Nursing practice teaching demonstration system based on big data

Abstract

The invention relates to the technical field of nursing practice teaching, and discloses a nursing practice teaching demonstration system based on big data, which comprises a teaching data acquisition module, a teaching path generation module, a dynamic evaluation module, a feedback optimization module and a path map correction module. The teaching data acquisition module acquires data through a preprocessing node and a multi-dimensional association engine and constructs a multi-dimensional teaching path topology; the teaching path generation module generates a teaching path by using a feature analysis unit and a scene reconstruction unit, and configures a multi-state switching link to cope with different scenes; the dynamic evaluation module generates a reference strategy through a strategy modeling node and a feedback optimization unit and dynamically evaluates the reference strategy; and the feedback optimization module and the path map correction module realize abnormity correction, parameter calibration and path state monitoring. According to the system, through multi-module cooperation and big data analysis, accurate processing of teaching data and dynamic generation and evaluation of teaching paths are realized, and the teaching quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of nursing practice teaching, and particularly to a nursing practice teaching demonstration system based on big data. Background Art

[0002] In the field of nursing education, practice teaching is the core link in cultivating professional nursing talents, and its teaching effect is directly related to the clinical operation ability and emergency handling level of nursing staff. Traditional nursing practice teaching mainly relies on on-site demonstrations by teachers, simulated operations by students, and teaching models with fixed processes, and there are many limitations that are difficult to break through.

[0003] From the perspective of teaching data processing, the traditional mode lacks the systematic collection and analysis ability of multi-dimensional data in the teaching scenario. Nursing practice involves a large number of complex operation behavior parameters (such as operation duration, movement amplitude, equipment adjustment accuracy, etc.) and dynamically changing teaching scenario data (such as fluctuations in the vital signs of simulated patients, sudden interference events in the teaching environment, etc.). Traditional teaching cannot capture these data in real time and build a correlation model, resulting in the difficulty of accurately reflecting the complexity and dynamics of actual nursing operations in the design of the teaching path.

[0004] In terms of generating the teaching path, traditional teaching often adopts a fixed process-based teaching mode and lacks targeted adaptation to individual differences and operation characteristics of students. There are significant differences among different students in terms of learning progress, operation habits, and problem-solving abilities, but the traditional system cannot dynamically adjust the teaching path according to the real-time operation data of students, making it difficult to achieve personalized teaching, and resulting in the learning effect of some students being affected due to the teaching content being too difficult or too easy.

[0005] The lack of a dynamic evaluation mechanism is also a major shortcoming of traditional nursing practice teaching. Traditional evaluation mainly relies on the subjective evaluation of teachers and post-operation assessments, and cannot monitor the operation status of students in real time during the teaching process and provide immediate feedback. This lagging evaluation is difficult to detect subtle deviations and potential problems in students' operations in a timely manner, resulting in the solidification of wrong operation habits and increasing the safety risks in clinical practice.

[0006] In addition, the traditional teaching system lacks effective coping strategies for abnormal situations during the teaching process. For example, when students make sudden mistakes during the operation or the teaching environment is interfered externally, the system cannot quickly identify the abnormality and adjust the teaching strategy, easily leading to the interruption of the teaching process or a decline in the teaching effect. At the same time, there is no effective data interaction and coordination mechanism among the various teaching modules of the traditional system, and it is impossible to form a closed-loop teaching feedback system, making it difficult to achieve continuous optimization of the teaching process.

[0007] With the rapid development of technologies such as big data and artificial intelligence, introducing advanced technologies into nursing practice teaching has become an inevitable trend to improve teaching quality. Although there are some teaching assistance systems based on information technology in the existing technologies, most of them stay at the level of simple data recording and process demonstration, and fail to make full use of the analysis ability of big data to realize the intelligentization, dynamicization and personalization of the teaching process. Therefore, there is an urgent need for a nursing practice teaching demonstration system that can integrate multi-dimensional teaching data, dynamically generate adaptive teaching paths, and achieve real-time and accurate evaluation, so as to solve the problems existing in traditional teaching such as insufficient data processing ability, rigid teaching paths, lagging evaluation, and weak abnormal response ability, and promote the development of nursing practice teaching towards a more efficient and intelligent direction. Summary of the Invention

[0008] The purpose of the present invention is to provide a nursing practice teaching demonstration system based on big data to solve the problems raised in the above background technology.

[0009] To achieve the above purpose, the present invention provides the following technical solution: A nursing practice teaching demonstration system based on big data, the system includes: A teaching data collection module, a teaching path generation module and a dynamic evaluation module; The teaching data collection module includes a preprocessing node and a multi-dimensional association engine; the teaching path generation module includes a feature analysis unit and a scenario reconstruction unit; the dynamic evaluation module includes a strategy modeling node and a feedback optimization unit; the preprocessing node is used to obtain teaching scenario data and operation behavior parameters in real time; according to the teaching stage identifier and the time series label, construct a multi-dimensional teaching path topology through the multi-dimensional association engine; the multi-dimensional association engine is composed of multiple teaching nodes based on feature thresholds in parallel; based on the real-time response characteristics of multiple teaching nodes, generate a benchmark teaching strategy through the strategy modeling node, and combine the reconstruction signal output by the scenario reconstruction unit, and dynamically cluster and sort the teaching path state through the feedback optimization unit.

[0010] Preferably, the system further includes a feedback optimization module and a path map correction module; the feedback optimization module includes a behavior isolation unit and an error compensation unit; According to the generated benchmark teaching strategy and the dynamic clustering and sorting results, the behavior isolation unit is used to perform distributed anomaly correction on the teaching path state, and the error compensation unit is used to dynamically calibrate the characteristic parameters of the teaching path generation module; at the same time, the chain stability evaluation network constructed by the path map correction module is used to monitor the path synchronization rate in real time, and the monitoring results are fed back to the strategy modeling node and the feedback optimization unit to perform closed-loop adjustment on the teaching strategy.

[0011] Preferably, the behavior isolation unit is composed of state regulation circuits corresponding to teaching nodes connected in parallel in the multi-dimensional association engine; the steps for constructing the multi-dimensional teaching path topology include: Collect teaching scenario data and operation behavior fluctuation signals, input them into the transfer learning models configured for each level of teaching nodes, generate data consistency calibration coefficients, and store them in the corresponding teaching nodes; Based on the consistency calibration coefficients stored in each level of teaching nodes, combined with the instantaneous change rate of time series tags, allocate the teaching path states to multiple teaching nodes through a dynamic feature detection algorithm to form a dynamic path matching topology.

[0012] Preferably, the conditions for constructing the teaching nodes configured in the preprocessing nodes include: the teaching scenario data is within a preset feature interval, the synchronization index corresponding to the teaching node is greater than a threshold, and the teaching stage identifier is less than the allowable deviation range; The feature analysis unit is also configured with a polymorphic switching link; the polymorphic switching link includes a steady-state teaching sub-link and an anomaly avoidance sub-link; the steady-state teaching sub-link is generated based on the matching degree between the chain stability signal of the teaching path generation module and the reference teaching strategy; the anomaly avoidance sub-link is constructed based on the correlation between the instantaneous anomaly risk of the teaching path state and external interference events.

