Medical training simulation system based on big data analysis

Abstract

The present application relates to a kind of medical training simulation simulation system based on big data analysis, specifically relates to the field of medical training simulation simulation, and the medical training simulation simulation system provides efficient, accurate and dynamic adaptive training environment by the cooperation of multiple modules;Data processing module ensures the high quality of input data, the generalization ability of model is improved by transfer learning and multi-task learning in construction module, the model prediction accuracy is improved by self-supervised learning and adjusting learning rate in dynamic optimization module, and real-time feedback module monitors student operation and provides personalized tutoring;The overall system combines high-quality data, intelligent training and real-time feedback, provides accurate and personalized support for medical education and training, and improves the operation skills and response capacity of students.

Description

Technical Field

[0001] This invention relates to the field of medical training simulation, and more specifically, to a medical training simulation system based on big data analysis. Background Technology

[0002] Existing medical training simulation systems based on big data analytics primarily rely on medical data extracted from large amounts of historical cases, using advanced algorithms to simulate patients. The aim of these systems is to assist doctors in skills training and decision-making practice by providing highly realistic virtual patients. Utilizing medical big data, the systems can generate diverse simulated cases, covering various age groups, genders, and disease types, thus providing doctors with a training experience close to real clinical scenarios. Current simulation systems apply traditional machine learning methods in areas such as image recognition, disease prediction, and treatment decision-making, which can improve doctors' diagnostic abilities and clinical response speed to some extent.

[0003] However, existing medical training simulation systems still have some problems, especially in terms of model accuracy and training data quality. Due to the diversity and complexity of medical data itself, existing training data often suffers from issues such as missing labels, mislabeling, or insufficient data volume, which may prevent the trained model from accurately reflecting the diversity of diseases or the real situation of patients. In addition, traditional supervised learning methods often encounter overfitting or underfitting problems when dealing with large-scale, multi-source medical data, failing to effectively improve the model's generalization ability. Furthermore, existing data cleaning and feature extraction techniques have limited effectiveness and often fail to fully utilize the potential value of the data. Therefore, how to improve the quality of training data, optimize model accuracy, and still achieve high training results under conditions of data scarcity has become the main challenge facing existing medical training simulation systems. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a medical training simulation system based on big data analysis. By comprehensively applying technologies such as multimodal data processing, transfer learning, multi-task learning, data augmentation, and generative adversarial networks, it can provide effective support in solving data scarcity, improving data quality, and optimizing model accuracy, thereby overcoming the main challenges of existing medical training simulation systems.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: it specifically includes a data processing module, a construction module, a dynamic optimization module, and a real-time feedback module;

[0006] Data processing module: Acquires multimodal medical data from medical data sources, including medical images, patient medical records, diagnostic reports, and time-series vital sign data. It uses a denoising convolutional neural network to remove artifacts from medical image data. Next, it uses an autoencoder to reconstruct the time-series vital sign data and remove outliers. Finally, it sends the processed medical data to the construction module via an AES-256 encrypted wireless transmission protocol.

[0007] The module constructs the medical training simulation model after receiving the processed medical data as input. When the amount of data is small, it borrows a model pre-trained on a large medical dataset and uses transfer learning to transfer it to a new medical task. At the same time, it uses a multi-task learning method to build a multi-task learning framework, which learns multiple medical tasks simultaneously by sharing a network structure. After the medical training model is built, if the training data is insufficient to support the training of the medical training simulation model, it performs image enhancement processing on the existing data to generate different variants to increase the diversity of the dataset. At the same time, it uses generative adversarial network technology to generate pathology-specific synthetic data to expand the training set, thereby obtaining the trained medical training simulation model.

[0008] Dynamic optimization module: continuously optimizes the parameters of the medical training simulation model through self-supervised learning and dynamic learning rate adjustment mechanism;

[0009] Real-time feedback module: Monitors student operation steps in real time, and analyzes the differences between the step sequence and the standard medical procedure. When the alarm is triggered, an audio-visual warning signal is activated, and remedial teaching videos of related cases are pushed. Based on the prediction results of the medical training simulation model, the lesion area is visualized by activating the heat map through gradient weighting. The operation error warning is associated with the lesion heat map. When the warning is triggered, the confidence level of the corresponding misdiagnosed area in the heat map is automatically highlighted.

