Newborn nursing skill training method, system and device based on virtual reality technology and medium

Abstract

The invention relates to a newborn nursing skill training method, system and device based on a virtual reality technology and a medium. The method comprises the following steps: based on a newborn physiological digital twinborn basic model set, performing parameter identification generation on newborn personalized indexes to obtain a personalized newborn digital twinborn model; based on a scene library, according to the current node of the training path, generating a VR scene in combination with a personalized neonatal digital twin model; vR tracking equipment action data flow and real-time physiological data flow are acquired, and a newborn physiological response state is calculated based on the personalized newborn digital twinborn model; reconstructing audio-visual feedback corresponding to the VR scene according to the physiological response state of the newborn; and in response to an end mark of the scene time window, calculating an evaluation index according to the action data flow and the real-time physiological data flow of the VR tracking equipment, and obtaining a training result report. By adopting the method, the practice times and scene coverage can be amplified in a virtual environment, and the dependence on a real scene is reduced.

Description

Technical Field

[0001] This invention belongs to the field of simulation training technology, and in particular relates to a method, system, equipment and medium for training newborn care skills based on virtual reality technology. Background Technology

[0002] With the development of Virtual Reality (VR) technology, skills training technology based on immersive simulation environments has emerged. Through 3D modeling, interactive perception, motion capture, and real-time feedback, highly realistic operation scenarios can be constructed, allowing trainees to obtain sensory experiences and operational feedback similar to those in a real environment in a virtual space. This breaks through the bottleneck of limited traditional physical teaching resources and is widely used in many fields such as surgical training, industrial assembly training, and fire safety drills.

[0003] However, in neonatal care training, existing training methods mainly rely on verbal instructions from medical staff, reading of written materials, or short-term practical observation in a hospital environment. In traditional techniques, trainees usually practice basic care steps by observing demonstrations by professional nurses or on static infant models. They lack interactive experiences that match the reactions of real infants and cannot simulate responses to sudden emergencies. Trainees often do not have the opportunity to practice safely and repeatedly, and can only rely on theoretical understanding and limited situational demonstrations. Summary of the Invention

[0004] Therefore, it is necessary to address the aforementioned technical issues by providing a method, system, equipment, and medium for newborn care skills training based on virtual reality technology that can achieve high simulation, personalization, data-driven, and quantifiable results.

[0005] Firstly, this application provides a method for training newborn care skills based on virtual reality technology, including:

[0006] Personalized indicators for newborns are obtained, and based on the basic set of newborn physiological digital twin models, parameters are identified and generated for the personalized newborn indicators to obtain a personalized newborn digital twin model. The personalized newborn digital twin model includes a set of variable parameters that characterize the physiological characteristics of newborns. The set of variable parameters includes, but is not limited to, infant weight, lung compliance, baseline waveforms of heart rate and blood oxygen saturation, and sensitivity thresholds for response actions.

[0007] Based on the scene library, VR scenes are generated according to the current node of the training path and combined with the personalized newborn digital twin model; the VR scene includes sudden emergency events within the preset scene time window and their corresponding noise levels, available devices and interference;

[0008] The system acquires VR tracking device motion data streams and real-time physiological data streams of trainees in response to VR scenarios, and calculates the newborn's physiological response status in real time based on a personalized newborn digital twin model, using the VR tracking device motion data streams as operational input.

[0009] Reconstruct the audiovisual feedback corresponding to the VR scene based on the physiological response state of the newborn;

[0010] In response to the end of the scene time window, evaluation metrics are calculated based on the VR tracking device motion data stream and real-time physiological data stream to obtain a training result report; the evaluation metrics include key performance indicators and physiological stress state.

[0011] In one embodiment, before obtaining a personalized newborn digital twin model by acquiring newborn personalized indicators and generating parameters for these indicators based on a newborn physiological digital twin basic model set, the process further includes:

[0012] Acquire the resource library; the resource library includes a scene library, an operation action library, a concurrent event atomic library, a basic set of neonatal physiological digital twin models, and training path configurations; the scene library includes metadata of each sudden emergency event and its corresponding scene element; the operation action library includes first aid actions and their corresponding quantifiable action features; quantifiable action features include timing, force, angle, and contact position; the concurrent event atomic library includes concurrent emergency events and their configuration trigger modes and probability distributions; the basic set of neonatal physiological digital twin models includes multi-scale models and their parameterized templates;

[0013] Acquire relevant data about the trainees and perform baseline calibration based on the data to obtain baseline physiological data, baseline movement samples, and training paths. The relevant data about the trainees includes trainee identity, training objectives, baseline knowledge test results, basic physiological data, and initial position data of the VR tracking device.

[0014] In one embodiment, based on a personalized newborn digital twin model, the newborn's physiological response state is calculated in real time using VR tracking device motion data stream as operational input, including:

[0015] The motion data stream of the VR tracking device is analyzed to extract operational features, including timing, contact position, force intensity, and motion angle.

[0016] Operational features are input into a personalized newborn digital twin model to obtain instantaneous physiological parameters corresponding to each time step; the personalized newborn digital twin model performs physiological and pathological evolution calculations on the operational feature input based on a set of variable parameters;

[0017] The various instantaneous physiological parameters are aggregated into the newborn's physiological response state; the newborn's physiological response state includes, but is not limited to, the result of the action response, heart rate, blood oxygen saturation, respiratory rate and body temperature.

[0018] In one embodiment, in response to the end marker of the scene time window, evaluation metrics are calculated based on the VR tracking device motion data stream and real-time physiological data stream to obtain a training result report, including:

[0019] The motion data stream of the VR tracking device is individually calibrated based on baseline motion samples to obtain calibrated motion performance characteristics;

[0020] Based on the quantifiable action features defined in the action library, the action performance features are compared with the corresponding quantifiable action features to obtain key performance indicators; key performance indicators include, but are not limited to, action positioning accuracy, force and force deviation, action timing and continuity, key response delay and process integrity.

[0021] Based on baseline physiological data, physiological variability is calculated using real-time physiological data, and physiological stress status is obtained by assessing physiological variability.

[0022] Training results are reported based on key performance indicators and physiological stress states.

[0023] In one embodiment, reconstructing the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state further includes:

[0024] Based on a sliding time window with a preset step size, short-term key performance indicators are calculated according to the real-time VR tracking device motion data stream, and the physiological and pathological evolution rate of the personalized newborn digital twin model is updated according to the short-term key performance indicators; the physiological and pathological evolution rate is used to characterize the rate at which physiological response state quantities change over time.

[0025] The short-term physiological stress state corresponding to the short-term key performance indicators is calculated based on the real-time physiological data stream, and the concurrent event trigger probability is calculated based on the short-term key performance indicators and the short-term physiological stress state; the trigger probability is obtained by combining the probability distribution of the concurrent event atom library that changes over time with the adjustment factor based on the current training performance;

[0026] When the concurrent event trigger probability meets the configured trigger mode, the parameter configuration corresponding to the concurrent emergency event is extracted from the concurrent event atomic library and injected into the VR scene and the personalized newborn digital twin model;

[0027] In response to short-term physiological stress exceeding a preset stress threshold, the sudden emergency event in the VR scene is replaced with a calming training scene; the calming training scene includes breathing regulation and situational decompression; situational decompression includes generating action guidance prompts.