[0013] Preferably, the policy modeling node is configured with a multi-modal fusion model; the multi-modal fusion model includes a feature-time mapping table corresponding to the teaching data collection module and an anomaly avoidance parameter library of the teaching path generation module; the steps for generating the reference teaching strategy include: According to the real-time feature data and clustering sorting results of the teaching data collection module, calculate the upper limit of the theoretical synchronization range of each level of teaching nodes in the steady-state teaching sub-link; Input the upper limit of the theoretical synchronization range and the transient time tag into the multi-modal fusion model to generate a chain teaching reference map, and correct the anomaly risk of the reference map through the anomaly avoidance sub-link to form a reference teaching strategy.

[0014] Preferably, the parameter update steps of the multi-modal fusion model include: When it is detected that the teaching scenario data exceeds the preset interval or the path synchronization rate exceeds the threshold, trigger the first anomaly avoidance instruction, that is, start the alternative teaching unit and adjust the parsing granularity of the teaching path generation module; If the path state still fails to return to the allowable interval after the execution of the first anomaly avoidance instruction, trigger the second anomaly avoidance instruction, that is, switch to the anomaly avoidance sub-link through the polymorphic switching link, and reallocate the topological weights of the teaching path state based on the duration of the external interference event.

[0015] Preferably, the parameter update steps of the multi-modal fusion model further include: When the synchronization index of any level of teaching node in the teaching data acquisition module is lower than the threshold, a third abnormal avoidance instruction is triggered, that is, the path output channel of the node is blocked, and the corresponding data is transferred to other teaching nodes through the behavior isolation unit; During the execution of the third abnormal avoidance instruction, if it is detected that other teaching nodes are overloaded, a fourth abnormal avoidance instruction is triggered, that is, the path graph correction module is called to degrade the output strategy of the feedback optimization unit and limit the synchronization requirements of the edge state.

[0016] Preferably, the degradation processing logic of the path graph correction module includes: Construct a path state safety level table according to the synchronization rate data output by the chain stability evaluation network and the off-chain characteristic indicators; When the fourth abnormal avoidance instruction is triggered, the teaching path priority is dynamically degraded according to the safety level table, and the parsing frequency of the teaching path generation module is synchronously adjusted to match the degraded teaching strategy.

[0017] Preferably, the construction steps of the teaching node further include: Generate a feature gradient adaptability index table based on the spatial distribution result of teaching scenario data and the historical operation behavior range; According to the difference in the consistency calibration coefficients of each teaching node in the index table, the topological weights of the state control circuits of multiple teaching nodes are assigned through a dynamic weight distribution algorithm, and redundant transmission links are established between the teaching nodes.

[0018] Preferably, the generation process of the chain-type teaching reference graph includes: After inputting the upper limit of the theoretical synchronization range and the transient time label into the multi-modal fusion model, the preset teaching medium database and the external interference event layer are synchronously loaded; Based on the encrypted peak frequency data output by the feature parsing unit, the abnormal threshold of the teaching path state in the reference graph is corrected, and the time gradient parameter is superimposed in the abnormal avoidance sub-link for secondary calibration.

[0019] Compared with the prior art, the beneficial effects of the present invention are: In terms of teaching data collection and processing, the teaching data collection module obtains teaching scene data and operation behavior parameters in real time through preprocessing nodes, and uses a multidimensional association engine to build a multidimensional teaching path topology. This mechanism can fully capture the complex dynamic data in nursing practice teaching, such as operation behavior fluctuation signals, spatial distribution characteristics of teaching scenes, etc., and generate data consistency calibration coefficients through transfer learning models to achieve efficient calibration and association analysis of multi-source heterogeneous data. This refined data processing capability lays a solid foundation for the precise generation and dynamic evaluation of subsequent teaching paths, enabling the teaching system to more realistically reflect the actual scenes of nursing operations and the operation characteristics of students.

[0020] The feature parsing unit and scene reconstruction unit of the teaching path generation module cooperate with each other to realize the dynamic adaptive adjustment of the teaching path based on the polymorphic switching link. The steady-state teaching sub-link can generate a stable teaching path according to the matching degree between the chain stability signal and the benchmark teaching strategy to ensure the orderly progress of the teaching process; while the abnormal avoidance sub-link can quickly build a targeted avoidance path based on the correlation between the instantaneous abnormal risk of the teaching path state and the external interference event. This dual-link mechanism enables the system to flexibly respond to normal teaching scenarios and sudden abnormal situations in the teaching process, which not only ensures the stability of teaching, but also enhances the anti-interference ability and adaptability of the teaching system. For example, when it is detected that the teaching scene data exceeds the preset interval, the system can switch to the abnormal avoidance sub-link in time through the polymorphic switching link, adjust the parsing granularity and topological weight of the teaching path, and avoid teaching interruption or effect decline due to abnormal situations.

[0021] The dynamic evaluation module realizes real-time monitoring and dynamic evaluation of the teaching path status through the collaborative work of the strategy modeling node and the feedback optimization unit. The multimodal fusion model configured by the strategy modeling node can integrate the feature-time mapping table of the teaching data acquisition module and the anomaly avoidance parameter library of the teaching path generation module to generate a scientific and reasonable benchmark teaching strategy. At the same time, the feedback optimization unit performs distributed anomaly correction and dynamic calibration of feature parameters on the teaching path status through the behavior isolation unit and the error compensation unit to ensure the accuracy and effectiveness of the teaching strategy. This closed-loop dynamic evaluation mechanism can detect deviations and problems in students' operations in real time during the teaching process, and correct them in time through feedback adjustment, which helps students form correct operating habits and improve teaching quality and learning effects.

[0022] The settings of the feedback optimization module and the path map correction module in the system further enhance the stability and reliability of the teaching system. The behavior isolation unit can precisely correct abnormal teaching path states through the state control circuits corresponding to the teaching nodes connected in parallel in the multi-dimensional correlation engine, avoiding the spread and influence of abnormal states among teaching nodes. The chain-type stability evaluation network constructed by the path map correction module can monitor the path synchronization rate in real time, and construct a path state security level table based on the synchronization rate data and off-chain characteristic indicators, realizing the dynamic downgrading of the teaching path priority and the adjustment of the parsing frequency. This hierarchical processing mechanism can reasonably allocate resources when the system is overloaded or abnormal, ensure the stable operation of critical teaching paths, and improve the overall robustness of the system.

[0023] The parameter update mechanism and the abnormal avoidance instruction system of the multi-modal fusion model enable the system to have strong adaptability and fault tolerance. When various abnormal situations occur during the teaching process, the system can trigger different levels of abnormal avoidance instructions in sequence according to the severity and duration of the abnormalities, such as starting alternative teaching units, switching teaching links, blocking overloaded nodes, etc. The orderly execution of these instructions can gradually resolve abnormal risks and ensure the continuity and stability of the teaching process. At the same time, through the dynamic weight allocation algorithm and the establishment of redundant transmission links, flexible data transmission and load balancing can be achieved between teaching nodes, further improving the reliability and teaching efficiency of the system. Brief Description of the Drawings

[0024] Figure 1 It is the working principle diagram of the big data-based nursing practice teaching demonstration system described in the present invention; Figure 2 It is the working principle diagram of feedback optimization and correction; Figure 3 It is the flow chart of multi-dimensional teaching path topology construction; Figure 4 It is the flow chart of benchmark teaching strategy generation. Detailed Embodiments

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0026] Please refer to Figures 1 - 4, the nursing practice teaching demonstration system based on big data involved in the present invention includes a teaching data collection module, a teaching path generation module, and a dynamic evaluation module. Each module cooperates with each other to achieve big data-driven nursing practice teaching demonstration. The specific implementation steps are as follows: The teaching data collection module includes a preprocessing node and a multi-dimensional association engine. The preprocessing node can obtain teaching scenario data in real time, such as image frames of nursing operation videos, timestamps of operation steps, parameters such as operation force and angle collected by device sensors, etc. At the same time, operation behavior parameters are obtained, such as the operation duration of learners, the number of errors, action coherence indicators, etc. Then, according to the teaching stage identifier (such as the unique identifier of different teaching stages such as basic nursing and specialty nursing) and the time series label (time mark accurate to the millisecond level), a multi-dimensional teaching path topology is constructed through the multi-dimensional association engine. Among them, the multi-dimensional association engine is composed of multiple teaching nodes based on feature thresholds in parallel, and each teaching node corresponds to different teaching feature dimensions, such as the needle insertion angle node for intravenous puncture, the process compliance node for aseptic operation, etc. Each teaching node performs real-time screening and correlation analysis on the input data based on a preset feature threshold (such as the needle insertion angle threshold is 30°-45°).

