AI-based nursing teaching scene construction and evaluation system

Abstract

The invention relates to the technical field of intelligent education, in particular to an AI-based nursing teaching scene construction and evaluation system. According to the system, a dynamic case script is generated through historical case data, and a virtual patient physiological model is constructed; creating three teaching scenes of community follow-up visit, emergency rescue and traditional Chinese medicine nursing fusion by using a mixed reality technology; a personalized learning portrait of the student is constructed through a pre-test, and accurate matching between the ability of the student and the teaching content is formed; a double-track system skill training module is built, and AI tutor guidance and teacher guidance are integrated; and an ability evaluation system based on operation data is established, and intelligent pushing of adaptive training content is realized. The problems that in traditional nursing teaching, scene authenticity is insufficient, personalized training is insufficient, and clinical thinking is difficult to cultivate are solved, and the clinical operation skill and decision judgment ability of students are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent education technology, and in particular to an AI-based nursing teaching scenario construction and evaluation system. Background Technology

[0002] Nursing education, as an important component of the medical education system, directly impacts the clinical practice abilities and professional ethics of future nursing professionals. Traditional nursing teaching models primarily rely on a combination of classroom lectures, skills training, and clinical internships. While this has resulted in a relatively mature teaching system, it has gradually revealed several pressing issues when facing complex and ever-changing clinical environments and increasing medical demands.

[0003] Existing simulation-based teaching methods often employ pre-set cases and standardized patients, resulting in a simplistic case progression path that fails to dynamically reflect the complexity and uncertainty of real patient conditions. While students trained in this static environment may master basic operational skills, they often demonstrate insufficient ability to cope with common clinical emergencies, complication developments, and the interplay of multiple factors. Furthermore, traditional teaching methods do not adequately reflect the individual characteristics of patients; factors such as cultural background and emotional state often overlook the influence of nursing plans, leading to a need to improve students' humanistic care and personalized nursing abilities in real clinical settings.

[0004] At the same time, due to limited teaching resources, teachers find it difficult to provide customized training content for each student. Students have different levels of basic knowledge, proficiency in operational skills, and characteristics of clinical reasoning, but under the traditional teaching model, they often receive the same teaching content and training intensity, making it difficult to achieve individualized instruction.

[0005] Furthermore, in terms of practical teaching, insufficient clinical internship opportunities have become a common problem. Due to patient safety concerns and strained medical resources, students have limited opportunities for hands-on practice, especially in high-risk scenarios such as emergency resuscitation and critical care nursing, where it is even more difficult to obtain sufficient practical training. Although virtual simulation technology alleviates this contradiction to some extent, existing virtual training systems often lack realism and interactivity, making it difficult to simulate the complex factors and stressful situations in clinical environments.

[0006] This invention aims to solve the technical problems existing in related technologies. To this end, it proposes an AI-based nursing teaching scenario construction and evaluation system. Summary of the Invention

[0007] The purpose of this invention is to provide an AI-based nursing teaching scenario construction and evaluation system to solve the problems existing in the prior art.

[0008] To achieve the above objectives, this invention provides an AI-based nursing teaching scenario construction and evaluation system, comprising a simulated case module, an intelligent scenario module, a dual-track skills training module, and an adaptive module. Specifically: the simulated case module generates dynamic case scripts based on historical case data, constructs patient physiological models, and predicts the probability of complication occurrence to form simulated cases for skills training; the intelligent scenario module constructs intelligent scenarios based on the simulated cases, mixing real-world scenarios with student profiles; the dual-track skills training module, relying on the intelligent scenario, provides students with a virtual drug compatibility system, dietary plan rationality assessment, and injection angle detection functions; and the adaptive module assesses the student's operational data in the dual-track skills training module and adaptively pushes personalized training content packages to the student based on the assessment results.

[0009] Furthermore, the simulated cases integrate case data from tertiary hospitals and are generated by AIGC into dynamic case scripts containing multiple complication evolution paths; the patient physiological model can simulate the multi-factor interaction of diet, exercise, and medication; and the complication probability prediction is based on the standards and data of evidence-based medicine guidelines.

[0010] Furthermore, the mixed reality scenario includes at least one of the following: a community follow-up scenario integrating natural language interactive virtual patients and environmental noise simulation; an emergency scenario integrating biosensor clothing pressure feedback and multi-parameter monitor linkage; and a traditional Chinese medicine nursing integration scenario integrating tongue image recognition AI and moxibustion heat simulation system.