[0028] In one embodiment, the method further includes:

[0029] Based on key performance indicators, the VR tracking device's motion data stream and the corresponding motion videos of the newborn's physiological response state are visualized and annotated to obtain training feedback results; the visualized and annotated feedback includes annotating and quantifying the gap based on key performance indicators.

[0030] Weaknesses are identified based on the training results report, and the next node of the training path and the corresponding expected key performance indicators are adjusted accordingly.

[0031] In one embodiment, the method further includes:

[0032] In response to the end of the training path, the relevant data of the trainees are reacquired, and a new training path is obtained based on the training path configuration, training objectives and baseline knowledge test results.

[0033] Secondly, this application also provides a newborn care skills training system based on virtual reality technology, including:

[0034] The newborn personalization module is used to acquire newborn personalized indicators and, based on the newborn physiological digital twin basic model set, to identify and generate parameters for the newborn personalized indicators to obtain a personalized newborn digital twin model. The personalized newborn digital twin model includes a set of variable parameters that characterize the newborn's physiological characteristics. The set of variable parameters includes, but is not limited to, infant weight, lung compliance, heart rate and blood oxygen saturation waveform baseline, and response action sensitivity threshold.

[0035] The virtual reality module is used to generate VR scenes based on a scene library and the current node of the training path, combined with a personalized newborn digital twin model. The VR scene includes sudden emergency events within a preset scene time window and their corresponding noise levels, available devices, and interference.

[0036] The response evolution module is used to acquire the VR tracking device action data stream and real-time physiological data stream of the trainee in response to the VR scene, and calculate the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device action data stream as the operation input.

[0037] The real-time update module is used to reconstruct the audiovisual feedback corresponding to the VR scene based on the physiological response status of the newborn.

[0038] The training evaluation module is used to calculate evaluation indicators based on the VR tracking device's motion data stream and real-time physiological data stream in response to the end of the scene time window, and to obtain a training result report; the evaluation indicators include key performance indicators and physiological stress state.

[0039] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above-described methods for training newborn care skills based on virtual reality technology.

[0040] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for training newborn care skills based on virtual reality technology.

[0041] The aforementioned methods, systems, equipment, and media for neonatal care skills training based on virtual reality technology can present differentiated behaviors for different neonatal types through personalized neonatal digital twin models, training scenarios, and physiological responses. This makes the situations faced by trainees closer to their future actual encounters, thereby improving the adaptability and transferability of learning. By establishing a real-time closed loop from action to audiovisual reconstruction, trainees can immediately see the direct physiological consequences of their specific actions, promoting immediate correction based on results and deep causal cognition, thus improving operational accuracy and decision-making quality. By incorporating sudden emergency events and assessment mechanisms, trainees can repeatedly practice emergency handling procedures such as choking on milk and apnea without real risk of harm, significantly improving their proficiency and psychological preparedness for rare but critical events. The evaluation of action data streams and real-time physiological data streams and the generation of training result reports enable explicit quantification of training effectiveness, progress tracking, and traceable recording, facilitating course management, qualification assessment, and continuous improvement. Compared to extensive one-on-one demonstrations by instructors and real clinical observation, the neonatal care skills training method based on virtual reality technology can amplify the number of practice sessions and the scope of scenarios in a virtual and safe environment, reducing reliance on real hospital beds and real infants, thereby reducing teaching organization costs and potential risks. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart illustrating the newborn care skills training method based on virtual reality technology according to the present invention.

[0044] Figure 2 This is a flowchart illustrating the steps of step S103.

[0045] Figure 3This is a flowchart illustrating the steps of step S105.

[0046] Figure 4 This is a structural diagram of the newborn care skills training system based on virtual reality technology according to the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0048] In one embodiment, such as Figure 1 As shown, a method for training newborn care skills based on virtual reality technology is provided. This embodiment illustrates the method by applying it to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0049] S101. Obtain personalized indicators for newborns, and based on the basic set of newborn physiological digital twin models, perform parameter identification and generation on the personalized indicators to obtain a personalized newborn digital twin model. The personalized newborn digital twin model includes a set of variable parameters that characterize the physiological characteristics of newborns. The set of variable parameters includes, but is not limited to, infant weight, lung compliance, baseline waveforms of heart rate and blood oxygen saturation, and sensitivity threshold for response actions.

[0050] Indicatively, neonatal personalized indicators refer to the basic input dataset used to characterize the individual differences of the virtual infants being trained. This includes, but is not limited to, infant weight, gestational age, past medical history or characteristic markers, initial heart rate and blood oxygen saturation range, initial estimates of lung compliance, and initial sensitivity thresholds to external stimuli or manipulations. This ensures that the physiological behaviors presented in subsequent simulations are both universal and individualized, allowing the training scenario to reflect the differences in physiological responses between preterm and full-term infants, thereby improving the relevance and realism of the training. Specifically, a basic set of neonatal physiological digital twin models is retrieved from a resource library. This set contains a set of parameterized multi-scale physiological model templates, such as simplified mathematical models or parameterized curves describing sub-modules like respiration, blood flow, oxygenation, and metabolism. Furthermore, using the obtained neonatal personalized indicators as constraints or priors, the basic templates are mapped to individualized models through parameter recognition methods. Optionally, parameter identification can employ system identification or Bayesian updating to fit the observed baseline physiological data with the model output, determining the optimal values ​​of the variable parameter set. Reasonable parameter estimates can be generated based on weighted priors or template interpolation, resulting in a personalized neonatal digital twin model containing the variable parameter set and a dynamic model body. This model serves as the core computational unit that maps trainee actions to physiological consequences during runtime, providing a unified physiological semantic foundation for subsequent scene generation, real-time physiological computation, and training evaluation. For example, when weight and lung compliance parameters are input, the model can produce corresponding predicted curves of blood oxygen saturation and respiratory waveform changes upon receiving induced ventilation.

[0051] S102. Based on the scene library, generate a VR scene according to the current node of the training path and the personalized newborn digital twin model; the VR scene includes sudden emergency events within the preset scene time window and their corresponding noise levels, available devices and interference.

[0052] The scenario library is a modular collection of knowledge resources used to describe metadata for different training environments and contextual elements. It includes typical scenarios such as a family bedroom and a neonatal intensive care unit. Each scenario contains multiple templates for potential emergency events, a list of available equipment such as oxygen masks, suction devices, and monitors, noise level definitions, and potential interference factors such as visitor entry, equipment alarms, and lighting changes. The current node in the training path indicates the current teaching objective and difficulty level in the course design, such as basic exercises, complication identification, or high-pressure comprehensive response. Each node has different requirements regarding scenario complexity, concurrent event frequency, and prompting strategies. The purpose of combining personalized neonatal digital twin models with a scene library to generate VR scenes, as illustrated, is to maintain the realism of the context while ensuring that the neonatal responses within the scene are based on individualized physiological characteristics rather than a single template. This makes the situations faced by trainees more clinically valuable in terms of perception and decision-making. Specifically, the course manager reads the configuration of the current training node, selects or parameterizes matching scene elements from the scene library, and injects initial concurrent event configurations. Then, it uses the physiological parameter range provided by the personalized model to fine-tune the scene parameters, such as parameterizing the initial noise level, the operability of available devices, and the baseline probability of concurrent event triggering, forming the final scene description file sent to the VR engine. For example, for a training node in the daily care training path for premature infants, the scene generator will select a neonatal intensive care unit scene, set a higher oxygenation sensitivity threshold, and increase the baseline probability of hypothermia as a concurrent event, thereby encouraging trainees to practice more warming and temperature management.

