraction subunit is used to extract a training operation key frame dual-stream feature core clue encoding vector from the training operation key frame dual-stream feature map; A skeleton point-practical training operation core clue encoding subunit, used to construct a skeleton point-practical training operation core clue weaving template matrix between the skeleton point temporal core clue encoding vector and the practical training operation key frame dual-stream feature core clue encoding vector; A training operation key frame feature decomposition subunit, used for performing feature decomposition along the channel dimension on the training operation key frame dual-stream feature map to obtain a set of local dual-stream feature matrices of the training operation key frame; A skeleton point-training operation cross-modal encoding subunit is used to use the skeleton point time series feature vector as a query vector, each training operation key frame local two-stream feature matrix in the set of training operation key frame local two-stream feature matrices as a key matrix, and the skeleton point-training operation core clue weaving template matrix as a prior information constraint matrix, and perform cross-modal template constraint heterogeneous transformation on it to obtain a set of skeleton point-training operation cross-modal fine-grained interaction encoding vectors; The skeleton point-training operation significant interaction feature generation subunit is used to calculate the positional mean vector of the set of the skeleton point-training operation cross-modal fine-grained interaction encoding vectors to obtain the skeleton point-training operation dual-stream semantic cross-modal significant interaction encoding vector.

7. The data management system for stoma care teaching according to claim 6, characterized in that: The skeleton point-practice operation core clue encoding subunit includes: The core association matrix constructs a secondary subunit, which is used to process the skeleton point temporal core clue encoding vector and the training operation key frame dual-stream feature core clue encoding vector using a feature mapping function to obtain a skeleton point-training operation core association matrix; The core association matrix global constraint secondary subunit is used to perform feature processing on the skeleton point-practical operation core association matrix based on global semantic constraints to obtain the skeleton point-practical operation core clue weaving template matrix.

8. The data management system for stoma care teaching according to claim 7, characterized in that: The core association matrix globally constrains the secondary subunit, which is used to: Using the residual mechanism and the feature mapping function, the skeleton point temporal core clue encoding vector and the training operation key frame dual-stream feature core clue encoding vector are preliminarily structured semantically aligned to obtain a skeleton point-training operation semantic alignment vector and a training operation-skeleton point semantic alignment vector; Calculating a skeleton point-training operation explicit global semantic correlation matrix between the skeleton point-training operation semantic alignment vector and the training operation-skeleton point semantic alignment vector; Based on the skeleton point-training operation explicit global semantic relevance matrix, the skeleton point-training operation core association matrix is ​​significantly fine-grained compensated to obtain the skeleton point-training operation core clue weaving template matrix.

9. The data management system for stoma care teaching according to claim 8, characterized in that: The teaching feedback module is used to: Generate the practical training operation suggestion based on the stoma nursing practical training feedback result, and retrieve the stoma nursing teaching video corresponding to the stoma nursing practical training feedback result in the stoma nursing teaching video collection; The training operation suggestions and the stoma care teaching video are displayed on the virtual training interface of the target nursing student, and the stoma care training feedback results and the training operation suggestions are stored in a database.

10. A data management method for stoma nursing teaching, characterized in that: include: Storing user identity information of nursing students and teachers, as well as stoma nursing teaching video collections and virtual training tasks; Collect and transmit the training operation videos of the target nursing students during the virtual training process; Receive the training operation video of the target nursing student object, and perform core information significant interaction analysis based on the training operation features and the skeleton point position features on the training operation video of the target nursing student object to generate stoma nursing training feedback results, including: sampling the key frames of the training operation video to obtain a time series of training operation key frames; extracting the key joint skeleton point position information of the time series of the training operation key frames to obtain a skeleton point time series feature vector; extracting dual-stream features of each training operation key frame to obtain a training operation key frame dual-stream feature map; performing skeleton point-training operation core information significant interaction on the skeleton point time series feature vector and the training operation key frame dual-stream feature map to obtain a skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector; passing the skeleton point-training operation dual-stream semantic cross-modal significant interaction coding vector through a classifier-based stoma nursing training feedback generator to obtain a stoma nursing training feedback result; A practical training operation suggestion is generated based on the stoma care practical training feedback result, and the practical training operation suggestion is displayed and stored.

Images