A simulation-based clinical skills multi-station examination system and method

Abstract

The application discloses a simulation type clinical skill multi-station examination system, builds a physical-virtual hybrid simulation multi-station examination platform, and fuses a high-fidelity physiological simulator, an AR / VR scene and an IoT multi-sensor array; generates a personalized examination script based on a Bayesian knowledge tracking and a genetic algorithm; collects operation, voice, eye movement and physiological data in a multi-mode; constructs an AI scoring engine with three parallel channels; realizes instant visual feedback and emotion self-adaptive adjustment; adopts PPO reinforcement learning dynamic iteration examination difficulty; and combines blockchain storage and federal learning cross-institutional collaborative modeling to form a full-link closed-loop intelligent examination system.

Description

Technical Field

[0001] This invention relates to the field of clinical medical assessment technology, specifically to a simulation-based multi-station clinical skills assessment system and method. Background Technology

[0002] With the rapid development of modern medical technology and the continuous improvement of medical service demands, higher requirements are being placed on the clinical skills training of medical personnel. In clinical medical education, the scientific and objective assessment of medical personnel's professional skill levels is of great significance for improving medical quality and ensuring patient safety. Especially in collaborative medical scenarios such as surgery and emergency care, medical personnel not only need excellent professional skills but also strong teamwork abilities.

[0003] Currently, in the field of medical education, the clinical competence of medical students and resident physicians has long been evaluated mainly through "bedside examination" or "paper-and-pencil / oral examination + simple model". These methods have the following inherent defects: (1) Single and non-reproducible cases: The evolution of the real patient's condition is uncontrollable, and it is impossible to practice repeatedly on the same case or compare the performance of different candidates; (2) Ethical and safety risks: Teaching operations on real patients pose risks of privacy infringement and medical safety; (3) Strong subjectivity in evaluation: The examiner's score is easily affected by personal experience, fatigue and "halo effect", resulting in insufficient reliability and validity; (4) Lack of immediate feedback: Candidates often only know the results several days after the examination, missing the best time for correction; (5) Limited data dimensions: Traditional scoring sheets can only record discrete indicators such as "whether it is completed" and "quality of completion", and it is difficult to capture high-resolution process data such as operation trajectory, eye movement, and voice decision-making. Summary of the Invention

[0004] The present invention aims to at least partially solve one of the technical problems existing in the related art.

[0005] The purpose of this invention is to provide a simulation-based multi-station clinical skills assessment system and method, which solves the problems of fixed assessment scenario scripts, strong subjectivity of human examiner scoring, and lack of data sharing among multiple hospitals in existing clinical assessments.

[0006] To achieve the above objectives, the present invention provides a simulation-based multi-station clinical skills assessment system, comprising:

[0007] The assessment data acquisition module is used to obtain patient data and scenario parameters corresponding to clinical assessment tasks from medical databases;

[0008] The physical-virtual hybrid simulation platform is used to construct a physical-virtual hybrid simulation assessment environment based on the data acquired by the assessment data acquisition module and according to the clinical assessment tasks using virtual simulation technology.

[0009] The multimodal data acquisition module is used to collect behavioral and physiological data generated by examinees during the simulation platform examination in real time;

[0010] The scenario script arrangement module is used to generate personalized and dynamic clinical scenario scripts for candidates based on their behavioral and physiological data during the exam.

[0011] The multimodal scoring module is used to construct three parallel scoring channels—spatiotemporal convolution, Transformer, and multi-task learning—based on candidates' behavioral and physiological data during the exam, to assess operational skills, clinical reasoning, and communication ethics, respectively.

[0012] The adaptive iterative module is used to dynamically adjust the exam difficulty, context complexity, and scoring weights based on group and individual data using the PPO reinforcement learning algorithm.

[0013] Preferably, the physical-virtual hybrid simulation platform includes no fewer than three assessment sites, which are set up in different medical assessment units, and the data between the sites is shared.

[0014] Preferably, the multimodal data acquisition module includes an IoT sensor array for real-time acquisition of behavioral data generated by examinees during the simulation platform examination; the behavioral and physiological data includes operation trajectory, voice decision-making, eye movement, and physiological indicators.