[0053] S103. Obtain the VR tracking device motion data stream and real-time physiological data stream of the trainee in response to the VR scene, and calculate the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device motion data stream as the operation input.

[0054] Optionally, the motion data stream for VR tracking devices comes from head-mounted displays, hand / hand controller trackers, and optional external motion capture sensors, such as hand position, posture, tactile contact points, and force estimation. When combined with mixed reality (MR) mode, it can also come from a sensorized dummy infant equipped with chest pressure sensors, mask positioning sensors, etc. The aforementioned devices output position, angle, and motion events at a high frequency.

[0055] The trainees’ real-time physiological data streams come from wearable physiological sensors, such as heart rate, heart rate variability, skin conductance, breathing belts, etc., or external physiological monitoring devices, all of which are timestamped to support multi-source synchronization.

[0056] Indicatively, a low-latency data bus is established during runtime to receive and time-synchronize action and physiological data streams. The action streams are then converted into operational inputs for the digital twin model according to a scene mapping matrix, such as performing a mask compression, a suctioning action, or a body position adjustment. For example, when a trainee performs positive pressure ventilation on a newborn in a VR scene, the tracking device records the mask position and compression frequency. Using the mask position and compression frequency as inputs, the digital twin outputs changes in heart rate and blood oxygen saturation at the next time step.

[0057] S104. Reconstruct the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state.

[0058] Indicatively, a newborn's physiological response includes a series of instantaneous physiological values, such as heart rate, blood oxygen saturation, respiratory rate, body movement, or muscle tone. These instantaneous physiological values ​​are mapped to visual elements in the VR scene, such as changes in skin color and chest rise and fall, and auditory elements, such as breathing sounds, cries, or heart sounds, through a mapping function. If tactile feedback is present, it can also include tactile sensations such as vibrations or changes in resistance. Trainees receive immediate feedback on the quality of their actions through intuitive perception without viewing abstract instruments, thus adjusting their action strategies more effectively. Optionally, the mapping function can be a rule-based threshold mapping, such as displaying a bluish skin tone and adding breathing difficulties sound effects when blood oxygen saturation falls below a certain threshold; a parameterized lookup table; or a lightweight neural network that maps continuous physiological values ​​to complex sensory representations. For example, when the digital twin predicts a significant decrease in respiratory rate and blood oxygen saturation, the VR scene can instantly amplify the chest drop, play faint breathing sounds, and trigger a visual effect of paleness, allowing the trainee to quickly detect and take appropriate resuscitation actions.

[0059] S105. In response to the end marker of the scene time window, calculate the evaluation indicators based on the VR tracking device motion data stream and real-time physiological data stream, and obtain the training result report; the evaluation indicators include key performance indicators and physiological stress state.

[0060] Indicatively, the scene time window can be determined by the arrival of a preset duration, a predetermined task completion signal, or an event-based external command triggered by the occurrence of concurrent physiological and pathological evolution reaching a termination condition. After the scene time window ends, the VR tracking device motion data stream and real-time physiological data stream collected during this session are aggregated and processed to calculate evaluation indicators for quantifying training effectiveness. Specifically, by eliminating the influence of scale differences such as height and arm length on motion quantification in the motion data stream, key performance indicators (KPIs) are calculated based on the target motion standard, such as motion positioning accuracy, force / intensity deviation, motion timing and continuity, key response delay, and process integrity. At the same time, the psychological and physiological load is assessed based on real-time physiological data to obtain the physiological stress state. Optionally, the KPIs and physiological stress state are aggregated into a training result report according to preset or configurable weighting rules. The report includes both quantitative scores and targeted suggestions or training modules that need to be strengthened. This report is written into the trainee's training file for the course manager to use for the next round of training path adjustment. For example, if the report shows that the trainee's mask fit positioning error increases and heart rate variability decreases under high noise interference, it may be recommended to add repetitive exercises under low noise or calm training scenarios to the next training node to improve stress regulation ability.

[0061] In the aforementioned neonatal care skills training method based on virtual reality technology, the generation of personalized models through parameter recognition based on the neonatal physiological digital twin basic model set can quickly customize general physiological models into individualized simulations, achieving accurate modeling of different physiological baselines and sensitivity thresholds. VR scenes are generated by combining a scene library with the current node of the training path using personalized models. Scene configuration can simultaneously reflect teaching objectives and individualized physiological characteristics, thus achieving precise matching of context and difficulty. Synchronous VR tracking action data streams and real-time physiological data streams are collected, achieving temporal coupling of multimodal data, enabling the quantification, reconstruction, and playback of the operation-physiological causal chain. Using the action data stream as input to drive personalized digital twins and calculate physiological response states in real time, operational behaviors are directly transformed into physiological effects, allowing trainees to verify whether their actions are physiologically effective or harmful during training, promoting integrated learning of skills and physiological benefits. Reconstructing audiovisual feedback based on physiological response states can improve trainees' speed of perceiving physiological abnormalities and the quality of their decision-making responses through intuitive multi-sensory presentation. Based on motion and physiological data, KPIs and physiological stress states are calculated and training result reports are generated to provide objective, comparable, and traceable performance measurement, supporting quantitative assessment, level certification, and the development of targeted training plans.

[0062] In one embodiment, before obtaining a personalized newborn digital twin model by acquiring newborn personalized indicators and generating parameters for these indicators based on a newborn physiological digital twin basic model set, the process further includes:

[0063] S11. Obtain the resource library; the resource library includes a scene library, an operation action library, a concurrent event atomic library, a basic set of neonatal physiological digital twin models, and training path configuration; the scene library includes metadata of each sudden emergency event and its corresponding scene element; the operation action library includes first aid actions and their corresponding quantifiable action features; quantifiable action features include timing, force, angle, and contact position; the concurrent event atomic library includes concurrent emergency events and their configuration trigger modes and probability distributions; the basic set of neonatal physiological digital twin models includes multi-scale models and their parameterized templates.

[0064] Indicatively, the resource library comprises several sub-libraries: a scenario library, an operation action library, a concurrent event atomic library, a basic set of neonatal physiological digital twin models, and a training path configuration. The scenario library defines different training environments and their components. Each emergency event in a scenario is described by metadata, which records the event's typical triggering conditions, duration, interaction with other scenario elements, configurable noise levels, and potential interference factors. The operation action library collects first aid and nursing actions in a structured format and defines quantifiable action characteristics for each action, such as action sequence, range of force, reference angle, and contact point location, for use in action recognition and performance comparison during runtime. The concurrent event atomic library decomposes complex situations into atomic events and their triggering patterns and probability distributions, enabling the system to generate concurrent pathological sequences in a controllable and combinable manner. The basic set of neonatal physiological digital twin models provides a set of parameterized multi-scale model templates and parameterized interpolation methods as a starting point for building personalized models. The training path configuration defines the course nodes, objectives, difficulty levels, and evaluation criteria.