[0010] In a preferred embodiment, the data processing module includes medical data sources such as a data open platform and hospitals. The denoising convolutional neural network comprises five residual blocks, each consisting of two 3×3 convolutional layers, a batch normalization layer, and a ReLU activation function. Its loss function is:

[0011] ;

[0012] in, This represents the loss function value of the denoising convolutional neural network. Represents the position of all pixels in a medical image. Perform double summation. Indicates the location of the original medical image. Pixel value at that location, Indicates the location of the denoised medical image. Pixel value at that location, These represent the height and width dimensions of the medical image, respectively. Represents the total pixels Take the average value;

[0013] The residual blocks of the denoising convolutional neural network use skip connections, and the last layer has 64 convolutional kernels.

[0014] In a preferred embodiment, the medical tasks include disease classification, symptom regression, and pathology report generation, all sharing a feature extraction layer. This feature extraction layer employs a Transformer architecture and fuses image features with textual medical record features through a multi-head attention mechanism. Its multi-task loss function is:

[0015] ;

[0016] in, This represents the total loss value for multi-task learning. This represents the weight coefficient of each task. The cross-entropy loss function represents the task of disease classification. This represents the loss function for the symptom regression task. This represents the negative log-likelihood loss function for the pathology report generation task.

[0017] In a preferred embodiment, the image enhancement process specifically includes rotation, scaling, and grayscale perturbation.

[0018] In a preferred embodiment, the generator objective function of the generative adversarial network is:

[0019] ;

[0020] in, This represents the total loss function of the generator. Represents the noise vector and pathological condition labels The joint distribution takes the expected value. The generator is based on noise. and pathological conditions The generated synthetic data, This represents the probability that the discriminator distinguishes the generator. Indicates the counter-loss term. express Regularization coefficient, and , express Reconstruction loss.

[0021] In a preferred embodiment, the dynamic optimization module employs a mask self-supervised learning strategy to randomly mask the input data. The region updates the parameters of the medical training simulation model by predicting the content to be occluded.

[0022] In a preferred embodiment, the dynamic learning rate adjustment mechanism is specifically as follows:

[0023] If the validation set loss does not decrease for three consecutive epochs, the learning rate is adjusted accordingly. Attenuation, in which Indicates the number of decay cycles.

[0024] In a preferred embodiment, in the real-time feedback module, the degree of difference The calculation is performed using the dynamic time warping algorithm, and the specific formula is as follows:

[0025] ;

[0026] in, This indicates the total number of time steps in the operation. Indicates the time index of the operation step. Indicates the range from 1 to 1 The differences in each step are summed. Indicates the student's steps The operational feature vector, Indicate steps The standard operating feature vector.

[0027] In a preferred embodiment, the formula for calculating the heat map is:

[0028] ;

[0029] in, Representing a heat map The activation intensity value, Represents the linear rectified function. Indicates to The activation values ​​of each feature map are weighted and summed. Indicates the first Each feature map Indicates the first Each feature map at location Activation value at that location, Indicates the classification weight.

[0030] In a preferred embodiment, the real-time feedback module automatically matches the heatmap display mode according to the warning type, specifically including:

[0031] S1. If a misdiagnosis warning is issued, the heat map will focus on the area of ​​missed lesions;

[0032] S2. If an operation timeout warning is issued, the heat map will mark key anatomical structures to guide the path.

[0033] The beneficial effects of this invention are as follows: This medical training simulation system provides an efficient, accurate, and dynamically adaptable training environment through the collaborative work of multiple modules; the data processing module ensures high-quality input data; the construction module improves the generalization ability of the model through transfer learning and multi-task learning; the dynamic optimization module improves the model's prediction accuracy through self-supervised learning and adjusting the learning rate; and the real-time feedback module monitors student operations and provides personalized guidance. The overall system combines high-quality data, intelligent training, and real-time feedback to provide precise and personalized support for medical education and training, thereby improving students' operational skills and coping abilities. Attached Figure Description

[0034] Figure 1 This is a flowchart of the method of the present invention;

[0035] Figure 2 This is a block diagram of the system structure of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0038] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0039] Example 1

[0040] This embodiment provides, for example Figure 1-2 The medical training simulation system based on big data analysis is shown, specifically including: a data processing module, a construction module, a dynamic optimization module, and a real-time feedback module;

[0041] The data processing module acquires multimodal medical data from medical data sources, including medical images, patient medical records, diagnostic reports, and time-series vital sign data. It uses a denoising convolutional neural network to remove artifacts from medical image data while preserving key features. This improves the accuracy of medical image diagnosis, especially in identifying lesion areas. Secondly, since time-series vital sign data often contains noise and sudden outliers, an autoencoder is used to reconstruct the data and remove outliers, ensuring the stability of the training data. Finally, the processed medical data is transmitted to the construction module via an AES-256 encrypted wireless transmission protocol. This module's design ensures data quality in practical applications, laying a solid foundation for subsequent training and model optimization.