[0011] Furthermore, the construction of the trainee profile in the intelligent scenario building module is completed through pre-testing. The trainee profile includes a knowledge reserve radar chart, clinical thinking characteristic analysis results, and skill operation baseline data. The simulated case generation module can automatically generate personalized features for simulated cases. The personalized features include the simulated patient's cultural background, complication risk level, and emotional state.

[0012] Furthermore, the dual-track skills training module also includes a teacher guidance module, which is used to guide trainees in clinical thinking, conduct integrated teaching of traditional Chinese medicine and nursing, and simulate virtual family members' sudden emotional conflicts and organize trainees to conduct coping training; the personalized training content package pushed by the ability assessment and content push module includes micro-lecture videos, VR intensive training scenarios and medication safety challenge games.

[0013] Furthermore, the complication probability prediction is achieved through a dynamic risk assessment model, which uses an exponential decay algorithm to accumulate historical nursing data to reflect the timeliness of nursing risks; at the same time, an event association weighting mechanism is introduced to enhance the model's response speed and accuracy to acute medical events.

[0014] Furthermore, the dynamic risk assessment model enhances its sensitivity to instantaneous high-risk fluctuations by constructing a mean neighborhood cumulative risk value, and uses a nonlinear mapping with a fused time interval sensitive kernel to process non-equal interval nursing data.

[0015] Furthermore, it also includes a virtual patient adaptive matching module, which is used to dynamically adjust the difficulty of the virtual patient according to the trainee's operation level; the virtual patient adaptive matching module constructs a virtual patient profile and a trainee's ability profile, and determines the appropriate virtual patient difficulty level based on the matching degree between the two.

[0016] Furthermore, the virtual patient adaptive matching module dynamically adjusts the nursing difficulty parameters of the virtual patient based on the trainee's real-time operation score in skills training, including: increasing the nursing difficulty when the operation score is consistently higher than the threshold, decreasing the nursing difficulty when the operation score is consistently lower than the threshold, and updating the node radius and nursing demand complexity in the virtual patient profile accordingly.

[0017] Furthermore, in the virtual patient adaptive matching module, the specific process of determining the adaptation level based on the matching degree between the trainee's ability profile and the virtual patient profile is as follows: extract the feature parameters of each node in the virtual patient profile and match them with the node graph in the trainee's ability profile; count the matching frequency of the two and calculate the adaptation degree based on the matching frequency; and assign virtual cases of appropriate difficulty level to the trainee based on the adaptation degree value.

[0018] Compared with existing technologies, this invention provides an AI-based nursing teaching scenario construction and evaluation system, which has the following beneficial effects: 1. Based on historical case data, dynamic case scripts are generated and patient physiological models are constructed. Combined with mixed reality technology, multimodal clinical scenarios are created. At the same time, through personalized learning profiles and virtual patient feature generation, students of different levels can obtain training content that is precisely matched to their abilities in a highly realistic dynamic environment, which effectively improves clinical adaptability and learning efficiency.

[0019] 2. Integrating AI tutor guidance and teacher guidance functions not only ensures standardized training of basic operations but also strengthens the cultivation of clinical thinking; at the same time, the ability assessment based on operation data and the adaptive training package push mechanism realize real-time monitoring of teaching effectiveness and dynamic adjustment of training content, significantly improving students' skill proficiency and problem-solving ability.

[0020] 3. A dynamic risk assessment model is adopted, which accurately predicts the probability of complications through exponential decay accumulation and event correlation weighting techniques. Combined with the dynamic difficulty adjustment mechanism established by the virtual patient adaptive matching module, students can cultivate forward-looking nursing thinking and risk management capabilities through progressive challenges, effectively improving the quality of clinical decision-making.