[0065] Specifically, templates and metadata can be encoded into machine-readable configuration files through expert knowledge engineering and clinical data processing, and maintained using version control. Before system deployment, the engineering team and medical experts will jointly review and annotate each sub-library entry. For example, home care scenarios in the scenario library can include infant bathing, massage, and feeding. The feeding scenario includes metadata items such as: indoor background noise = 35 decibels, available equipment = {feeding bottle, warmer (off)}, and potential concurrent event template = {choking: baseline probability 0.02, triggering condition: improper feeding posture}.

[0066] S12. Obtain relevant data of the trainees and perform baseline calibration based on the relevant data of the trainees to obtain baseline physiological data, baseline movement samples and training paths; the relevant data of the trainees include trainee identity, training objectives, baseline knowledge test results, basic physiological data and initial position data of VR tracking device movements.

[0067] As an example, an individualized initial profile is created for each trainee to enable personalized calibration and fair evaluation of motor performance and physiological responses during training. Trainee-related data includes trainee identity information, training objectives, baseline knowledge test results, basic physiological data, and initial position data from the VR tracking device. The trainee can be a neonatal healthcare worker or a newborn's parent.

[0068] Furthermore, considering the differences in physical stature, existing knowledge level, and psychological state among trainees, baseline calibration typically consists of two parts: information and ability baseline collection, and movement and physiological baseline collection. Information and ability baseline collection obtains the trainee's training goals, past experience, and theoretical knowledge level through questionnaires or short knowledge tests. Movement and physiological baseline collection collects data through short tasks—for example, a set of standardized movements performed after demonstration, such as a standard hugging posture or simulated mask placement—while the trainee wears physiological sensors and VR tracking devices. This data records the trainee's natural movement trajectory without external interference, the initial sensor baseline level, and the initial positional offset and calibration parameters of the VR tracking device. The collected baseline physiological data is used to form the prior distribution of a personalized model or as a reference range for subsequent physiological stress assessment. Baseline movement samples are used for individualized calibration during the assessment phase, such as considering differences in movement curvature due to height. The training path selects or generates corresponding course nodes in the system based on the trainee's abilities and goals.

[0069] Optionally, the calibration process should ensure repeatability and traceability, with collected data timestamped and written into the trainee's file for later retrieval. For example, during calibration, the system may require the trainee to complete three standard hold positions and record the relative hand positions and duration each time, while simultaneously collecting heart rate and skin conductance as resting reference values. These records will constitute the trainee's baseline movement sample and baseline physiological data for subsequent movement calibration and stress threshold setting.

[0070] In one embodiment, such as Figure 2 As shown, based on a personalized newborn digital twin model, the physiological response state of the newborn is calculated in real time using the motion data stream from a VR tracking device as the input, including:

[0071] S201. Analyze the motion data stream of the VR tracking device to extract operation features; operation features include timing, contact position, force intensity and motion angle.

[0072] Indicatively, the motion data stream of a VR tracking device consists of the posture of the head-mounted display, the 3D position-pose information of the hand or controller trackers, and force / torque readings from haptic sensors or force estimators. Specifically, the VR tracking device motion data stream undergoes preprocessing, including time synchronization, signal denoising, drift correction, and coordinate transformation. Further, motion segmentation and event detection methods are used to decompose the continuous stream into several basic motion segments or events, such as a suctioning action or a mask compression process. Segmentation can be triggered by threshold detection, energy envelope, or short-time window statistics. For each motion segment, a structured operational feature vector is extracted, including the motion timing (start and end times and duration), contact position (coordinates of the contact point in the scene / model coordinate system), force / intensity, motion angle, motion rate and acceleration, and motion completion indicators. To improve robustness, the operational feature extraction employs a multi-sensor fusion strategy, fusing inertial information (IMU) with optical / position tracking results using filters to obtain stable feature estimates even in the presence of occlusion or short-term loss. For example, in a simulated mask sealing action, the analysis module will identify the starting point of the action, the time sequence of mask position and force changes during the continuous process, and output key instantaneous features such as average pressing force of 2.8N, maximum displacement error of 12mm, and mask sealing angle of 5°.

[0073] S202. Input the operational features into the personalized newborn digital twin model to obtain the instantaneous physiological parameters corresponding to each time step; the personalized newborn digital twin model performs physiological and pathological evolution calculations on the operational feature input based on the variable parameter set.

[0074] The personalized neonatal digital twin model inherits a pre-generated set of variable parameters. Internally, several sub-modules work collaboratively to simulate the neonatal physiological processes such as respiration, circulation, oxygenation, and body temperature. The purpose of using operational features as model input is to transform the mechanical / temporal characteristics of nursing actions into driving effects on the infant's physiology through physical or semi-physical mapping, thereby generating predictable physiological responses. Specifically, input-influence mapping rules or control interfaces are defined for each type of action. For example, the peak force and fit of mask compression are mapped to an adjustment coefficient for airway ventilation efficiency; the depth and duration of suctioning are mapped to airway cleanliness or short-term heart rate fluctuations caused by stimulation. Furthermore, for model calculation, iterative updates are performed using discrete time steps. At each time step, the model calculates instantaneous physiological parameters for the next time step, such as heart rate, blood oxygen saturation, respiratory rate, airway resistance, alveolar ventilation, and estimated body surface temperature, based on the current set of variable parameters, the physiological state of the previous step, and the current action input. For example, based on closed-loop or iterative solutions to simplified physical equations, the physics module can also be combined with a lightweight machine learning residual model. To ensure real-time performance, the model computation uses a numerically stable and computationally cost-controlled solver, and is equipped with a safety rule layer to provide conservative default responses or alarms in the event of input anomalies or model overflow. For example, under a positive pressure ventilation input, the model converts the pressure and frequency into hourly changes in alveolar ventilation over several consecutive time steps, thereby calculating the corresponding curves of slow rise or fall in blood oxygen saturation and the immediate impact on heart rate, thus outputting a series of instantaneous physiological parameters.

[0075] S203. Aggregate the instantaneous physiological parameters into the newborn's physiological response state; the newborn's physiological response state includes, but is not limited to, action response results, heart rate, blood oxygen saturation, respiratory rate and body temperature.

[0076] Schematic, instantaneous physiological parameters are several scalar or vector values ​​output by the model at each discrete time step. Aggregating instantaneous values ​​forms a higher-level understanding of the neonatal physiological response state. Specifically, the aggregation process typically includes short-term smoothing / filtering such as sliding window averaging or exponentially weighted moving averages to reduce measurement noise. Further, based on baseline physiological data, instantaneous values ​​are standardized by normalizing them to a percentage change relative to the individual baseline or a Z-score, followed by event detection and state labeling. Optionally, aggregation can calculate several derived indicators, such as motor response results, physiological fluctuation amplitude, and short-term recovery rate, used to drive audiovisual / tactile reconstruction, and as the raw data source for calculating KPIs and physiological stress states in the assessment module. Furthermore, the aggregation process needs to explicitly define the time window length, such as a short window of 5-30 seconds for detecting immediate responses, a medium window of 30-120 seconds for assessing recovery trends, and applying a priority strategy for sudden events, meaning events with severe physiological thresholds can bypass smoothing and directly trigger warnings and concurrent event injections. For example, if the instantaneous blood oxygen saturation output by the model is lower than the baseline limit three times consecutively after a suctioning operation, the aggregation module can label the neonatal physiological response state as a short-term hypoxia event and calculate the response time.