[0042] The construction module receives processed medical data and uses it as input to begin building a medical training simulation model. When the data volume is small, it borrows a model pre-trained on a large medical dataset and uses transfer learning to transfer it to new medical tasks. For example, a convolutional neural network model trained on a large amount of chest X-ray data can be transferred to the medical task of lung disease detection, reducing the dependence on large-scale data. Simultaneously, a multi-task learning framework is built using multi-task learning methods, learning multiple medical tasks simultaneously by sharing the network structure, improving the model's generalization ability. After the medical training model is built, it is used in medical... In teaching, especially with medical imaging data, there is often a limited sample size. If the training data is insufficient to support the training of the medical training simulation model, image enhancement processing is performed on the existing data to generate different variants, thereby increasing the diversity of the dataset. At the same time, generative adversarial network technology is used to generate pathology-specific synthetic data to expand the training set. For practical training, this can generate diverse simulation data to help students practice and test in different scenarios, thereby obtaining a trained medical training simulation model. This module can provide diverse training data as needed to ensure that the medical training simulation system can cope with different training scenarios.

[0043] Dynamic optimization module: Continuously optimizes the parameters of the medical training simulation model through self-supervised learning and dynamic learning rate adjustment mechanism; this module can improve the prediction accuracy of the simulation system and its adaptability to dynamic changes, and is very effective for the constantly changing data flow in practical applications;

[0044] Real-time feedback module: Monitors student operation steps in real time, and analyzes the differences between the step sequence and the standard medical procedure. At that time, its difference threshold Based on clinical expert experience, the system triggers audio-visual warning signals and pushes remedial teaching videos for related cases. According to the prediction results of the medical training simulation model, the system activates the visualized lesion area on the heat map through gradient weighting. The system associates the operation error warning with the lesion heat map. When the warning is triggered, the system automatically highlights the confidence level of the corresponding misdiagnosed area on the heat map. This module design can improve the user experience, especially in education and practical training environments.

[0045] In this embodiment, the data processing module needs to be specifically described. The medical data sources include a data open platform and hospitals. The denoising convolutional neural network contains 5 residual blocks. Each residual block consists of two 3×3 convolutional layers, a batch normalization layer, and a ReLU activation function. Its loss function is:

[0046] ;

[0047] in, This represents the loss function value of the denoising convolutional neural network, used to evaluate the denoised image. With original medical images The smaller the value, the better the noise reduction effect. Represents the position of all pixels in a medical image. Perform a double summation, covering the data from row 1, column 1 to column 2. Line number All pixels in the column, Indicates the location of the original medical image. Pixel value at that location, Indicates the location of the denoised medical image. The pixel value at that location is the output of the denoising convolutional neural network. These represent the height and width dimensions of the medical image, respectively. Represents the total pixels Take the average value;

[0048] The residual blocks of the denoising convolutional neural network use skip connections, and the last layer has 64 convolutional kernels.

[0049] Before using an autoencoder, it is necessary to preprocess the time series vital signs data. The specific steps include:

[0050] S1. Use linear interpolation, an interpolation method, to handle missing data points;

[0051] S2. Standardize the data so that the autoencoder can effectively process input data of different dimensions;

[0052] S3. Sliding window: For time series biometric data, the sliding window method is used to divide the data, which facilitates model training. The biometric data of the time series in each window will be used as an input sample of the autoencoder.

[0053] Secondly, the goal of the autoencoder is to reconstruct the input data, which here refers to time-series life characteristic data. Therefore, the difference between the input data and the reconstructed data (i.e., the reconstruction error) can be used to detect outliers. In this application, the mean squared error between the input data and the reconstructed data at each time point is calculated. If the reconstruction error at a certain time point is greater than a preset threshold, then that point is considered an outlier. The formula for calculating the mean squared error is as follows:

[0054] ;

[0055] in, Raw biometric data representing time series. This represents the output data of the reconstruction. Indicates the number of data points;

[0056] Meanwhile, to enable more flexible and accurate outlier detection, this application designs an adaptive threshold formula based on the distribution of reconstruction error. This formula dynamically adjusts the threshold according to the error distribution of the training data, thereby improving the accuracy of the model. Specifically, the reconstruction error follows a normal distribution, and a threshold is dynamically set by calculating the mean and standard deviation of the error. The formula for the adaptive threshold is as follows:

[0057] ;

[0058] in, This represents the mean reconstruction error of the training data. The standard deviation of the reconstruction error is represented by the standard deviation of the reconstruction error. This represents an adjustable hyperparameter that controls the detection sensitivity. Furthermore, in this formula, The larger the value, the higher the threshold, and the fewer outliers will be detected. The smaller the value, the lower the threshold, and the more outliers will be detected. The most suitable value is selected through cross-validation.