[0027] The teaching path generation module includes a feature analysis unit and a scenario reconstruction unit. The feature analysis unit deeply analyzes the multi-dimensional data output by the teaching data collection module, and extracts key feature parameters, such as the spatial trajectory features of operation actions, the rhythm features of time series, etc. The scenario reconstruction unit then reconstructs a virtual operation environment highly similar to the actual teaching scenario, including ward scenarios, operating room scenarios, etc., through 3D modeling and virtual reality technology, and dynamically displays the teaching path therein according to these feature parameters.

[0028] The dynamic evaluation module includes a strategy modeling node and a feedback optimization unit. The strategy modeling node generates a benchmark teaching strategy through machine learning algorithms based on the real-time response characteristics of multiple teaching nodes, such as the response time of each node to operation behavior parameters, the data processing volume, etc. This strategy covers the presentation order of teaching content, the emphasis methods of key points and difficulties, the setting of interactive links, etc. The feedback optimization unit combines the reconstruction signals output by the scenario reconstruction unit (such as the completion status signal of operation steps in the virtual scenario) to perform dynamic clustering and sorting on the teaching path status, and classify similar teaching path statuses for subsequent evaluation and optimization.

[0029] The present invention will be further described in detail below with specific embodiments.

[0030] Embodiment 1: In this embodiment, the system adds a feedback optimization module and a path map correction module to improve the stability of the teaching path and the closed-loop nature of strategy adjustment. The feedback optimization module includes a behavior isolation unit and an error compensation unit, and the two cooperate with the path map correction module to form a multi-level control mechanism for the state of the teaching path.

[0031] After the system generates the benchmark teaching strategy, it is necessary to monitor and correct the state of the teaching path in combination with the dynamic clustering sorting results. The core of the behavior isolation unit is the state control circuit corresponding to each parallel teaching node in the multi-dimensional association engine. These circuits have data transmission control functions and can perform differential processing for different types of state anomalies of teaching nodes. For example, when a certain teaching node (such as the "aseptic operation process compliance node") detects that the operation behavior parameter (such as the omission of the glove wearing step) exceeds the preset feature interval, the state control circuit corresponding to this node can immediately activate the isolation mechanism: by cutting off the output channel of the abnormal data, preventing the unqualified operation data from spreading to other nodes, and at the same time triggering a warning signal to the error compensation unit.

[0032] The role of the error compensation unit is to dynamically calibrate the characteristic parameters of the teaching path generation module. Specifically, when the feature analysis unit analyzes the teaching scenario data (such as the force sensor data of the intravenous puncture operation), the extraction of characteristic parameters may deviate due to equipment accuracy fluctuations or environmental interference. The error compensation unit establishes a parameter deviation correction model by collecting historical calibration data. When receiving the abnormal warning from the behavior isolation unit, this unit will perform real-time adjustment on the characteristic parameters corresponding to the abnormal node (such as the calculation threshold of the needle insertion angle): if the deviation between the measured angle data of an operation and the theoretical value exceeds the preset range, the error compensation unit will automatically fine-tune the weight coefficient of the feature analysis algorithm to reduce the calculation error and make the subsequent analyzed characteristic parameters closer to the actual operation standard.

[0033] The chain stability evaluation network constructed by the path map correction module collects the path synchronization rate data of each teaching node in real time through distributed sensors. The synchronization rate index covers parameters such as data transmission delay and state consistency between nodes. For example, in the simulated operating room scenario, the teaching node of the instrument transfer step needs to be synchronized with the aseptic operation node. If the deviation between the transfer timestamp and the operation process timestamp exceeds the threshold, it is determined as a synchronization anomaly. The evaluation network feeds the real-time monitoring data back to the strategy modeling node and the feedback optimization unit: the strategy modeling node adjusts the generation logic of the benchmark teaching strategy according to the type of synchronization anomaly, such as inserting a delay compensation link in the teaching path; the feedback optimization unit re-optimizes the dynamic clustering sorting rules according to the synchronization rate fluctuation, and classifies the path state with synchronization anomaly into a specific clustering group for targeted adjustment.

[0034] In the specific execution process, the system first obtains real-time operation data through the teaching data collection module, and generates a preliminary teaching path topology after being processed by the multi-dimensional correlation engine. The dynamic evaluation module generates a benchmark teaching strategy based on this topology, and executes the first round of anomaly correction through the feedback optimization module. For example, when a learner is performing a catheterization operation and enters the intubation step without disinfecting the vulva according to the procedure, this abnormal behavior is captured by the "operation process compliance node", and the status control circuit immediately isolates the abnormal data of this node to prevent it from affecting the normal operation of the "aseptic operation specification node". At the same time, the error compensation unit calibrates the parameters of "operation process sequence analysis" in the feature analysis unit to ensure the accuracy of subsequent operation process judgment.

[0035] The intervention of the path map correction module further enhances the system's self-adaptability. The chain stability evaluation network continuously monitors the synchronization status of each node. If it is found that the synchronization rate of multiple teaching nodes decreases due to the spread of abnormal data (such as the timestamp matching error between the nursing document writing node and the operation step node accumulates more than 5%), the evaluation network will feedback this status to the policy modeling node. The policy modeling node then initiates a policy adjustment process: according to the historical synchronization anomaly case library, it adjusts the time allocation weights of each link in the benchmark teaching strategy, such as extending the interval time between the operation step demonstration and document writing, to improve the learner's step connection ability. The feedback optimization unit then recalculates the priorities of each clustering group based on the new synchronization data, sets the path status with frequent synchronization anomalies as the high-priority correction object, and gradually restores the stability of the system through the coordinated action of the behavior isolation unit and the error compensation unit.

[0036] In addition, there is a two-way communication mechanism between the behavior isolation unit and the path map correction module. When the behavior isolation unit performs multiple anomaly isolation operations, the system will trigger a deep evaluation of the path map correction module: by analyzing the distribution characteristics of abnormal data (such as concentrated in a certain teaching stage or a certain type of operation), the evaluation network can identify the weak links in the teaching path topology. For example, if the "first aid operation process node" frequently has data anomalies due to external device interference (such as signal fluctuations of the simulated electrocardiogram monitor), the path map correction module will instruct the policy modeling node to generate targeted strategies, such as adding a simulated scenario for coping with device interference in the first aid teaching path, and at the same time adjusting the parameters of the status control circuit of this node to improve its anti-interference ability.