[0077] In one embodiment, such as Figure 3 As shown, in response to the end marker of the scene time window, evaluation metrics are calculated based on the VR tracking device's motion data stream and real-time physiological data stream to obtain a training result report, including:

[0078] S301. Perform individualized calibration of the VR tracking device motion data stream based on baseline motion samples to obtain calibrated motion performance characteristics.

[0079] Indicatively, baseline motion samples are reference trajectories and biomechanical characteristics of trainees acquired during the calibration phase within a standardized set of motions. These samples characterize the trainee's natural movement scale and habitual biases under conditions free from external interference. Individualized calibration aims to eliminate or reduce systematic biases caused by differences in trainee physique, equipment installation, or initial positional offsets. This allows subsequent motion performance evaluations to be based on objective measurements relative to the trainee's own baseline, rather than direct scoring against the group average, thereby improving the fairness and relevance of the evaluation. Specifically, the VR tracking device's motion data stream and baseline motion samples are time-aligned and interpolated by timestamps to handle different sampling rates or brief packet loss. Further geometric scale normalization and coordinate system registration are performed, mapping the trainee's current motion trajectory to the baseline reference system via rigid body / affine transformation, or scaling the trajectory by limb length using a normalization factor to make features such as displacement and angle comparable. For temporal differences, dynamic time warping (DTW) or relative temporal comparison can be used to align the motion phases. For force / force characteristics, if the force calibration of the sensor or estimation model is inconsistent, linear or nonlinear correction is performed using baseline mechanical measurements. After calibration, motion performance feature vectors are extracted from the calibrated trajectory, such as normalized position error, relative force deviation, timing offset, motion smoothness index, and completion percentage. For example, if the baseline motion sample shows that a trainee naturally leans forward 5° in a standard holding motion, then when evaluating the same motion, this 5° is first used as an individual baseline offset correction term to calculate the true deviation value relative to their natural posture to determine the motion quality.

[0080] S302. Based on the quantifiable action features defined in the operation action library, compare and calculate the action performance features with the corresponding quantifiable action features to obtain key performance indicators; key performance indicators include, but are not limited to, action positioning accuracy, force and force deviation, action timing and continuity, key response delay and process integrity.

[0081] In a schematic way, the calibrated motion performance characteristics are quantified into a set of comparable and traceable Key Performance Indicators (KPIs) to measure the accuracy, timeliness, and process integrity of technical actions. The operation motion library predefines quantifiable motion characteristics for each standard motion. During evaluation, the calibrated performance characteristics of each motion are compared item by item with the corresponding quantifiable motion characteristics. This can be done by calculating the root mean square error (RMSE), maximum deviation, and mean absolute error of position or trajectory; force / force error is expressed as absolute difference or percentage difference; timing indicators are expressed as response delay, duration deviation, or phase error; and continuity / completeness is expressed as the missed step rate or the number of missing sub-motions. Furthermore, to ensure that KPIs reflect both individual details and facilitate comprehensive evaluation, standardization and weighted synthesis are employed. Each comparison result is mapped to a score of 0-100 according to the tolerance range of the motion characteristics. Then, weighted averages are calculated based on the importance of the motion or the teaching objective to obtain motion-level KPIs. Finally, these are summarized into session-level KPIs by task or scenario. KPIs can take three forms simultaneously: scalar scores, vectorized sub-scores, and trend curves, so as to provide both an overall score and identify weaknesses.

[0082] S303. Based on baseline physiological data, calculate physiological variability according to real-time physiological data, and assess physiological variability to obtain physiological stress state.

[0083] The real-time physiological signals of trainees during training are transformed into quantitative indicators reflecting psychophysiological load or stress response. Baseline physiological data serves as a reference range for individuals at rest or low intensity. The real-time physiological data stream undergoes preprocessing steps such as artifact removal, missing data imputation, and synchronization. Physiological variability indicators, such as the short-term heart rate variability index (RMSSD), instantaneous heart rate deviation, peak and frequency of skin conductance response, and respiratory rate fluctuation amplitude, are calculated within a short time window. Physiological variability is further standardized to convert it into comparable quantities, and then the standardized values ​​are mapped to continuous or graded physiological stress state scores based on empirical thresholds or a trained lightweight classifier. For example, different physiological indicators can be assigned different weights, and the final physiological stress score can be generated through a fusion strategy. If a trainee completes the mask-fitting action under high noise conditions, and their heart rate increases by 12% from the baseline, RMSSD decreases significantly, and is accompanied by multiple short-term peaks in skin conductance, then this period is assessed as a moderate to high physiological stress state.

[0084] S304. Report the training results based on key performance indicators and physiological stress state.

[0085] Indicatively, various KPIs and physiological stress scores are combined to construct a comprehensive scoring system according to pre-set or configurable weighting rules. Optionally, an overall pass / fail judgment can be provided. The report includes an overall score, a list of sub-KPIs and their corresponding scores and tolerance ranges, a physiological stress timeline, a key event replay index, and corrective actions and training suggestions based on rule engines or strategy modules.

[0086] In one embodiment, reconstructing the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state further includes:

[0087] S21. Based on a sliding time window with a preset step size, calculate short-term key performance indicators according to the real-time VR tracking device motion data stream, and update the physiological and pathological evolution rate of the personalized newborn digital twin model according to the short-term key performance indicators; the physiological and pathological evolution rate is used to characterize the rate at which the physiological response state quantity changes over time.

[0088] Short-term key performance indicators (SKIs) refer to the immediate quantitative assessment of a trainee's movement quality within a short time window, reflecting the effectiveness and stability of operations over the most recent few seconds to tens of seconds. The purpose of using a sliding time window is to smooth out transient noise while ensuring real-time performance, promptly reflecting sudden changes in trainee performance without being misjudged by brief anomalies. Specifically, with a preset step size, the received VR tracking device's motion data stream is segmented by a rolling window during runtime. Key performance characteristics, such as average position error, average pressure, movement completion rate, movement rate consistency, and missed steps rate, are extracted and summarized for each movement or movement segment within the window, and then the SKIs are calculated. The combination of sliding windows and step sizes can generate a continuous sequence of short-term KPIs. Based on short-term KPIs, the physiological and pathological evolution rate of the personalized neonatal digital twin model is adjusted. This rate refers to the scaling factor of parameters representing the speed of physiological changes over time, such as the rate of decrease in blood oxygen saturation, the rate of heart rate decline, or the rate of change in body temperature. Dynamically adjusting this rate allows the model to evolve faster or slower under the current trainee's performance, achieving real-time matching of neonatal performance after real-time operations. For example, an exponentially weighted moving average (EWMA) is used to smooth the short-term KPIs, and then the physiological and pathological evolution rate is adjusted proportionally based on the deviation of the KPI from the target threshold. Simultaneously, safety constraints are applied, namely maximum and minimum rate boundaries and physiological acceptability checks, to avoid unrealistic physiological evolution. For example, if the average mask fit KPI within the short-term window is 0.2, lower than the target of 0.8, the physiological and pathological evolution rate can be increased by 20%, causing the neonatal's blood oxygen saturation to decrease more rapidly, thus forcibly pushing the training scenario towards a more urgent response.