[0021] 4. Through multimodal data fusion and real-time feedback, it supports the synchronous analysis of multi-dimensional data such as operational behavior and cognitive decision-making; it establishes a three-party collaborative mechanism among teachers, AI, and students to balance technology empowerment and humanistic education; it integrates intelligent assessment functions of traditional Chinese and Western medicine nursing to build a distinctive nursing education system and provide comprehensive support for cultivating compound nursing talents. Attached Figure Description

[0022] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 A schematic diagram of the overall structure of an AI-based nursing teaching scenario construction and evaluation system provided by the present invention; Figure 2 This invention provides a flowchart of a simulated case generation system for constructing and evaluating nursing teaching scenarios based on AI. Detailed Implementation

[0023] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0024] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0025] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0026] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0027] Please see Figure 1 As shown, this embodiment provides an AI-based nursing teaching scenario construction and evaluation system, including a simulated case module, an intelligent scenario module, a dual-track skills training module, and an adaptive module. The simulated case module generates dynamic case scripts based on historical case data, constructs patient physiological models, and predicts the probability of complications to form simulated cases for skills training. The intelligent scenario module builds intelligent scenarios based on the simulated cases, mixing real-world scenarios with student profiles. The dual-track skills training module, relying on the intelligent scenario, provides students with a virtual drug compatibility system, dietary plan rationality assessment, and injection angle detection functions. The adaptive module assesses the student's operational data in the dual-track skills training module and adaptively pushes personalized training content packages to the student based on the assessment results.

[0028] Specifically, taking diabetes complication nursing training as an example, the simulated case module generates simulated cases that meet teaching needs by relying on historical case data and AI technology. First, historical case data is collected and preprocessed. Nearly 12,000 clinically valid diabetes cases from 30 top-tier hospitals in China over the past 10 years are integrated through hospital information systems and electronic medical record systems. These cases cover various types of diabetes and eight common complications. After desensitization and cleaning, a structured case database containing fields such as patient age and disease duration is formed. Next, the database is processed using the GPT-4 medical-specific fine-tuning model to generate dynamic case scripts containing the evolution paths of six types of complications. Taking a type 2 diabetes case with diabetic foot risk as an example, the script includes the initial scenario, disease evolution branches, corresponding physiological indicator changes, and key nursing observation points. Then, based on the dynamic case script, a virtual patient blood glucose fluctuation model is constructed using a multi-factor coupling algorithm. This model uses diet, exercise, and medication as core input parameters to simulate blood glucose change curves under different conditions. Finally, a complication probability prediction algorithm based on the "Guidelines for the Prevention and Treatment of Type 2 Diabetes in China" is embedded. A logistic regression model analyzes relevant factors of the simulated patient to output the probability of complication occurrence, ultimately forming a complete simulated case.

[0029] The intelligent scenario module centers on a generated simulated case of type 2 diabetes with diabetic foot risk, combining mixed reality scenarios with student profiles to build an immersive intelligent teaching environment. The mixed reality scenario construction includes a community follow-up scenario and a specialized diabetic foot care scenario, integrating a natural language interactive virtual patient unit and an environmental noise simulation unit, enabling real-time dialogue and community environmental noise simulation. A 1:1 virtual patient foot model is constructed and integrated with haptic feedback gloves. Student profile construction is completed through a 60-minute, 100-point pre-test, which includes three types of questions: knowledge reserve test, clinical reasoning test, and skills operation test. The test generates a knowledge reserve radar chart, analyzes clinical reasoning characteristics, and collects baseline data for skills operation. The system also automatically adjusts the mixed reality scenario parameters based on the simulated case and student profile to achieve precise matching between the scenario and the student's abilities.

[0030] The dual-track skills training module, based on a built-in intelligent scenario, provides trainees with a dual-track training mode of self-training and teacher guidance. Regarding the activation of core functional modules, after trainees enter the intelligent scenario, the system automatically activates the virtual drug compatibility system, the dietary plan rationality assessment function, and the injection angle detection function. The virtual drug compatibility system provides a library of commonly used drugs and judges the rationality of drug combinations; the dietary plan rationality assessment function calculates dietary nutrition and calories and generates an assessment report; and the injection angle detection function monitors the injection angle in real time and prompts adjustments. When trainees encounter clinical thinking blockages, operational errors, or inappropriate responses to emergency scenarios during self-training, the system automatically triggers the teacher guidance module. Nursing instructors intervene through a remote collaboration platform, guiding trainees to organize their thoughts, demonstrating standardized operations, explaining communication skills, and organizing repeated drills.