[0037] The dynamic calibration process of the error compensation unit has the characteristics of progressive learning. The system accumulates the operation data of different learners to establish a multi-dimensional characteristic parameter deviation database. When common characteristic deviations occur in similar operations (such as intravenous punctures by different learners), the error compensation unit will automatically update the parameters of the correction model to form a universal calibration rule. For example, if it is found that most beginners have a relatively small needle insertion angle, the error compensation unit will increase the angle correction coefficient in the feature analysis algorithm to make the analyzed angle data closer to the standard value, thereby guiding learners to gradually correct their operation habits.

[0038] The feedback mechanism of the path map correction module also involves the long-term optimization of teaching strategies. By continuously analyzing the historical data of the chain stability evaluation network, the system can identify structural problems in the teaching path, such as a long-term high failure rate of node synchronization in a certain teaching stage. At this time, the path map correction module will cooperate with the strategy modeling node to make structural adjustments to the benchmark teaching strategy, such as splitting complex teaching nodes and adding intermediate verification links, to reduce the synchronization pressure between nodes and improve the stability of the overall teaching path.

[0039] Embodiment 2: In this embodiment, the specific composition of the behavior isolation unit and the construction steps of the multi-dimensional teaching path topology form a dynamic allocation mechanism for data processing to ensure the accurate mapping and abnormal isolation of teaching scenario data and operation behavior parameters. The behavior isolation unit is composed of the state control circuits corresponding to each parallel teaching node in the multi-dimensional association engine. The state control circuit of each teaching node includes a data filtering module, an abnormal marking module, and a path switching module. The data filtering module performs real-time screening of the input data based on preset feature thresholds (such as the safe range of intravenous puncture depth and the compliance interval of aseptic operation time intervals). The abnormal marking module generates a timestamped abnormal label when it detects that the data exceeds the threshold. The path switching module then automatically switches the data transmission path according to the abnormal label and directs the abnormal data to a dedicated correction node.

[0040] The construction of the multi-dimensional teaching path topology starts from the data acquisition layer. The teaching data acquisition module collects teaching scenario data (such as the three-dimensional spatial coordinates of nursing operations and the waveform data of instrument contact forces) and operation behavior fluctuation signals (such as the number of learner operation pauses and the sudden change value of action acceleration) in real time through cameras, pressure sensors, motion capture devices, etc. deployed in the virtual simulation environment. These raw data are input into the transfer learning model of each teaching node in the form of data streams. The transfer learning model performs feature extraction and consistency analysis on the real-time data based on a pre-trained standard model of nursing operations (such as a basic nursing operation specification database) to generate a data consistency calibration coefficient. This coefficient is a value in the range of 0 - 1, reflecting the degree of coincidence between the current operation data and the historical standard data. For example, a coefficient of 0.9 indicates that the operation process matches the standard process by 90%, and a coefficient of 0.4 indicates a significant deviation. The calibration coefficient is stored synchronously with the real-time data in the corresponding teaching node to form a dynamically updated node feature library.

[0041] Based on the consistency calibration coefficients stored in each teaching node, the system enters the dynamic path allocation stage. The instantaneous change rate calculation module of the time series tag performs differential processing on the operation timestamp to obtain the change amplitude of the time tag per unit time (such as whether the time advance per millisecond is uniform). The dynamic feature detection algorithm combines the consistency calibration coefficient and the instantaneous change rate to perform multi-node allocation of the teaching path status. The specific rules are as follows: When the consistency calibration coefficient of a certain operation data is ≥ 0.8 and the instantaneous change rate ≤ the preset threshold, it is determined to be in a normal state and assigned to the "steady-state teaching node" for standard process processing; when the consistency calibration coefficient is between 0.5 - 0.8 or the instantaneous change rate exceeds the threshold, it is determined to be in a warning state and assigned to the "transition teaching node" for deviation analysis; when the consistency calibration coefficient < 0.5 and the instantaneous change rate ≥ twice the threshold, it is determined to be in an abnormal state and assigned to the "abnormal teaching node" for isolation and correction. For example, in the teaching scenario of "cardiopulmonary resuscitation operation", if the consistency calibration coefficient of the learner's pressing frequency is 0.6 and the time interval change rate is large, the system assigns this operation state to the transition teaching node to trigger a real-time prompt for the deviation of the pressing frequency; if the calibration coefficient is only 0.3 and there are continuous pressing interruptions, it is assigned to the abnormal teaching node to pause the current operation process and start the error correction guidance.

[0042] The formation of the dynamic path matching topology depends on the real-time communication mechanism between teaching nodes. Each teaching node is connected by a high-speed data bus, and the data transmitted between nodes includes information such as calibration coefficients, time tags, and status identifiers. When a teaching node receives the allocated path status data, it first performs secondary verification through the filtering module of the status control circuit: if the data is distorted during transmission (such as a jump in the pressure sensor value), the filtering module will call the calibration coefficient of the previous moment for interpolation and completion; if the verification passes, the data will be input into the feature analysis unit for in-depth processing. For example, in the "urinary catheterization" teaching, when an abnormal teaching node receives data with an incorrect glove-wearing order, after the filtering module confirms the data validity, it triggers the abnormal marking module to generate a "procedure reversed" label, and through the path switching module, directs the data to the "aseptic operation specification correction node", while sending visual step prompt information to the learner terminal.

[0043] The continuous training mechanism of the transfer learning model ensures the accuracy of data consistency calibration. The system regularly collects the operation data of different learners, classifies and labels them according to teaching stages (basic, specialist, comprehensive) and operation types (injection, puncture, first aid), and forms an incremental training dataset. When the standard deviation of the calibration coefficient of a teaching node exceeds the preset value for 3 consecutive teaching cycles, the system automatically triggers the model update process: fine-tuning the transfer learning model with the new data and adjusting the weight parameters of the feature extraction layer to make the model adapt to the operation habit differences of different learners. For example, for the characteristics of smaller movement amplitudes of elderly learners, the transfer learning model of the "venous puncture angle node" can automatically adjust the calculation method of the angle threshold, dynamically expanding the standard range from 30° - 45° to 25° - 50° for a more inclusive standard for consistency assessment.

[0044] The parameter adaptive adjustment of the dynamic feature detection algorithm is a key link in topology construction. The algorithm dynamically adjusts the threshold of status allocation by monitoring the load rate of each teaching node (such as the proportion of the data processed per unit time in the maximum processing capacity of the node). When the load rate of a steady-state teaching node exceeds 80%, the algorithm automatically reallocates some low-risk warning states (such as calibration coefficients of 0.7 - 0.8 and a change rate close to the threshold) to the transition teaching node to balance the processing pressure between nodes; when an abnormal teaching node processes the same type of abnormal state 5 consecutive times, the algorithm will lower the determination threshold of this type of abnormality by 5% to identify potential operation errors in advance. For example, in the "infusion operation" teaching, if multiple learners cause frequent triggering of the abnormal teaching node due to incomplete air exhaustion, the algorithm will lower the detection threshold of the "air bubble residue amount" from 0.5 ml to 0.3 ml to improve the sensitivity of error identification.

[0045] The state control circuit of the teaching node also has a hot standby function. When the state control circuit of a certain node fails due to hardware failure or data overload, the system automatically enables the redundant circuit of the parallel node to take over the data processing task. For example, when the main circuit of the "disinfection range node" fails, the standby circuit is immediately activated to process the data based on the same characteristic threshold and calibration coefficient, ensuring that the evaluation of the teaching path state is not interrupted. The redundancy mechanism is implemented through the circuit parallel design at the hardware level and the fault detection protocol at the software level. The fault detection period is 50 milliseconds, and the main-standby switch can be completed in a very short time.