[0059] In this embodiment, the construction module needs to be specifically described. The medical tasks include disease classification, symptom regression, and pathology report generation, all sharing a feature extraction layer. The feature extraction layer adopts a Transformer architecture and fuses image features and text medical record features through a multi-head attention mechanism. Its multi-task loss function is:

[0060] ;

[0061] in, This represents the total loss value for multi-task learning. This represents the weight coefficient of each task, used to adjust the proportion of contribution of different tasks to the total loss, and satisfies... This ensures weight normalization and prevents the training process from being dominated by a single task. The cross-entropy loss function for disease classification tasks is calculated using the following formula:

[0062] ;

[0063] in, This represents the unique hot code of the true label, where the disease category is... ,but Otherwise, it is 0. This represents the probability of disease categories predicted by the medical training simulation model;

[0064] The loss function for the symptom regression task, in this application, uses mean squared error, and the specific calculation formula is as follows:

[0065] ;

[0066] in, Indicates the first The true symptom scores of each sample, such as dyspnea index and pain level. This represents the symptom score predicted by the medical training simulation model;

[0067] The negative log-likelihood loss function represents the pathology report generation task, and its specific calculation formula is as follows:

[0068] ;

[0069] in, This indicates the first [item] in the pathology report. One word, This indicates that the medical training simulation model is based on the preceding words. and input data The probability of generating the current word;

[0070] Image enhancement processing specifically includes rotation Scaling and grayscale value perturbation ;

[0071] The generator objective function of a generative adversarial network is:

[0072] ;

[0073] in, This represents the generator's total loss function, used to guide generator optimization so that the generated synthetic data can both fool the discriminator and approximate the real data distribution. Represents the noise vector and pathological condition labels The joint distribution takes the expected value. The generator is based on noise. and pathological conditions The generated synthetic data, Let represent the discriminant's probability of classifying the generator, where This represents the confidence level at which the discriminator considers the data to be true. This represents the adversarial loss term, the purpose of which is to force the discriminator to make a judgment probability on the synthetic data by minimizing this term. The value approaches 0, meaning the discriminator struggles to distinguish between real and synthetic data. express Regularization coefficient, and The purpose is to prevent the generator from over-optimizing to combat loss and ignoring the authenticity of the data. express Reconstruction loss is used to calculate the generated data. With real data The absolute difference at the pixel level, among which .

[0074] In this embodiment, the dynamic optimization module needs to be specifically explained. Because medical data is constantly updated, the medical training simulation model also needs to continuously adapt to new data. Therefore, the dynamic optimization module adopts a masking self-supervised learning strategy to randomly mask the input data. area, of which The range of values ​​for is: The parameters of the medical training simulation model are updated by predicting the occluded content.

[0075] The dynamic learning rate adjustment mechanism is as follows:

[0076] If the validation set loss does not decrease for three consecutive epochs, the learning rate is adjusted accordingly. Attenuation, in which Indicates the number of decay cycles.

[0077] In this embodiment, the real-time feedback module and the degree of difference need to be specifically explained. The calculation is performed using the dynamic time warping algorithm, and the specific formula is as follows:

[0078] ;

[0079] in, This indicates the total number of time steps in the operation. Indicates the time index of the operation steps, from , Indicates the range from 1 to 1 The differences in each step are summed. Indicates the student's steps The operation feature vector contains operation data in multiple dimensions, among which typical features include tool selection, operation position, operation force, and operation duration. Indicate steps The standard operating characteristic vector is specifically defined by expert experience or clinical guidelines;

[0080] The formula for calculating a heat map is:

[0081] ;

[0082] in, Representing a heat map The activation intensity value also indicates the importance of that location to the model's prediction results. A higher value, i.e., a warmer color, indicates a greater contribution of the lesion area to the model's decision. Represents a linear rectified function, located as The purpose is to filter out negative values ​​and retain only positive activations, ensuring that the heatmap only highlights regions that contribute positively to the classification. Indicates to The activation values ​​of each feature map are weighted and summed. Indicates the first Each feature map Indicates the first Each feature map at location Activation value at that location, Indicates the classification weight;

[0083] The real-time feedback module automatically matches the heatmap display mode according to the warning type, specifically including:

[0084] S1. If a misdiagnosis warning is issued, the heat map will focus on the area of ​​missed lesions;

[0085] S2. If an operation timeout warning is issued, the heat map will mark key anatomical structures to guide the path.