[0015] Preferably, the scenario script arrangement module is used to generate personalized dynamic clinical scenario scripts for candidates based on their behavioral and physiological data during the exam, using a Bayesian knowledge tracing model combined with Gaussian process regression and genetic algorithms.

[0016] As a preferred option, the generated clinical scenario scripts should meet the following difficulty level:

[0017]

[0018] in, For the knowledge dimension difficulty, Difficulty in terms of time pressure For the psychological load dimension of difficulty, , , These are the weighting coefficients determined using the Delphi method.

[0019] Another aspect of the present invention provides a multi-station assessment method for simulation-based clinical skills, characterized by comprising the following steps:

[0020] S1. Based on clinical assessment tasks, obtain patient data from the medical database and set up examination scenarios according to the historical assessment data of groups and individuals within the system;

[0021] S2. Arrange for candidates to undergo assessment and simultaneously collect behavioral and physiological data of candidates during the assessment process;

[0022] S3. Based on the candidates' behavioral and physiological data, three parallel scoring channels are constructed: spatiotemporal convolution, Transformer, and multi-task learning, to evaluate the candidates' operational skills, clinical reasoning, and communication ethics, respectively.

[0023] As a preferred option, step S1 specifically involves:

[0024] S11. Based on clinical assessment tasks and group assessment data within the system, a reward function is constructed using skill improvement, patient safety, and psychological load. The difficulty, situational complexity, and scoring weight of the assessment sites in this round are dynamically determined through PPO reinforcement learning.

[0025] S12. Based on the candidate's historical assessment data and the Bayesian knowledge tracing model, the candidate's weaknesses are predicted through Gaussian process regression. The genetic algorithm is used to optimize and obtain the three-dimensional difficulty coefficients of knowledge, time pressure, and psychological load, and to generate personalized medical history, physical signs evolution, and emergency event scripts.

[0026] Beneficial effects: This invention achieves a dynamic eco-city with personalized assessment scripts by combining a Bayesian knowledge tracing model with Gaussian process regression and genetic algorithms; it replaces traditional manual checklists with multimodal data and three-channel AI fusion scoring, enabling real-time visual feedback and adaptive emotion adjustment; and it uses PPO reinforcement learning to dynamically iterate the assessment difficulty. Attached Figure Description

[0027] Figure 1 This is a flowchart of a multi-station assessment method for clinical skills based on simulation, according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, embodiments of this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. In the description of this invention, it should be understood that the terminology used is for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0029] The following is combined Figure 1 This invention describes a simulation-based multi-station clinical skills assessment system and method.

[0030] Example 1: This example provides a simulation-based multi-station clinical skills assessment system, including: an assessment data acquisition module, a physical-virtual hybrid simulation platform, a multimodal data acquisition module, a scenario script arrangement module, a multimodal scoring module, and an adaptive iteration module; wherein:

[0031] The assessment data acquisition module is used to obtain patient data and scenario parameters corresponding to clinical assessment tasks from medical databases;

[0032] The physical-virtual hybrid simulation platform is used to construct a physical-virtual hybrid simulation assessment environment based on data acquired by the assessment data acquisition module and according to clinical assessment tasks using virtual simulation technology. The platform includes at least three assessment stations, each located in a different medical assessment unit, and data is shared between stations. Each station is coupled with a high-fidelity physiologically driven humanoid simulator, an AR / VR scene rendering subsystem, and an IoT sensor array. The physiologically driven simulator can present over 200 physiological parameters in real time, including hemodynamics, respiration, and pharmacological responses, and supports complex operations such as cardiopulmonary resuscitation, endotracheal intubation, and puncture. The AR / VR scene rendering can overlay CT / MRI images onto the simulator or construct virtual wards, emergency resuscitation rooms, and other dynamic environments. IoT multimodal sensing devices, such as six-dimensional force sensors, UWB positioning, and ground skin response (GSR), can collect behavioral and physiological data, including examinee operation trajectories, voice decisions, eye movements, and physiological indicators, at the millisecond level.

[0033] The scenario script arrangement module is used to generate personalized and dynamic clinical scenario scripts for candidates based on their behavioral and physiological data during the exam, using a Bayesian knowledge tracing model combined with Gaussian process regression and genetic algorithms.