[0046] The visual presentation of the multi-dimensional teaching path topology provides an intuitive basis for teaching evaluation. The system displays the state distribution of each teaching node, the data flow direction, and the association strength between nodes through a graphical interface. For example, in the teaching scenario of "operating room nursing cooperation", the visual interface can display the synchronization status of the "instrument transfer node" and the "operation step node" in real time: when the difference between the calibration coefficients of the two exceeds 0.2, the connection line between the nodes turns red and flashes, indicating that there is a problem with the process connection for the teacher or learner. This visualization mechanism helps to quickly locate the bottleneck links in the teaching path and provides data support for teaching strategy adjustment.

[0047] Example 3: In this embodiment, the teaching node construction conditions configured by the preprocessing node and the polymorphic switching link of the feature analysis unit together constitute an intelligent screening and adaptive switching mechanism for the teaching path, ensuring the stability and robustness of the system in different teaching states. The construction of the teaching node by the preprocessing node needs to meet three conditions: the teaching scenario data is within the preset feature interval, the synchronization index corresponding to the teaching node is greater than the threshold, and the teaching stage identifier is less than the allowable deviation range. Among them, the preset feature interval of the teaching scenario data is predefined according to the nursing operation standards. For example, in the "intradermal injection" operation, the preset interval for the injection angle is 5°-15°, and the preset interval for the operation duration is 10-15 seconds; the synchronization index is an indicator to measure the data interaction efficiency between the teaching node and other nodes, and the calculation formula is:

[0048] In the formula, represents the synchronization index, is the number of successful data communications between nodes, is the total number of data transmissions. This index is calculated by real-time statistics of the sending and receiving of data packets between nodes. When It is determined that the synchronization requirement is met when; the teaching stage identifier is the only code representing the current teaching progress (for example, the code for the basic nursing stage is B01, and the code for the specialized nursing stage is Z02). The allowable deviation range is set by the system according to the teaching syllabus, usually the Hamming distance between the current stage code and the target stage code does not exceed 1 (for example, the transition from B01 to B02 is regarded as a valid deviation).

[0049] The polymorphic switching link configured by the feature analysis unit includes a steady-state teaching sub-link and an anomaly avoidance sub-link, and the two achieve seamless switching through the state determination module. The generation of the steady-state teaching sub-link is based on the matching degree between the chain stability signal of the teaching path generation module and the reference teaching strategy. The chain stability signal is generated by monitoring the state consistency of each teaching node. For example, in the "aseptic operation" teaching, if the calibration coefficients of the teaching nodes of 5 consecutive operation steps are all greater than 0.8 and the synchronization index remains above 0.95, it is determined to be in a steady state. At this time, the steady-state teaching sub-link is activated, and the system advances the teaching according to the preset process of the reference teaching strategy (such as operation demonstration → step-by-step practice → comprehensive assessment). The calculation of the matching degree involves the Euclidean distance of the characteristic parameters, and the formula is:

[0050] In the formula, is the matching degree value, is the measured characteristic parameter of the current teaching node (such as disinfection time, instrument holding force), is the standard characteristic parameter in the reference teaching strategy, is the number of dimensions of the characteristic parameter. When ( is the preset matching threshold, usually taking 0.3), it is determined to be a high matching degree, and the steady-state sub-link operation is maintained.

[0051] The construction of the anomaly avoidance sub-link is based on the correlation between the instantaneous anomaly risk of the teaching path state and external interference events. The instantaneous anomaly risk is identified by calculating the fluctuation amplitude of the characteristic parameters in real time. For example, in the "cardiopulmonary resuscitation" operation, if the measured value of the chest compression depth fluctuates by more than 30% of the standard value within 200 milliseconds, an anomaly risk warning is triggered. External interference events include network latency, equipment failures, virtual scene data transmission interruptions, etc. The system captures the interference type and duration in real time through the event listening module. When the anomaly risk level (comprehensively determined by the fluctuation amplitude and duration) reaches the preset threshold and an external interference event is detected, the state determination module triggers a switching instruction to redirect the data flow to the anomaly avoidance sub-link. For example, when the virtual blood pressure monitor causes data interruption due to equipment failure, the anomaly avoidance sub-link automatically calls historical simulation data to fill the interrupted interval, and at the same time adjusts the teaching path to insert a theoretical explanation session on "equipment failure response" to avoid teaching process interruption.

[0052] In the specific implementation process, the preprocessing node first preliminarily filters the teaching scenario data collected in real time (such as the puncture speed and liquid medicine flow rate of "intravenous infusion operation"), and only allows the data that meets the preset feature interval to enter the teaching node. For example, if the puncture speed exceeds 60 mm / s (outside the preset interval of 20 - 50 mm / s), the data is marked as abnormal and temporarily stored in the buffer queue until subsequent calibration processing. At the same time, the synchronization index calculation module monitors the data packet transmission situation between nodes in real time. If the exceeds 1.5 times for 10 consecutive times, the system determines that the node is out of synchronization and temporarily freezes its data input until the synchronization index resumes above the threshold.

[0053] The polymorphic switching link of the feature parsing unit dynamically evaluates the link state during the teaching process. When the steady-state teaching sub-link is running, the feature parsing unit parses the feature parameters of each operation step in sequence according to the process of the benchmark teaching strategy. For example, in the operation of "laying a sterile drape", the parameters such as the opening method of the sterile towel, the order of placing items, and the edge folding length are parsed in sequence, and it is judged whether it meets the standard through the matching degree calculation. If the value exceeds 3 times in a row, the system determines that there is a steady-state deviation and starts the abnormal risk assessment: if an operation behavior fluctuation signal is detected at the same time (such as the hand shaking amplitude exceeds 0.5 cm), the abnormal avoidance sub-link is triggered, and the learner is guided to correct the operation deviation through voice prompts, and at the same time, the parsing accuracy of the subsequent teaching nodes is adjusted (such as switching from three-dimensional space coordinate parsing to two-dimensional contour analysis to reduce the calculation load).

[0054] The execution of the abnormal avoidance sub-link includes two stages: interference response and path reconstruction. In the interference response stage, the system calls the corresponding coping strategy according to the type of external interference event: if it is the virtual scene freeze caused by network delay, it automatically switches to the static operation schematic diagram cached locally; if it is the data loss caused by sensor failure, the feature parameters of adjacent nodes are used for interpolation estimation (such as inferring the puncture angle through the data of the wrist motion sensor). In the path reconstruction stage, the system re-plans the teaching path based on the duration of the interference event (such as the network delay lasts for 5 seconds) and the abnormal risk level: if the interference duration is short (<10 seconds), then skip the non-critical links and directly resume the current step after the interference is eliminated; if the duration is long (≥10 seconds), then insert a review link to quickly review the learned steps to ensure the coherence of the learner's knowledge chain.

[0055] The collaborative effect of the construction conditions of teaching nodes and the polymorphic switching link also manifests in the smooth transition of the teaching stage. When teaching progresses from the basic nursing stage (B01) to the specialized nursing stage (Z02), the system first verifies the construction conditions of the teaching nodes in the new stage: checks whether the preset interval of the characteristic parameters of the specialized operation is loaded (such as the catheter insertion depth interval for "tracheal intubation"), whether the synchronization index of the corresponding node meets the standard (confirmed through pre-communication tests), and whether the stage identification deviation is within the allowable range (the Hamming distance between B01 and Z02 is 2, and when it exceeds the allowable range, it needs to be phased over, such as switching to B02 first and then transitioning to Z02). If the conditions are not met, the system automatically delays the stage switch until the node initialization and parameter calibration are completed, avoiding process interruptions caused by unready teaching nodes.