[0089] S22. Calculate the short-term physiological stress state corresponding to the short-term key performance indicators based on the real-time physiological data stream, and calculate the concurrent event trigger probability based on the short-term key performance indicators and the short-term physiological stress state; the trigger probability is obtained by combining the probability distribution of the concurrent event atom library that changes over time with the adjustment factor based on the current training performance.

[0090] Short-term physiological stress is obtained by rapidly extracting and standardizing features from real-time physiological data within a short window. Simple motor performance cannot fully reflect the psychological load of trainees under high-pressure situations. The probability of concurrent event triggering should consider both the trainee's operational ability and psychological resilience to achieve an adaptive situational injection that is challenging yet not overwhelming. Specifically, short-window statistics are performed on real-time physiological data and normalized with baseline physiological data to calculate a physiological stress score. Short-term KPIs are then normalized into short-term performance factors. An adjustment factor f(KPI, Stress) for concurrent event triggering is calculated according to a predefined combination model, such as multiplication or weighted summation. For example, f = 1 + w1·(KPI) ref -KPI now )+w2·Stress score Where w1 and w2 are configurable weights, and KPIs are... ref The target KPI for this training node, KPI now Stress is a short-term key performance indicator. score This represents a short-term physiological stress state. The concurrent event atomic library provides the time-related baseline probability distribution P for each concurrent event. base (t), combining the baseline probability with the adjustment factor to obtain the actual trigger probability P. trigger For example, P trigger =clamp(P base (t)×f(KPI,Stress),0,1), where clamp represents limiting the result to [0,1]. To prevent frequent or extreme injections, injection cooldown time, maximum concurrent event limit, and safety rules should be set. For example, if the physiological and pathological evolution rate exceeds a certain threshold, new high-risk event injections should be suspended.

[0091] S23. When the concurrent event trigger probability meets the configured trigger mode, extract the parameter configuration corresponding to the concurrent emergency event from the concurrent event atomic library, and inject the parameter configuration into the VR scene and the personalized newborn digital twin model.

[0092] Each atomic event in the concurrent event atomic library contains not only the event type but also a parameter configuration template for injection, such as the start time offset, severity level, duration, accompanying noise and device failure parameters, and the amount of perturbation the event causes to the model's initial state. When determining P... triggerOnce the injection conditions are met, the trigger mode is satisfied, and a sampling / selection strategy is executed. For example, weighted sampling can be used, prioritizing events that are most challenging to the current teaching objective or best expose weaknesses, or events that can be completed based on the remaining time window of the scene. After extraction, the parameter configuration of the selected event is sent to the configuration file that updates the VR scene in real time, and the perturbation parameters corresponding to the event are injected into the personalized neonatal digital twin model, causing the model to begin evolving its physiological state according to the new conditions at the next time step. For example, when choking on milk is extracted as a moderately severe event, the injected parameters might include a sudden increase in airway resistance by 20%, a decrease in the decibel level of the accompanying crying sound, and a short-term blood oxygen saturation decrease driver in the digital twin model, forcing the trainee to quickly take actions such as suctioning and repositioning to correct the situation.

[0093] S24. In response to a short-term physiological stress state exceeding a preset stress threshold, the sudden emergency event in the VR scene is replaced with a calming training scene; the calming training scene includes breathing regulation and situational decompression; situational decompression includes generating action guidance prompts.

[0094] When the short-term physiological stress level exceeds a pre-set stress threshold, the system prioritizes the trainee's psychological and physiological tolerance, automatically downgrading the current high-pressure operational scenario to a calming training scenario to avoid excessive pressure or triggering an overly strong stress response. The calming training scenario is designed to help trainees recover to an acceptable stress level through short-term physiological regulation and cognitive decompression, thus enabling them to continue effective learning. Illustratively, alternative steps include pausing or reverting to the further evolution of concurrent events, switching the VR scenario to a low-stimulation version, and triggering breathing regulation guidance on the interface, including a visual breathing rhythm guide bar, voice / text guidance on breathing rhythm, and situational decompression exercises such as guiding phrases, advanced motion guidance, or relaxation exercises. Situational decompression can also include motion guidance prompts, i.e., generating step-by-step visual guidance, such as highlighting arrows indicating the correct contact position of the mask, and dynamic prompt bars indicating the pressure range, presented in an overlay format in the VR view, thereby reducing the trainee's physiological stress while providing clear and actionable corrective suggestions. Optionally, the duration and content of the calming training scenario can be dynamically adjusted based on the trainee's stress recovery. Once the trainee has recovered to a safe level, the course strategy can be used to decide whether to return to the original concurrent event to continue training, reduce the severity of the concurrent event to continue, or proceed to the next training node.

[0095] In one embodiment, the method further includes:

[0096] S31. Based on key performance indicators, visualize and annotate the motion data stream of the VR tracking device with the motion video corresponding to the physiological response state of the newborn to obtain training feedback results; the visualized annotation feedback includes annotating and quantifying the gap based on key performance indicators.

[0097] Training feedback results refer to feedback products presented to trainees and instructors in a combined visual and quantitative format. The aim is to make technical performance deficiencies and physiological consequences readily apparent, thereby supporting rapid correction and learning. Illustratively, this involves summarizing the motion data stream, calibrated motion performance characteristics, calculated key performance indicators (KPIs), and the time series of neonatal physiological responses from this training session. These time-series data are then precisely aligned with the motion video along a unified timeline. The generation of visual annotation feedback includes multiple levels of presentation. For example, keyframes or short windows related to KPIs are annotated on the video timeline with time markers and graphical overlays. Dynamic charts directly indicate motion deviations, insufficient / excessive force, and changes in neonatal physiological quantities at corresponding time points on the video frames. Optionally, summary views, such as motion score bar charts, KPI trend curves, and physiological stress heatmaps, are generated to allow trainees to quickly locate the frequency and extent of problems. Specifically, the visual annotations are synthesized by the rendering engine during post-processing and output in an interactive format with a timeline, while simultaneously generating structured metadata for retrieval and archiving. For example, after a suctioning training session, the trainee can see in the playback interface that at t=12.4 seconds, the suctioning depth exceeds the target by 6 mm in red. At the same time, the corresponding blood oxygen saturation decreases by 4% between t=12 and 20 seconds is displayed below, and suggested actions are listed on the right, such as reducing the insertion depth by 5 mm and maintaining the duration ≤2 seconds as immediate feedback.

[0098] S32. Identify weaknesses based on the training result report, and adjust the next node of the training path and the corresponding expected key performance indicators based on the weaknesses.