[0031] The adaptive module collects real-time operational data from trainees in the dual-track skills training module, completing competency assessments and personalized training content delivery. Operational data is collected via the system's backend module at a frequency of once per minute, recording core data items such as skill operation data, thought process and decision-making data, and learning behavior data. Nursing competency assessment uses a weighted scoring method, evaluating four dimensions: operational standardization, decision-making accuracy, learning efficiency, and emergency response capability. Scores for each dimension are calculated using corresponding formulas and summed to obtain a comprehensive score, thus determining the competency level. For example, a trainee scoring 72.3 points corresponds to an intermediate level. Based on the competency level and weaknesses in each dimension, the system adaptively delivers personalized training content packages containing micro-lecture videos, VR reinforcement scenarios, medication safety challenge games, and cloud-based collaborative tasks. These content packages have a 7-day validity period, during which trainees must complete the training. Instructors will provide timely feedback on trainees' questions and collaborative task solutions.

[0032] Therefore, the simulated case module integrates case data, preprocesses it, and uses an AI model to generate a dynamic script containing the evolution path of complications. Combined with a blood glucose fluctuation model and evidence-based medicine guidelines, it predicts the probability of complications, forming a simulated case. The intelligent scenario module uses this simulated case as the core and uses mixed reality devices to build specific scenarios for community follow-up and diabetic foot care. It constructs student profiles and adapts them to scenario parameters through pre-testing. The dual-track skills training module relies on the intelligent scenario to provide self-training functions such as virtual drug compatibility, dietary assessment, and injection angle detection. When students encounter problems such as mental blocks or operational errors, teachers can intervene and provide guidance through the remote module. The adaptive module collects student operation data in real time, evaluates the ability level from four dimensions such as operation standardization and decision accuracy, and then pushes personalized training packages containing micro-lecture videos, VR reinforcement scenarios, etc. according to the level.

[0033] Specifically, the simulated cases integrate case data from tertiary hospitals and are generated by AIGC into dynamic case scripts containing multiple complication evolution paths; the patient physiological model can simulate the multi-factor interaction of diet, exercise, and medication; complication probability prediction is based on evidence-based medicine guidelines and data. The mixed reality scenarios include at least one of the following: a community follow-up scenario integrating natural language interactive virtual patients and environmental noise simulation; an emergency scenario integrating biosensor clothing pressure feedback and multi-parameter monitoring; and a TCM nursing integration scenario integrating tongue image recognition AI and moxibustion heat simulation system. The construction of the trainee profile in the intelligent scenario building module is completed through pre-testing. The trainee profile includes a knowledge reserve radar chart, clinical thinking characteristic analysis results, and skill operation baseline data; and the simulated case generation module can automatically generate personalized features for the simulated cases, including the simulated patient's cultural background, complication risk level, and emotional state.

[0034] See Figure 2 It is known that the data source is clinical case data of type 2 diabetes from 15 tertiary-level hospitals in China, collected through the interfaces of hospital information systems and electronic medical record systems. A total of 18,000 valid cases were screened, covering different age groups, disease stages, and complication types, ensuring that the data covers common clinical scenarios. The data underwent preprocessing. Data analysis tools were used to process the raw data. The first step involved using a data anonymization algorithm to remove private information such as patient names, ID numbers, and hospital numbers, retaining only treatment-related fields such as case ID, gender, age, disease course, blood glucose monitoring data, medication records, and complication occurrence time. The second step involved using an outlier detection algorithm to remove invalid data, for example, marking and deleting abnormal records with fasting blood glucose values ​​<2.8 mmol / L or >22.2 mmol / L. Finally, a structured medical record database was formed and stored in the database, supporting subsequent data retrieval.

[0035] A medical AIGC model was selected and a training set of 18,000 preprocessed case data was used. A training corpus was constructed in conjunction with authoritative materials such as the "Guidelines for the Prevention and Treatment of Type 2 Diabetes in China" and the "Guidelines for Diabetes Nursing Practice". The model was fine-tuned through supervised learning and three types of training tasks were set up: case scenario generation, complication evolution path design, and dynamic description of physiological indicators. This enabled the model to generate dynamic scripts that conform to clinical logic. Taking the risk of type 2 diabetes complicated with diabetic foot as an example, the AIGC model is input to generate a type 2 diabetes case script containing the early-mid-late evolution path of diabetic foot. It needs to reflect the impact of diet, exercise, and medication on blood glucose, including dynamic changes in key physiological indicators. The dynamic script output by the model includes an initial scenario such as a 55-year-old patient with an 8-year history of type 2 diabetes, a fasting blood glucose of 9.2 mmol / L, and occasional numbness in the foot. The branch evolution path is as follows: Path 1: Uncontrolled diet and missed doses of hypoglycemic drugs lead to a rise in fasting blood glucose to 13.5 mmol / L after 72 hours, with worsening numbness in the foot and purplish discoloration of the skin; Path 2: Regular exercise and standardized medication lead to a drop in fasting blood glucose to 6.8 mmol / L after 72 hours, with relief of foot symptoms, supporting subsequent linkage with the patient's physiological model.