[0056] The hardware implementation of the polymorphic switching link is based on the dynamic routing mechanism of a programmable logic controller (PLC). The steady-state teaching sub-link and the anomaly avoidance sub-link correspond to different data processing threads, and priority management is achieved through a thread scheduler. When the system is in a steady state, the steady-state sub-link thread has the highest priority and preferentially obtains computing resources; when the anomaly avoidance sub-link is activated, the scheduler automatically raises its thread priority to the highest and temporarily freezes non-critical threads in the steady-state sub-link (such as the high-definition rendering thread of the operation video) to ensure the real-time nature of anomaly handling. This resource scheduling strategy is implemented through the real-time kernel of the operating system, and the thread switching delay is controlled within 10 milliseconds.

[0057] The parameter self-adaptation mechanism of the feature parsing unit further enhances the compatibility of the system. For learners of different body types and strengths, the system establishes personalized characteristic parameter intervals for each learner through historical data statistics. For example, considering the generally smaller strength of female learners, the preset interval of the injection force for the "intramuscular injection" operation can be dynamically adjusted to 80%-100% of the standard value (the standard interval is 5-8N, and after adjustment, it is 4-8N). When calculating the synchronization index, the learner's individual number is introduced as a weight factor, and the formula is:

[0058] In the formula, is the adjusted synchronization index, is the individual difference coefficient (the value range is -0.1 to 0.1), is the hash value modulus of the learner's individual number (used to quantify individual differences). This mechanism enables the system to adapt to the operation habits of different learners and avoid misjudgments caused by individual differences.

[0059] Example 4: In this embodiment, the multi-modal fusion model configured by the policy modeling node integrates the spatio-temporal characteristics of teaching data and the exception handling rules, and the process of generating the benchmark teaching policy covers the entire process of feature analysis, graph construction, and risk correction. The multi-modal fusion model includes two core data parts: one is the feature-time mapping table corresponding to the teaching data collection module, which records the standard distribution patterns of various feature parameters (such as needle insertion angle, disinfection time, action coherence) in the time dimension during nursing operations. For example, during the "venous puncture" operation, the time for the needle insertion angle to increase from 0° to 30° should be controlled within 0.5 - 1 second; the other is the exception avoidance parameter library of the teaching path generation module, which stores the coping strategies for common exception scenarios in historical teaching, such as the supplementary hand disinfection steps when the "aseptic operation process is interrupted" and the step-back guiding logic when the "operation sequence is reversed".

[0060] Taking the teaching scenario of "Cardiopulmonary Resuscitation (CPR)" as an example, the generation of the benchmark teaching policy is described. First, the system obtains the operation data of the learner in real time through the teaching data collection module, including feature data such as the depth of chest compressions (standard is 5 - 6 cm), frequency (100 - 120 times per minute), and the time ratio of compression to ventilation (30:2), as well as the dynamic clustering and sorting results (such as classifying the operations with abnormal compression frequencies into the "insufficient rhythm control" group). Based on these data, the feature analysis unit calculates the upper limit of the theoretical synchronization range for each teaching node in the steady-state teaching sub-link. For example, the upper limit of the synchronization range for the "compression depth node" is that the deviation between the measured value and the standard value does not exceed ±0.5 cm, and the upper limit of the synchronization range for the "compression frequency node" is that the deviation does not exceed ±5 times per minute.

[0061] Next, the upper limit of the theoretical synchronization range and the transient time tags (such as the time nodes of the 30th second, 60th second, etc. when the current operation is in progress) are input into the multi-modal fusion model. The model first calls the feature-time mapping table to retrieve the distribution of feature parameters of the standard CPR operation at the corresponding time points. For example, at the 30th second, the compression depth should be stable at about 5.5 cm and the frequency should be maintained at 110 times per minute. Based on these standard data, the model generates a chained teaching reference graph. This graph takes the time axis as the horizontal axis and the feature parameter dimension as the vertical axis, uses curves to mark the ideal change trajectories of each parameter, and represents the allowable fluctuation range of the parameters through color gradients (such as the green area is the normal range, yellow is the warning range, and red is the abnormal range). In the CPR example, the graph will show the ideal curve of the compression depth changing with time and the allowable interval of the frequency fluctuation.

[0062] After generating the reference atlas, the system corrects the risks through the anomaly avoidance sub-link. The anomaly avoidance sub-link first analyzes the instantaneous anomaly risks of the teaching path status. For example, it detects that the pressing depth of the learner suddenly drops to 4 cm (below the standard lower limit) at the 45th second and lasts for more than 2 seconds, triggering the anomaly risk of "insufficient pressing depth". At the same time, the sub-link queries the external interference event log to confirm whether there is equipment failure (such as signal fluctuation of the pressure sensor) or learner operation error (such as reduced strength due to arm fatigue). If it is determined as an operation error, the anomaly avoidance sub-link calls the "depth deficiency correction strategy" in the anomaly avoidance parameter library: superimpose a red warning mark at this time node in the atlas, generate a voice prompt "Please increase the pressing force" at the same time, and reproduce the depth change of the standard pressing action through a semi-transparent virtual image in the virtual scene to guide the learner to adjust.

[0063] The dynamic adaptability of the multi-modal fusion model is reflected in the strategy adjustment for different teaching stages. Taking the "intravenous infusion" teaching in the "basic nursing" and "specialty nursing" stages as an example: in the basic stage, the benchmark teaching strategy generated by the model focuses on the integrity of the operation process. The ideal value of the "exhaust time" in the reference atlas is 8 - 10 seconds, and the allowable fluctuation range is relatively wide (±3 seconds). The anomaly avoidance strategy mainly provides step-by-step prompts; while in the specialty stage, the strategy turns to operation efficiency and accuracy. The ideal value of the exhaust time is compressed to 5 - 7 seconds, and the allowable fluctuation range is narrowed (±1.5 seconds). The anomaly avoidance strategy adds real-time feedback from the pressure sensor, and dynamically displays the operation effect through the liquid level fluctuation of the virtual infusion tube to strengthen the learner's control over details.

[0064] The update mechanism of the feature-time mapping table is based on the continuous accumulation of teaching data. The system regularly collects historical data of each teaching node, classifies and statistically analyzes them according to the operation type (such as injection, puncture, first aid) and learner level (beginner, intermediate, advanced), and generates a probability distribution model of each feature parameter. For example, by analyzing the "intradermal injection" data of 1000 beginner learners, it is found that the average value of the injection angle is 12° (the standard value is 5° - 15°), and the standard deviation is 3°. Accordingly, the ideal curve of this operation in the "angle-time" dimension in the feature-time mapping table is adjusted to make it closer to the actual operation level of beginners and avoid learning frustration caused by overly strict standards.

[0065] The risk correction process of the abnormal avoidance sub-link is hierarchical. Taking the example of "incorrect order of wearing sterile gloves" in the teaching of "urethral catheterization": the first-level correction is visual prompt (marking the correct wearing order with arrows on the virtual glove icon), the second-level correction is voice explanation (detailedly explaining the aseptic principle of each step of the operation), and the third-level correction is to force a pause in the operation and start the step-by-step practice module. The system automatically upgrades the correction level according to the number of occurrences of the abnormality. If the same learner makes mistakes in the order three times during the practice, the sub-link will determine it as a stubborn problem, trigger a personalized training strategy, add targeted simulation scenarios (such as a virtual link for practicing glove wearing alone), and adjust the parsing focus of the subsequent teaching path to prioritize strengthening the monitoring of the characteristic parameters of the aseptic operation process.