[0099] This example illustrates how, using training result reports as input, the system automatically identifies trainees' weaknesses by analyzing the sub-KPIs, physiological stress timelines, and metadata from visualized annotations listed in the reports. These weaknesses are those actions or abilities that repeatedly fail to meet standards or deviate significantly from the target KPIs in multiple practice sessions or key scenarios. The identification process prioritizes rule-based screening; for example, a weakness is identified when a sub-KPI falls below the target threshold in more than three training sessions without improvement in variability. Clustering or simple statistical learning methods can be combined, such as clustering by item scores to identify clustered low-scoring action categories, thus classifying and prioritizing weaknesses. After identifying weaknesses, a correction strategy is automatically generated based on the training path configuration and teaching objectives. For instance, the next node in the training path is adjusted—that is, a more targeted training node is selected or generated, and a clear expected key performance indicator is set for that next node.

[0100] In one embodiment, the method further includes:

[0101] S41. In response to the end of the training path, reacquire the relevant data of the trainee, and conduct training planning based on the training path configuration, training objectives and baseline knowledge test results to obtain a new training path.

[0102] Indicatively, the training path can end either by completing all preset teaching nodes or by being triggered by a session-level assessment, such as a training results report indicating whether the predetermined goals have been achieved or not, or by a human tutor determining that the training has entered the review phase. Once the assessment is complete, an automated reacquisition and reassessment process begins to ensure that the next stage of training is based on the latest and most representative trainee status. Specifically, the complete dataset of this training session is retrieved from the trainee's file, including VR tracking device motion data streams, real-time physiological data streams, KPIs and physiological stress curves from the training results report, motion videos, and telemetry overlay view indexes. If necessary, trainees are required to perform short-term recalibration tests or supplementary tests, such as repeating key movements several times or re-acquiring resting heart rate and skin conductance baselines, to obtain the latest observations that can be used to update baseline physiological data and baseline motion samples.

[0103] After obtaining the latest trainee-related data, a new training path is generated based on the training path configuration, training objectives, and baseline knowledge test results, using a hybrid strategy of rule engine and recommendation algorithm. Specifically, the rule engine selects candidate nodes based on the course outline and safety constraints, and then the recommendation algorithm sorts and parameterizes the candidate nodes, prioritizing the recommendation of training modules that can maximize short-term learning gains or specifically repair the identified weaknesses.

[0104] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0105] Based on the same inventive concept, this application also provides a virtual reality-based newborn care skills training system for implementing the aforementioned virtual reality-based newborn care skills training method. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the virtual reality-based newborn care skills training system provided below can be found in the limitations of the virtual reality-based newborn care skills training method described above, and will not be repeated here.

[0106] In one exemplary embodiment, such as Figure 4 As shown, a newborn care skills training system based on virtual reality technology is provided, including:

[0107] The newborn personalization module 401 is used to acquire newborn personalized indicators and, based on the newborn physiological digital twin basic model set, to perform parameter identification and generation on the newborn personalized indicators to obtain a personalized newborn digital twin model; the personalized newborn digital twin model includes a set of variable parameters that characterize the newborn's physiological characteristics; the set of variable parameters includes, but is not limited to, infant weight, lung compliance, heart rate and blood oxygen saturation waveform baseline, and response action sensitivity threshold;

[0108] Virtual reality module 402 is used to generate VR scenes based on a scene library and the current node of the training path, combined with a personalized newborn digital twin model; the VR scene includes sudden emergency events within a preset scene time window and their corresponding noise levels, available devices and interference;

[0109] The response evolution module 403 is used to acquire the VR tracking device action data stream and real-time physiological data stream of the trainee in response to the VR scene, and calculate the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device action data stream as the operation input.

[0110] Real-time update module 404 is used to reconstruct the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state;

[0111] The training evaluation module 405 is used to calculate evaluation indicators based on the VR tracking device motion data stream and real-time physiological data stream in response to the end of the scene time window, and to obtain a training result report; the evaluation indicators include key performance indicators and physiological stress state.

[0112] In one embodiment, an initialization module is also included, for:

[0113] Acquire the resource library; the resource library includes a scene library, an operation action library, a concurrent event atomic library, a basic set of neonatal physiological digital twin models, and training path configurations; the scene library includes metadata of each sudden emergency event and its corresponding scene element; the operation action library includes first aid actions and their corresponding quantifiable action features; quantifiable action features include timing, force, angle, and contact position; the concurrent event atomic library includes concurrent emergency events and their configuration trigger modes and probability distributions; the basic set of neonatal physiological digital twin models includes multi-scale models and their parameterized templates;

[0114] Acquire relevant data about the trainees and perform baseline calibration based on the data to obtain baseline physiological data, baseline movement samples, and training paths. The relevant data about the trainees includes trainee identity, training objectives, baseline knowledge test results, basic physiological data, and initial position data of the VR tracking device.

[0115] In one embodiment, the response evolution module 403 is further configured to:

[0116] The motion data stream of the VR tracking device is analyzed to extract operational features, including timing, contact position, force intensity, and motion angle.

[0117] Operational features are input into a personalized newborn digital twin model to obtain instantaneous physiological parameters corresponding to each time step; the personalized newborn digital twin model performs physiological and pathological evolution calculations on the operational feature input based on a set of variable parameters;

[0118] The various instantaneous physiological parameters are aggregated into the newborn's physiological response state; the newborn's physiological response state includes, but is not limited to, the result of the action response, heart rate, blood oxygen saturation, respiratory rate and body temperature.

[0119] In one embodiment, the training evaluation module 405 is further configured to:

[0120] The motion data stream of the VR tracking device is individually calibrated based on baseline motion samples to obtain calibrated motion performance characteristics;

[0121] Based on the quantifiable action features defined in the action library, the action performance features are compared with the corresponding quantifiable action features to obtain key performance indicators; key performance indicators include, but are not limited to, action positioning accuracy, force and force deviation, action timing and continuity, key response delay and process integrity.

[0122] Based on baseline physiological data, physiological variability is calculated using real-time physiological data, and physiological stress status is obtained by assessing physiological variability.

[0123] Training results are reported based on key performance indicators and physiological stress states.

[0124] In one embodiment, the real-time update module 404 is further configured to:

[0125] Based on a sliding time window with a preset step size, short-term key performance indicators are calculated according to the real-time VR tracking device motion data stream, and the physiological and pathological evolution rate of the personalized newborn digital twin model is updated according to the short-term key performance indicators; the physiological and pathological evolution rate is used to characterize the rate at which physiological response state quantities change over time.

[0126] The short-term physiological stress state corresponding to the short-term key performance indicators is calculated based on the real-time physiological data stream, and the concurrent event trigger probability is calculated based on the short-term key performance indicators and the short-term physiological stress state; the trigger probability is obtained by combining the probability distribution of the concurrent event atom library that changes over time with the adjustment factor based on the current training performance;

[0127] When the concurrent event trigger probability meets the configured trigger mode, the parameter configuration corresponding to the concurrent emergency event is extracted from the concurrent event atomic library and injected into the VR scene and the personalized newborn digital twin model;

[0128] In response to short-term physiological stress exceeding a preset stress threshold, the sudden emergency event in the VR scene is replaced with a calming training scene; the calming training scene includes breathing regulation and situational decompression; situational decompression includes generating action guidance prompts.

[0129] In one embodiment, a training result feedback module is also included, for:

[0130] Based on key performance indicators, the VR tracking device's motion data stream and the corresponding motion videos of the newborn's physiological response state are visualized and annotated to obtain training feedback results; the visualized and annotated feedback includes annotating and quantifying the gap based on key performance indicators.