[0036] In one preferred embodiment, dietary, exercise, and medication parameters are converted into mathematical formulas; blood glucose monitoring data from real clinical cases are used as a validation set, and the blood glucose curves predicted by the model are compared with the real data. The accuracy of the model is evaluated by the root mean square error.

[0037] The specific implementation of the community follow-up scenario involves integrating iFlytek's medical speech recognition system with a natural language processing module to construct a virtual patient voice interaction logic. The virtual patient avatar is modeled using the Unity 3D engine, depicting the appearance of a 50-year-old diabetic patient in the community. The trainee asks, "How have your blood sugar monitoring results been over the past week?" After the system parses the semantics, it retrieves the corresponding data from the case script, and the virtual patient answers in natural language, such as, "I measure my fasting blood sugar every morning, and it's usually between 8.0 and 9.5. Sometimes it's even higher if I forget to take my medication." Multi-turn dialogue is also supported; for example, if the trainee asks, "How often do you forget to take your medication?" the virtual patient can respond, "About three times in the last two weeks, because I'm in a rush to take my grandson to school in the morning." Simultaneously, an environmental noise generator presets common community noise scenarios such as neighbors talking, vehicle noise, and square dance music. The noise intensity can be adjusted through the system backend. When the trainee communicates with the virtual patient, the system randomly plays low-intensity noise to simulate interference scenarios in a real community follow-up.

[0038] The specific implementation of the emergency scenario includes a biosensor garment, a multi-parameter monitor simulator, and a VR controller. The biosensor garment can collect physiological signals such as the trainee's heart rate and respiratory rate, with a sampling rate of up to 250Hz. The multi-parameter monitor simulator supports the display of simulated ECG, blood oxygen, and blood pressure parameters and can achieve data linkage with the system via a serial port. When the trainee performs operations such as chest compressions and intravenous punctures using the VR controller, the biosensor garment applies pressure feedback to the trainee's chest and arms through its built-in air pressure sensor. Simultaneously, the system adjusts the virtual patient's physiological parameters in real time based on the trainee's exertion, forming a closed-loop training cycle of operation-feedback-parameter change. When the simulated patient's condition changes, such as a sudden drop in blood sugar to 2.5 mmol / L leading to hypoglycemic shock, the system automatically sends a command to the monitor, causing it to display corresponding alarm information. The trainee must intervene according to the monitor's prompts. After the intervention is completed, the system sends a parameter update command to the monitor, showing that the blood sugar gradually returns to the normal range, recreating the equipment linkage scenario in emergency resuscitation.

[0039] The specific implementation of the integrated TCM nursing scenario includes a VR headset, a tongue image acquisition camera, and a moxibustion thermal simulator. The system integrates a tongue image recognition model based on a convolutional neural network. This model uses 20,000 tongue images from the "Atlas of Tongue Diagnosis in Traditional Chinese Medicine" as its training set and can identify features such as tongue color, shape, and coating. Trainees use the tongue image acquisition camera to capture virtual patient tongue images. The system inputs these images into the recognition model, outputs tongue diagnosis results, and pushes corresponding TCM nursing plans. The moxibustion thermal simulator has built-in heating elements and a temperature sensor. A moxibustion operation interface is built into the VR scene. After the trainee selects the moxibustion site and starts moxibustion, the system gradually adjusts the temperature of the thermal simulator according to the moxibustion duration. Simultaneously, visual effects such as moxibustion smoke and skin redness are displayed in the VR scene, simulating the warmth and operation experience of real moxibustion, thus training trainees' TCM nursing operation skills.