[0066] The collaborative operation of the strategy modeling node and the scenario reconstruction unit realizes the visual implementation of teaching strategies. After generating the benchmark teaching strategy, the scenario reconstruction unit dynamically adjusts the element presentation of the virtual teaching scenario according to the time-characteristic mapping relationship in the reference atlas. For example, at the time node of "compression-ventilation switch" in CPR teaching, the virtual human model will automatically trigger the chest recoil animation. At the same time, the timer in the scenario highlights the ventilation duration (the standard is 1 second for each ventilation), and the bubble prompt box shows the synchronization status of the current operation (such as "compression frequency synchronization rate 95%", "ventilation time deviation +0.2 seconds"). This multi-modal feedback mechanism enables learners to intuitively perceive the difference between their operation performance and the standard strategy and make adjustments in a timely manner.

[0067] The multi-modal fusion model also supports cross-node feature correlation analysis. In the teaching of "operating room nursing cooperation", the model can correlate the time parameters of the "instrument transfer node" with the process parameters of the "surgical step node". For example, analyze whether the time taken to transfer the scalpel matches the time interval of the skin incision step. If it is found that most learners need to wait for more than 2 seconds after transferring the instrument before starting the incision operation, the model will automatically adjust the benchmark teaching strategy and insert a 0.5-second prompt delay between the transfer link and the incision link to help learners establish a more compact process connection awareness. This strategy optimization based on data correlation breaks through the limitations of single-node evaluation and improves the integrity and practicality of teaching strategies.

[0068] Example 5: In this embodiment, the parameter update mechanism of the multi-modal fusion model and the degradation processing logic of the path atlas correction module form a multi-level abnormal response system of the system to ensure the continuity and stability of the teaching process in complex interference or node failure scenarios. Taking the teaching of "venous puncture" as an example, the system dynamically triggers different levels of abnormal avoidance instructions by real-time monitoring the teaching scenario data and node status, and realizes the adaptive adjustment of the strategy in combination with the path atlas correction.

[0069] When the teaching data acquisition module detects that a learner's puncture angle exceeds the preset range for 5 consecutive times (for example, the standard is 30° - 45°, and the measured value is 20° - 25°) or the path synchronization rate suddenly drops (for example, the data transmission delay between nodes increases from 10 ms to 50 ms), the first abnormal avoidance instruction is triggered. The system immediately activates the alternative teaching unit, such as switching from the real-time virtual operation scenario to the playback of a pre-recorded standard operation video, and adjusts the parsing granularity of the teaching path generation module, simplifying the parsing from three-dimensional space coordinates to two-dimensional angle measurement, reducing the computational load to quickly respond to the anomaly. At the same time, the multi-modal fusion model marks the anomaly type as "angle deviation", and records the time point and magnitude of the deviation in the feature-time mapping table to accumulate data for subsequent model training.

[0070] If the learner's puncture angle still fails to return to the allowable range (for example, continuously deviating from the standard value by more than 10°) after the execution of the first abnormal avoidance instruction, the system triggers the second abnormal avoidance instruction. At this time, the feature parsing unit automatically switches to the abnormal avoidance sub-link through the polymorphic switching link, and pauses the original steady-state teaching process. The abnormal avoidance sub-link reallocates the topological weights of the teaching path state based on the duration of the external interference event (such as the network delay has lasted for 8 seconds). For example, it increases the evaluation weight of the "puncture angle node" from the default 30% to 50%, while reducing the weight of the "needle-holding posture node" to 10%, focusing on solving the core problem. The auxiliary line for angle correction (such as a semi-transparent 30° standard angle guide) is synchronously displayed in the virtual scene, and the puncture resistance at the correct angle is simulated through the force feedback device to help the learner establish muscle memory.

[0071] When the synchronization index of any level of teaching node (such as the "tourniquet tying time node") in the teaching data acquisition module is lower than the threshold (such as dropping from 0.9 to 0.7), the third abnormal avoidance instruction is triggered. The system immediately blocks the path output channel of this node to prevent low-synchronization data from affecting the overall teaching evaluation, and transfers the corresponding data (such as abnormal data with a tying time of 60 seconds) to other parallel nodes (such as the "operation process sequence node") for temporary storage through the behavior isolation unit. For example, in the "intramuscular injection" teaching, if the "disinfection time node" has abnormal synchronization, the collected disinfection duration data (such as only 10 seconds, the standard is 15 - 20 seconds) will be transferred to the "aseptic operation specification node" for secondary verification to confirm whether it is a random error or an operation mistake through historical data comparison.

[0072] During the execution of the third exception avoidance instruction, if other teaching nodes are detected to be overloaded (for example, the data volume processed by the "needle insertion depth node" exceeds 120% of the rated value), the fourth exception avoidance instruction is triggered. The path map correction module intervenes to perform a downgrade process: First, based on the synchronization rate data output by the chain stability evaluation network (for example, the overall synchronization rate drops from 92% to 85%) and off-chain characteristic indicators (for example, the CPU load rate reaches 85%), a path status safety level table is constructed, and the teaching path is divided into three levels: critical path (such as operation process compliance), sub-critical path (such as action coherence), and non-critical path (such as operation posture aesthetics). When the fourth instruction is triggered, the system dynamically downgrades the priority of the teaching path according to the safety level table, temporarily turns off the real-time evaluation of the non-critical path (such as hiding the feedback prompt of posture aesthetics), and synchronously adjusts the parsing frequency of the teaching path generation module from 20 times per second to 10 times per second to match the downgraded teaching strategy and avoid system crashes.

[0073] Taking the sudden scenario in the "aseptic technique" teaching as an example: When multiple learners perform the "laying out an aseptic tray" operation simultaneously, the system may cause the synchronization index of the "instrument placement order node" to drop sharply (to 0.6) due to a sudden increase in data acquisition volume, triggering the third exception avoidance instruction to block the node and transfer the data to the "aseptic towel folding node". If the "aseptic towel folding node" becomes overloaded due to additional load at this time (the processing time extends from 50 ms to 100 ms), triggering the fourth exception avoidance instruction, the system gives priority to ensuring the evaluation of the critical path (such as the alignment of the edges of the aseptic towel), downgrades the sub-critical path (such as the folding speed), and replaces the dynamic animation showing the instrument placement order with a text list in the virtual scene to reduce the consumption of graphic rendering resources.

[0074] The downgrade processing logic of the path map correction module is also reflected in the long-term optimization of teaching strategies. For example, in the "emergency care" teaching, if the "defibrillator operation node" frequently triggers the third and fourth exception avoidance instructions due to equipment driver failures, the system analyzes the historical downgrade records and finds that the synchronization stability of this node is insufficient in high-load scenarios. Then it automatically adjusts the benchmark teaching strategy: switches the simulation scenario of defibrillator operation from full simulation mode to process guidance mode to reduce the amount of real-time data interaction, and at the same time adds a single-person training session for this node in the teaching plan to disperse the system pressure during collective operations.