[0131] Weaknesses are identified based on the training results report, and the next node of the training path and the corresponding expected key performance indicators are adjusted accordingly.

[0132] In one embodiment, a phased training module is also included for:

[0133] In response to the end of the training path, the relevant data of the trainees are reacquired, and a new training path is obtained based on the training path configuration, training objectives and baseline knowledge test results.

[0134] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps in the above method embodiments.

[0135] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0136] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0137] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for training newborn care skills based on virtual reality technology, characterized in that, The method includes: Personalized indicators for newborns are obtained, and based on the basic set of newborn physiological digital twin models, parameters are identified and generated for the personalized newborn indicators to obtain a personalized newborn digital twin model. The personalized newborn digital twin model includes a set of variable parameters that characterize the physiological characteristics of newborns. The set of variable parameters includes, but is not limited to, infant weight, lung compliance, baseline waveforms of heart rate and blood oxygen saturation, and sensitivity threshold for response actions. Based on the scene library, a VR scene is generated according to the current node of the training path and the personalized newborn digital twin model; the VR scene includes sudden emergency events within a preset scene time window and their corresponding noise levels, available devices and interference; The system acquires VR tracking device motion data streams and real-time physiological data streams of trainees responding to the VR scene, and calculates the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device motion data streams as operation inputs. The audiovisual feedback corresponding to the VR scene is reconstructed based on the newborn's physiological response state. In response to the end marker of the scene time window, evaluation indicators are calculated based on the motion data stream of the VR tracking device and the real-time physiological data stream to obtain a training result report; the evaluation indicators include key performance indicators and physiological stress state.

2. The method according to claim 1, characterized in that, Before obtaining personalized newborn indicators and generating personalized newborn digital twin models by parameter identification based on the newborn physiological digital twin basic model set, the process further includes: Acquire a resource library; the resource library includes a scene library, an operation action library, a concurrent event atomic library, a basic set of neonatal physiological digital twin models, and training path configurations; the scene library includes metadata of each sudden emergency event and its corresponding scene element; the operation action library includes first aid actions and their corresponding quantifiable action features; the quantifiable action features include timing, force, angle, and contact position; the concurrent event atomic library includes concurrent emergency events and their configured trigger modes and probability distributions; the basic set of neonatal physiological digital twin models includes multi-scale models and their parameterized templates; Acquire relevant data about the trainees and perform baseline calibration based on the relevant data to obtain baseline physiological data, baseline movement samples, and training paths; the relevant data about the trainees includes trainee identity, training objectives, baseline knowledge test results, basic physiological data, and initial position data of VR tracking device movements.

3. The method according to claim 2, characterized in that, The process of calculating the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device's motion data stream as operational input, includes: The motion data stream of the VR tracking device is parsed to extract operational features; the operational features include timing, contact position, force intensity, and motion angle. The operational features are input into the personalized newborn digital twin model to obtain the instantaneous physiological parameters corresponding to each time step; the personalized newborn digital twin model performs physiological and pathological evolution calculations on the operational feature input based on the variable parameter set; The instantaneous physiological parameters are aggregated into the neonatal physiological response state; the neonatal physiological response state includes, but is not limited to, action response results, heart rate, blood oxygen saturation, respiratory rate and body temperature.

4. The method according to claim 2, characterized in that, In response to the end marker of the scene time window, the evaluation index is calculated based on the VR tracking device motion data stream and the real-time physiological data stream to obtain a training result report, including: The VR tracking device motion data stream is individually calibrated based on the baseline motion samples to obtain calibrated motion performance characteristics; Based on the quantifiable action features defined in the operation action library, the action performance features are compared with the corresponding quantifiable action features to obtain key performance indicators; the key performance indicators include, but are not limited to, action positioning accuracy, force and force deviation, action timing and continuity, key response delay and process integrity. Based on the baseline physiological data, physiological variability is calculated according to the real-time physiological data, and the physiological variability is evaluated to obtain the physiological stress state; The training result report is obtained based on the key performance indicators and the physiological stress state.

5. The method according to claim 4, characterized in that, The reconstructing of the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state further includes: Based on a sliding time window with a preset step size, short-term key performance indicators are calculated according to the real-time motion data stream of the VR tracking device, and the physiological and pathological evolution rate of the personalized newborn digital twin model is updated according to the short-term key performance indicators; the physiological and pathological evolution rate is used to characterize the rate at which the physiological response state quantity changes over time. The short-term physiological stress state corresponding to the short-term key performance indicator is calculated based on the real-time physiological data stream, and the concurrent event trigger probability is calculated based on the short-term key performance indicator and the short-term physiological stress state; the trigger probability is obtained by combining the probability distribution of the concurrent event atom library that changes over time with an adjustment factor based on the current training performance; When the concurrent event trigger probability meets the configured trigger mode, the parameter configuration corresponding to the concurrent emergency event is extracted from the concurrent event atomic library, and the parameter configuration is injected into the VR scene and the personalized newborn digital twin model; In response to the short-term physiological stress state exceeding a preset stress threshold, the sudden emergency event in the VR scene is replaced with a calming training scene; the calming training scene includes breathing regulation and situational decompression; the situational decompression includes generating action guidance prompts.

6. The method according to claim 1, characterized in that, The method further includes: Based on the key performance indicators, the motion data stream of the VR tracking device and the motion video corresponding to the physiological response state of the newborn are visualized and annotated to obtain training feedback results; the visualized and annotated feedback includes annotating and quantifying the gap based on the key performance indicators. Weaknesses are identified based on the training result report, and the next node of the training path and the corresponding expected key performance indicators are corrected based on the weaknesses.

7. The method according to claim 2, characterized in that, The method further includes: In response to the end of the training path, the relevant data of the trainee are reacquired, and a new training path is obtained based on the training path configuration, the training objective and the baseline knowledge test results.

8. A newborn care skills training system based on virtual reality technology, characterized in that, The system includes: The newborn personalization module is used to acquire newborn personalized indicators and, based on the newborn physiological digital twin basic model set, to perform parameter identification and generation on the newborn personalized indicators to obtain a personalized newborn digital twin model; the personalized newborn digital twin model includes a set of variable parameters characterizing the newborn's physiological characteristics; the set of variable parameters includes, but is not limited to, infant weight, lung compliance, heart rate and blood oxygen saturation waveform baseline, and response action sensitivity threshold; The virtual reality module is used to generate VR scenes based on a scene library and the current node of the training path, combined with the personalized newborn digital twin model; the VR scene includes sudden emergency events within a preset scene time window and their corresponding noise levels, available devices, and interference; The response evolution module is used to acquire the VR tracking device action data stream and real-time physiological data stream of the trainee in response to the VR scene, and to calculate the newborn's physiological response state in real time based on the personalized newborn digital twin model, using the VR tracking device action data stream as operation input. The real-time update module is used to reconstruct the audiovisual feedback corresponding to the VR scene based on the newborn's physiological response state; The training evaluation module is used to calculate evaluation indicators based on the VR tracking device motion data stream and the real-time physiological data stream in response to the end marker of the scene time window, and to obtain a training result report; the evaluation indicators include key performance indicators and physiological stress state.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images