[0040] Immersive environments were constructed for three scenarios: community follow-up, emergency care, and integration of traditional Chinese medicine nursing. These environments integrate natural language interaction, physiological feedback, and AI recognition, which not only reproduce the real operation and environmental characteristics of different clinical scenarios, but also realize closed-loop training of virtual operation, real-time feedback, and parameter linkage, effectively improving trainees' scenario adaptation and practical skills.

[0041] Specifically, the complication probability prediction is achieved through a dynamic risk assessment model. This model uses an exponential decay algorithm to accumulate historical nursing data to reflect the timeliness of nursing risks. Simultaneously, an event association weighting mechanism is introduced to enhance the model's response speed and accuracy to acute medical events. The dynamic risk assessment model also enhances its sensitivity to instantaneous high-risk fluctuations by constructing a mean neighborhood to accumulate risk values, and employs a nonlinear mapping with a fused time interval sensitivity kernel to process non-equidistant nursing data.

[0042] Therefore, this embodiment uses the probability prediction of diabetic foot and diabetic nephropathy in patients with type 2 diabetes as a scenario to construct a dynamic risk assessment model. First, 12 input variables across three dimensions—physiological indicators, treatment behaviors, and basic information—are determined. These variables are automatically collected from relevant system modules and updated daily. An exponential decay algorithm calculates historical data weight coefficients using a formula. The decay coefficients are calibrated according to the type of complication, and then weighted and integrated with historical data from the past 90 days to obtain the final input values ​​of the variables, reflecting the dominant influence of recent nursing data on risk. An event association weighting mechanism first defines and identifies three types of acute medical events in real time, setting association weighting coefficients according to severity. After an event is identified, the coefficient is multiplied by the basic predicted probability to obtain the final risk probability. An alert is also pushed and a corresponding training task is triggered. The model has been validated with 18,000 simulated cases, achieving a complication prediction accuracy of over 89% and an average acute event response delay of 3.2 minutes.

[0043] To address the issue of easily masked instantaneous high risks, the mean neighborhood cumulative risk value construction method sets a 24- or 72-hour neighborhood time window based on the complication type and divides it into 1-hour sub-windows. The risk values ​​of the current and the adjacent sub-windows are weighted and accumulated using a formula. A threshold is set, and if the threshold is exceeded, an early warning is triggered and the trainee is forced to intervene first, thereby enhancing the model's sensitivity to instantaneous high risks.

[0044] Specifically, it also includes a virtual patient adaptive matching module, used to dynamically adjust the difficulty of virtual patients based on the trainee's operational level. This module constructs a virtual patient profile and a trainee's ability profile, and determines the appropriate virtual patient difficulty level based on the matching degree between the two. The module dynamically adjusts the nursing difficulty parameters of the virtual patient based on the trainee's real-time operational scores during skills training, including: increasing the nursing difficulty when the operational score is consistently above a threshold, and decreasing the nursing difficulty when the operational score is consistently below a threshold, and updating the node radius and nursing requirement complexity in the virtual patient profile accordingly. The specific process of determining the adaptation level based on the matching degree between the trainee's ability profile and the virtual patient profile in the virtual patient adaptive matching module is as follows: extracting the feature parameters of each node in the virtual patient profile and matching them with the node set in the trainee's ability profile; counting the matching frequency and calculating the adaptation degree based on the matching frequency; and assigning a virtual case of appropriate difficulty level to the trainee based on the adaptation degree value.

[0045] Therefore, the following steps are taken: First, operational data from trainees in the dual-track skills training module is collected through a data interface. This includes skill indicators such as the accuracy rate of virtual drug compatibility, the rate of achieving the required injection angle, and the pass rate of dietary plan evaluation, as well as cognitive indicators such as the time spent on case analysis and the accuracy rate of complication risk assessment. Based on this, a trainee competency profile is constructed, covering three dimensions: skill mastery, clinical thinking efficiency, and risk identification ability. Each dimension is quantified and scored from 0 to 100. Simultaneously, the module constructs a virtual patient profile, defining virtual patient attributes from three dimensions: nursing difficulty coefficient, disease complexity, and emergency scenario density. The nursing difficulty coefficient relates to operational requirements; disease complexity corresponds to the number and rate of evolution of complications; and emergency scenario density refers to the frequency of unexpected situations during training. Each attribute is also graded from 0 to 100. Then, a cosine similarity algorithm is used to calculate the matching degree between the trainee competency profile and the virtual patient profile. Based on the matching degree results, the appropriate difficulty level is determined: a matching degree ≥85% is high difficulty, 60%-84% is medium difficulty, and <60% is low difficulty. Finally, trainees are assigned virtual patients of the corresponding difficulty level, achieving precise "competency-difficulty" matching.