[0034] The scenario script arrangement module uses historical assessment data and real-time multimodal assessment data as input, integrating Bayesian Knowledge Tracking (BKT), Gaussian Process Regression (GPR), and Genetic Algorithm (GA) in a three-layer algorithmic approach. Combined with a three-dimensional difficulty calculation formula, it accurately identifies candidates' skill gaps, quantifies assessment difficulty, and ultimately generates personalized, dynamically evolving clinical scenario assessment scripts. These scripts iterate in real-time based on the candidate's performance, completely resolving the pain points of traditional clinical assessment scripts being fixed, having a one-size-fits-all approach to difficulty, and failing to adapt to individual candidate ability differences. The formula for calculating the script difficulty value D is as follows:

[0035]

[0036] in, For the knowledge dimension difficulty, Difficulty in terms of time pressure For the psychological load dimension of difficulty, , , These are the weighting coefficients determined using the Delphi method.

[0037] The multimodal scoring module is used to construct three parallel scoring channels—spatiotemporal convolution, Transformer, and multi-task learning—based on candidates' behavioral and physiological data during the exam, to assess operational skills, clinical reasoning, and communication ethics, respectively. Multimodal data is input into a deep neural network engine, which evaluates operational techniques, diagnostic logic, and professional communication skills in three parallel paths. The training data consists of 100,000 hours of anonymized clinical videos, cross-labeled with experts using Cohen's κ ≥ 0.82, and federated learning for cross-center training.

[0038] The adaptive iterative module dynamically adjusts the exam difficulty, context complexity, and scoring weights based on group and individual data using the PPO reinforcement learning algorithm. The PPO reinforcement learning algorithm employs Proximal Policy Optimization (PPO), and the reward function R is defined as:

[0039]

[0040] in, To assess the extent to which candidates' skills have improved, The amount of improvement in patient safety indicators, To increase the psychological burden on test takers, , , The weights are determined using the Analytic Hierarchy Process (AHP).

[0041] The adaptive iterative module operates as follows: It collects individual data on examinees' actions, decisions, and physiological and psychological states in real time; retrieves historical baseline data from multiple platforms and institutions; integrates this data to generate real-time examinee ability and assessment environment status, providing a data foundation for algorithm iteration; presets adjustable system parameters, including adjustments to assessment difficulty, scenario complexity, and scoring weights; and sets upper and lower limits for difficulty based on clinical standards and patient safety requirements to ensure compliant and controllable iterative adjustments. Through a patented reward function, it uses improved examinee skills and clinical safety compliance as positive rewards and excessive psychological burden as negative penalties to quantify the merits of the current assessment scheme, clarify the direction of algorithm optimization, continuously sample assessment process data, and compare the current assessment with historical best strategies to evaluate its effectiveness. The adaptability of the core assessment scheme is controlled by the PPO cutoff mechanism to limit parameter adjustments, avoid sudden changes in assessment parameters, and ensure the stability of the assessment process. Core parameters are adaptively optimized based on evaluation results: the assessment difficulty and the complexity of the medical scenario are adjusted according to the candidate's ability level; the scoring weights for operational skills, clinical reasoning, and communication ethics are dynamically adjusted to address the candidate's weaknesses, accurately adapting to individual assessment needs. Once a single assessment achieves the optimal balance of difficulty, workload, and scoring discrimination, the specific assessment parameters are solidified. Simultaneously, assessment data is uploaded to the federated learning platform to update the group assessment baseline, which in turn empowers the front-end script orchestration module, forming a closed-loop optimization across the entire chain. This outputs personalized assessment difficulty, scenario complexity, three-dimensional dynamic scoring weights for each candidate, as well as assessment baseline parameters for global platform optimization.

[0042] Example 2: A multi-station assessment method based on simulation-based clinical skills, such as Figure 1 As shown, it includes the following steps:

[0043] S1. Based on clinical assessment tasks, obtain patient data from the medical database and set up examination scenarios according to the historical assessment data of groups and individuals within the system;

[0044] S11. Based on clinical assessment tasks and group assessment data within the system, a reward function is constructed using skill improvement, patient safety, and psychological load. The difficulty, situational complexity, and scoring weight of the assessment sites in this round are dynamically determined through PPO reinforcement learning.