[0075] The parameter update of the multi-modal fusion model has memory and heredity. When a certain type of anomaly (such as "puncture angle deviation") frequently triggers the first and second anomaly avoidance instructions, the model will automatically adjust the angle threshold range of this operation in the feature-time mapping table (such as expanding from 30° - 45° to 25° - 50°, and marked as "beginner-friendly interval"), and synchronize this adjustment to the teaching strategies of all primary learners. If subsequent data shows that this adjustment leads to a decrease in the operation pass rate, the model will roll back the parameters and activate a more stringent anomaly detection mechanism to form a closed loop of "detection - adjustment - verification - optimization".

[0076] In the team teaching scenario of "operating room nursing cooperation", the system may face synchronous anomalies at multiple nodes simultaneously (such as data delays occurring at both the "instrument transfer node" and the "surgical record node"). At this time, according to the safety level table, the path map correction module gives priority to ensuring the critical path evaluation of the "instrument transfer node" (which directly affects the safety of the surgical process), degrades the "surgical record node" (such as allowing the recording time to be delayed until after the operation for supplementary recording), and caches the delayed data through the message queue mechanism for batch processing after the system load decreases, ensuring that the core teaching links are not affected.

[0077] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

[0078] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A nursing practice teaching demonstration system based on big data, characterized in that, Including: A teaching data collection module, a teaching path generation module, and a dynamic evaluation module; The teaching data collection module includes a preprocessing node and a multi-dimensional correlation engine; The teaching path generation module includes a feature analysis unit and a scenario reconstruction unit; the dynamic evaluation module includes a strategy modeling node and a feedback optimization unit; The preprocessing node is used to obtain teaching scenario data and operation behavior parameters in real time; According to the teaching stage identifier and the time series label, a multi-dimensional teaching path topology is constructed through the multi-dimensional correlation engine; the multi-dimensional correlation engine is composed of multiple teaching nodes based on feature thresholds connected in parallel; based on the real-time response characteristics of multiple teaching nodes, a benchmark teaching strategy is generated through the strategy modeling node, and combined with the reconstruction signal output by the scenario reconstruction unit, the feedback optimization unit performs dynamic clustering and sorting on the teaching path state.

2. The nursing practice teaching demonstration system based on big data according to claim 1, wherein The system further includes a feedback optimization module and a path map correction module; the feedback optimization module includes a behavior isolation unit and an error compensation unit; According to the generated benchmark teaching strategy and the dynamic clustering and sorting results, the behavior isolation unit performs distributed anomaly correction on the teaching path state, and the error compensation unit dynamically calibrates the feature parameters of the teaching path generation module; at the same time, the chain stability evaluation network constructed by the path map correction module monitors the path synchronization rate in real time, and feeds back the monitoring results to the strategy modeling node and the feedback optimization unit to perform closed-loop adjustment on the teaching strategy.

3. The nursing practice teaching demonstration system based on big data according to claim 2, wherein, The behavior isolation unit is composed of state control circuits corresponding to the teaching nodes connected in parallel in the multi-dimensional correlation engine; The construction steps of the multi-dimensional teaching path topology include: Collect teaching scenario data and operation behavior fluctuation signals, input them into the transfer learning model configured for each level of teaching node, generate data consistency calibration coefficients and store them in the corresponding teaching node; Based on the consistency calibration coefficients stored in each level of teaching node, combined with the instantaneous change rate of the time series label, the teaching path state is assigned to multiple teaching nodes through the dynamic feature detection algorithm to form a dynamic path matching topology.

4. The nursing practice teaching demonstration system based on big data according to claim 3, characterized in that, The construction conditions of the teaching node configured by the preprocessing node include: the teaching scenario data is within the preset feature interval, the synchronization index corresponding to the teaching node is greater than the threshold, and the teaching stage identifier is less than the allowable deviation range; The feature analysis unit is also configured with a polymorphic switching link; the polymorphic switching link includes a steady-state teaching sub-link and an anomaly avoidance sub-link; the steady-state teaching sub-link is generated based on the matching degree between the chain stability signal of the teaching path generation module and the benchmark teaching strategy; the anomaly avoidance sub-link is constructed based on the correlation between the instantaneous anomaly risk of the teaching path state and external interference events.

5. The nursing practice teaching demonstration system based on big data according to claim 4, characterized in that, The strategy modeling node is configured with a multi-modal fusion model; the multi-modal fusion model includes a feature-time mapping table corresponding to the teaching data collection module and an anomaly avoidance parameter library of the teaching path generation module; The steps for generating the benchmark teaching strategy include: According to the real-time feature data and the clustering and sorting results of the teaching data collection module, calculate the upper limit of the theoretical synchronization range of each level of teaching node in the steady-state teaching sub-link; Input the upper limit of the theoretical synchronization range and the transient time tag into the multi-modal fusion model to generate a chain-type teaching reference map, and correct the abnormal risk of the reference map through the abnormal avoidance sub-link to form a benchmark teaching strategy.

6. The nursing practice teaching demonstration system based on big data according to claim 5, characterized in that The parameter update steps of the multi-modal fusion model include: When it is detected that the teaching scenario data exceeds the preset interval or the path synchronization rate exceeds the threshold, trigger the first abnormal avoidance instruction, that is, start the alternative teaching unit and adjust the parsing granularity of the teaching path generation module; If the path state still has not returned to the permitted interval after the execution of the first abnormal avoidance instruction, trigger the second abnormal avoidance instruction, that is, switch to the abnormal avoidance sub-link through the polymorphic switching link, and re-allocate the topological weight of the teaching path state based on the duration of the external interference event.

7. The demonstration system for nursing practice teaching based on big data according to claim 6, wherein The parameter update steps of the multi-modal fusion model also include: When the synchronization index of any level of teaching node in the teaching data acquisition module is lower than the threshold, trigger the third abnormal avoidance instruction, that is, block the path output channel of the node, and transfer the corresponding data to other teaching nodes through the behavior isolation unit; During the execution of the third abnormal avoidance instruction, if it is detected that other teaching nodes are overloaded, trigger the fourth abnormal avoidance instruction, that is, call the path map correction module to degrade the output strategy of the feedback optimization unit, and limit the synchronization requirements of the edge state.

8. The demonstration system for nursing practice teaching based on big data according to claim 7, characterized in that, The degradation processing logic of the path map correction module includes: Construct a path state safety level table according to the synchronization rate data output by the chain-type stability evaluation network and the off-chain feature indicators; When the fourth abnormal avoidance instruction is triggered, dynamically degrade the teaching path priority according to the safety level table, and synchronously adjust the parsing frequency of the teaching path generation module to match the degraded teaching strategy.

9. The nursing practice teaching demonstration system based on big data according to claim 3, characterized in that The construction steps of the teaching node also include: Generate a feature gradient adaptability index table based on the spatial distribution result of the teaching scenario data and the historical operation behavior range; According to the difference in the consistency calibration coefficients of each teaching node in the index table, perform topological weight allocation on the state control circuits of multiple teaching nodes through the dynamic weight allocation algorithm, and establish redundant transmission links between teaching nodes.

10. The nursing practice teaching demonstration system based on big data according to claim 5, characterized in that, The generation process of the chain-type teaching reference map includes: After inputting the upper limit of the theoretical synchronization range and the transient time tag into the multi-modal fusion model, synchronously load the preset teaching medium database and the external interference event layer; Based on the encrypted peak frequency data output by the feature parsing unit, correct the abnormal threshold of the teaching path state in the reference map, and perform secondary calibration by superimposing the time gradient parameter in the abnormal avoidance sub-link.

Images