[0046] In another preferred embodiment, if a trainee's real-time operation score is higher than the threshold corresponding to the current difficulty level for three consecutive times, the module automatically triggers a difficulty-increasing mechanism: adjusting the virtual patient profile parameters, including increasing the nursing difficulty coefficient by 10-15 points, the disease complexity by 10 points, and the emergency scenario density by 5-8 points; if the trainee's real-time operation score is lower than the threshold corresponding to the current difficulty level for three consecutive times, a difficulty-reducing mechanism is triggered: reducing the nursing difficulty coefficient and disease complexity by 10-12 points each, and reducing the emergency scenario density by 5 points. Simultaneously with the difficulty adjustment, the module updates the node radius parameters and nursing requirement complexity parameters in the virtual patient profile: the node radius parameter corresponds to the priority of key nursing nodes in the virtual patient; when the difficulty increases, the core node radius is expanded by 20% to enhance attention, and when the difficulty decreases, it is reduced by 15%; the nursing requirement complexity parameter is related to the complexity of nursing operation steps; when the difficulty increases, operation steps are added, and when the difficulty decreases, steps are simplified, ensuring that the virtual patient profile perfectly matches the adjusted difficulty level.

[0047] The working principle and process of this invention are as follows: First, clinical case data from tertiary hospitals are integrated and desensitized and cleaned to form a structured case database. Then, AIGC technology is used to process the database data to generate dynamic case scripts containing multiple complication evolution paths. At the same time, a patient physiological model that can simulate the interaction of multiple factors such as diet, exercise, and medication is constructed, and a dynamic risk assessment model based on evidence-based medicine guidelines is embedded. This model uses an exponential decay algorithm to accumulate historical nursing data to reflect the timeliness of risk and introduces an event association weighting mechanism to strengthen the response to acute medical events. Finally, the physiological model and risk assessment results are combined to form simulated cases for skills training.

[0048] The system builds at least one mixed reality scenario, including a community follow-up scenario integrating natural language interactive virtual patients and environmental noise simulation, an emergency scenario integrating biosensor clothing pressure feedback and multi-parameter monitor linkage, and a traditional Chinese medicine nursing integration scenario integrating tongue image recognition AI and moxibustion heat simulation system. At the same time, a student profile is built through pre-testing, which covers three categories: knowledge reserve, clinical thinking, and skills operation. Finally, by combining simulated cases and student profiles, the scenario parameters are adjusted to build an adapted intelligent teaching scenario.

[0049] Through the dual-track skills training module, on the one hand, it provides trainees with a virtual drug compatibility system, a function to assess the rationality of dietary plans, and a function to detect injection angles, supporting trainees' self-training; on the other hand, it sets up a teacher guidance module, so that when trainees encounter situations such as clinical thinking block, errors in operation procedures, or inappropriate emergency response, teachers can intervene to guide thinking, provide teaching guidance on traditional Chinese medicine nursing, or simulate virtual family emotional conflict scenarios to guide trainees in coping, thus forming a dual-track training model.

[0050] The system collects operational data from trainees in the dual-track training module and conducts a weighted evaluation based on four dimensions: operational standardization, decision-making accuracy, learning efficiency, and emergency response capability to determine the trainees' ability level. Based on the evaluation results, it adaptively pushes personalized training content packages containing micro-lecture videos, VR reinforcement scenarios, and medication safety challenge games. Some content packages also include cloud-based collaborative projects.

[0051] Meanwhile, the system dynamically adjusts the training difficulty through the virtual patient adaptive matching module: first, it constructs a profile of the trainee's ability and a profile of the virtual patient, and uses the cosine similarity algorithm to calculate the matching degree to determine the initial difficulty level; during the training process, it adjusts the difficulty based on the trainee's real-time operation scores, and updates the node radius and nursing demand complexity of the virtual patient profile in sync, to ensure that the training difficulty is continuously adapted to the trainee's ability and form a teaching closed loop.