[0045] S12. Based on the candidate's historical assessment data and the Bayesian knowledge tracing model, the candidate's weaknesses are predicted through Gaussian process regression. The genetic algorithm is used to optimize and obtain the three-dimensional difficulty coefficients of knowledge, time pressure, and psychological load, and to generate personalized medical history, physical signs evolution, and emergency event scripts.

[0046] S2. Arrange for candidates to undergo assessment and simultaneously collect behavioral and physiological data of candidates during the assessment process;

[0047] S3. Based on the candidates' behavioral and physiological data, three parallel scoring channels are constructed: spatiotemporal convolution, Transformer, and multi-task learning, to evaluate the candidates' operational skills, clinical reasoning, and communication ethics, respectively.

[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A simulation-based clinical skills multi-station examination system, characterized in that, include: The assessment data acquisition module is used to obtain patient data and scenario parameters corresponding to clinical assessment tasks from medical databases; The physical-virtual hybrid simulation platform is used to construct a physical-virtual hybrid simulation assessment environment based on the data acquired by the assessment data acquisition module and according to the clinical assessment tasks using virtual simulation technology. The multimodal data acquisition module is used to collect behavioral and physiological data generated by examinees during the simulation platform examination in real time; The scenario script arrangement module is used to generate personalized and dynamic clinical scenario scripts for candidates based on their behavioral and physiological data during the exam. The multimodal scoring module is used to construct three parallel scoring channels—spatiotemporal convolution, Transformer, and multi-task learning—based on candidates' behavioral and physiological data during the exam, to assess operational skills, clinical reasoning, and communication ethics, respectively. The adaptive iterative module is used to dynamically adjust the exam difficulty, context complexity, and scoring weights based on group and individual data using the PPO reinforcement learning algorithm.

2. The simulation-based multi-station clinical skills assessment system according to claim 1, characterized in that, The physical-virtual hybrid simulation platform includes no fewer than three assessment sites, which are set up in different medical assessment units, and the data between the sites is shared.

3. The simulation-based multi-station clinical skills assessment system according to claim 1, characterized in that, The multimodal data acquisition module includes an IoT sensor array, used to collect behavioral data generated by examinees during the simulation platform examination in real time; the behavioral and physiological data includes operation trajectory, voice decision-making, eye movement, and physiological indicators.

4. The simulation-based multi-station clinical skills assessment system according to claim 1, characterized in that, The scenario script arrangement module is used to generate personalized dynamic clinical scenario scripts for candidates based on their behavioral and physiological data during the exam, using a Bayesian knowledge tracing model combined with Gaussian process regression and genetic algorithms.

5. The simulation-based multi-station clinical skills assessment system according to claim 1, characterized in that, The generated clinical scenario scripts meet the following difficulty levels: ； in, For the knowledge dimension difficulty, Difficulty in terms of time pressure For the psychological load dimension of difficulty, , , These are the weighting coefficients determined using the Delphi method.

6. A multi-station assessment method based on simulation-based clinical skills, characterized in that, Includes the following steps: S1. Based on clinical assessment tasks, obtain patient data from the medical database and set up examination scenarios according to the historical assessment data of groups and individuals within the system; S2. Arrange for candidates to undergo assessment and simultaneously collect behavioral and physiological data of candidates during the assessment process; S3. Based on the candidates' behavioral and physiological data, three parallel scoring channels are constructed: spatiotemporal convolution, Transformer, and multi-task learning, to evaluate the candidates' operational skills, clinical reasoning, and communication ethics, respectively.

7. The multi-station assessment method for simulation-based clinical skills according to claim 6, characterized in that, Step S1 is as follows: S11. Based on clinical assessment tasks and group assessment data within the system, a reward function is constructed using skill improvement, patient safety, and psychological load. The difficulty, situational complexity, and scoring weight of the assessment sites in this round are dynamically determined through PPO reinforcement learning. S12. Based on the candidate's historical assessment data and the Bayesian knowledge tracing model, the candidate's weaknesses are predicted through Gaussian process regression. The genetic algorithm is used to optimize and obtain the three-dimensional difficulty coefficients of knowledge, time pressure, and psychological load, and to generate personalized medical history, physical signs evolution, and emergency event scripts.

Images