[0052] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An AI-based nursing teaching scene construction and evaluation system, characterized in that, The system comprises a simulated case module, an intelligent scene module, a double-track skill training module, and an adaptive module, wherein: The simulated case module generates a dynamic case script based on historical case data, builds a patient physiological model, and predicts the probability of complications to form a simulated case for skill training. The intelligent scene module builds an intelligent scene based on the simulated case, a mixed reality scene, and a student profile. The double-track skill training module provides a virtual drug compounding system, diet plan rationality evaluation, and injection angle detection function for students based on the intelligent scene. The adaptive module evaluates the student's operation data in the double-track skill training module and adaptively pushes personalized training content packages to the student according to the evaluation results. 2.The AI-based nursing teaching scenario construction and evaluation system according to claim 1, wherein, The simulated case integrates data from top three hospitals and generates a dynamic case script containing multiple complication evolution paths through AIGC; the patient physiological model can simulate the interaction of diet, exercise, and medication; and the complication probability prediction is based on the standards and data of evidence-based medicine guidelines. 3.The AI-based nursing teaching scenario construction and evaluation system according to claim 1, wherein, The mixed reality scene includes at least one of the following: a community follow-up scene integrating natural language interaction virtual patients and environmental noise simulation; an emergency scene integrating biological sensing clothing pressure feedback and multi-parameter monitor linkage; a traditional Chinese medicine nursing fusion scene integrating tongue image recognition AI and moxibustion heat simulation system. 4.The AI-based nursing teaching scenario construction and evaluation system according to claim 1, wherein, The student profile in the intelligent scene building module is constructed through pre-test, and the student profile includes knowledge reserve radar chart, clinical thinking feature analysis results, and skill operation baseline data; and the simulated case generation module can automatically generate personalized features for the simulated case, including the cultural background, complication risk level, and emotional state of the simulated patient. 5.The AI-based nursing teaching scenario construction and evaluation system according to claim 1, wherein, The double-track skill training module also includes a teacher guidance module for guiding clinical thinking, conducting traditional Chinese medicine nursing fusion teaching, and simulating virtual family emotional conflict scenes and organizing students for coping training; the personalized training content package pushed by the ability evaluation and content pushing module includes micro-lecture videos, VR reinforcement training scenes, and medication safety obstacle course games. 6.The AI-based nursing teaching scenario construction and evaluation system according to claim 2, wherein, The complication probability prediction is achieved through a dynamic risk assessment model that uses an exponential decay algorithm to accumulate historical nursing data to reflect the timeliness of nursing risk; and introduces an event correlation weighting mechanism to enhance the response speed and accuracy of the model to acute medical events. 7.The AI-based nursing teaching scenario construction and evaluation system according to claim 6, wherein, The dynamic risk assessment model also enhances the sensitivity to instantaneous high-risk fluctuations by building a mean neighborhood cumulative risk value, and uses a nonlinear mapping process that integrates time interval sensitive kernels to handle non-equidistant nursing data. 8.The AI-based nursing teaching scenario construction and evaluation system according to claim 1, wherein, It also includes a virtual patient adaptive matching module for dynamically adjusting the difficulty of virtual patients according to the student's operation level; the virtual patient adaptive matching module determines the adaptive virtual patient difficulty level by building a virtual patient profile and a student ability profile based on the matching degree between the two. 9.The AI-based nursing teaching scenario construction and evaluation system according to claim 8, wherein, The virtual patient adaptive matching module dynamically adjusts the nursing difficulty parameters of the virtual patient based on the real-time operation score of the trainee in skill training, including increasing the nursing difficulty when the operation score is continuously higher than the threshold, decreasing the nursing difficulty when the operation score is continuously lower than the threshold, and updating the node radius and nursing demand complexity in the virtual patient image accordingly. 10.The AI-based nursing teaching scenario construction and evaluation system according to claim 8, wherein, In the virtual patient adaptive matching module, the specific process of determining the adaptation level according to the matching degree of the trainee ability profile and the virtual patient profile is as follows: the feature parameters of each node in the virtual patient profile are extracted and matched with the node atlas in the trainee ability profile; the matching frequency of the two is counted, and the adaptation degree is calculated based on the matching frequency; and according to the adaptation degree value, the trainee is matched with a virtual case of adaptive difficulty level.

Images