[0086] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0087] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0088] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0091] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0092] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A medical training simulation system based on big data analysis, characterized in that, Specifically comprising: The data processing module, the construction module, the dynamic optimization module and the real-time feedback module; The data processing module: obtaining multi-modal medical data from a medical data source, the multi-modal medical data including medical images, patient medical records, diagnosis reports and time series vital sign data, removing artifacts in the medical image data using a denoising convolutional neural network, then using an autoencoder to reconstruct the time series vital sign data and remove outliers, and finally sending the processed medical data to the construction module through an AES-256 encryption wireless transmission protocol; The construction module: receiving the processed medical data as input and starting to build a medical training simulation model, when the data volume is small, borrowing a pre-trained model on a large medical data set, using transfer learning to migrate it to a new medical task, and using a multi-task learning method to build a multi-task learning framework to learn multiple medical tasks simultaneously, after the medical training model is built, if the training data is insufficient to support the medical training simulation model training, image enhancement processing is performed on the existing data to generate different variants to increase the diversity of the data set, and path-specific synthetic data is generated using a generative adversarial network technology to expand the training set, and then a trained medical training simulation model is obtained; The generator objective function of the generative adversarial network is: ； wherein, represents a total loss function of the generator, represents a joint distribution of the noise vector and the pathological condition label taking an expectation value, represents synthetic data generated by the generator from the noise and the pathological condition , represents a discrimination probability of the discriminator on the generator, represents an adversarial loss term, represents a regularization coefficient, and , represents a reconstruction loss; The dynamic optimization module: continuously optimizing the medical training simulation model parameters through self-supervised learning and dynamic learning rate adjustment mechanism; Real-time feedback module: real-time monitoring of student operation steps, when the difference between the step sequence and the standard medical process When the difference between the step sequence and the standard medical process is triggered, an audible warning signal is triggered, and a remedial teaching video associated with the case is pushed. According to the prediction result of the medical training simulation model, the lesion area is visualized through gradient weighted class activation heat map, the operation error warning is associated with the lesion heat map, and when the warning is triggered, the confidence of the corresponding misdiagnosis area in the heat map is automatically highlighted.

2. The medical training simulation system based on big data analysis according to claim 1, characterized in that: In the data processing module, the medical data source includes a data open platform and a hospital, the denoising convolutional neural network contains 5 residual blocks, each residual block is composed of two 3x3 convolution layers, a batch normalization layer and a ReLU activation function, and its loss function is: ； wherein, denotes a loss function value of the denoising convolutional neural network, denotes a double summation over all pixel positions of the medical image, denotes a pixel value of the original medical image at position , denotes a pixel value of the denoised medical image at position , denotes the height and width dimensions of the medical image, respectively, denotes a total pixel value of the medical image, and denotes an average value. The residual blocks of the denoising convolutional neural network use skip connection, and the number of convolution kernels in the last layer is 64.

3. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The medical tasks include disease classification tasks, symptom regression tasks and pathology report generation tasks, and share a feature extraction layer, the feature extraction layer uses a Transformer architecture, and fuses image features and text medical record features through a multi-head attention mechanism, and its multi-task loss function is: ； wherein, denotes a total loss value of multi-task learning, denotes a task weight coefficient, denotes a cross-entropy loss function of a disease classification task, denotes a loss function of a symptom regression task, denotes a negative log-likelihood loss function of a pathology report generation task.

4. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The image enhancement processing specifically includes rotation, scaling and gray value disturbance.

5. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The dynamic optimization module adopts a mask self-supervised learning strategy, randomly masks the input data Region, update the medical training simulation model parameters by predicting the masked content.

6. The big data analytics based medical training simulation system according to claim 1, wherein: The dynamic learning rate adjustment mechanism specifically is: When the validation set loss does not decrease for 3 consecutive epochs, then the learning rate is decayed by where denotes the number of decays.

7. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The real-time feedback module, the difference degree The dynamic time warping algorithm is calculated, and the specific calculation formula is: ； wherein, represents the total number of time steps of the operation step, represents the time index of the operation step, represents the cumulative sum of the difference from the first to the step, represents the operation feature vector of the student at step , represents the standard operation feature vector of step .

8. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The heat map calculation formula is: ； in, Representing a heat map The activation intensity value, Represents the linear rectified function. Indicates to The activation values ​​of each feature map are weighted and summed. Indicates the first Each feature map Indicates the first Each feature map at location Activation value at that location, Indicates the classification weight.

9. The big data analytics based medical training simulation system as claimed in claim 1, wherein: The real-time feedback module automatically matches the heat map display mode according to the warning type, specifically including: S1, if misdiagnosis warning, the heat map focuses on the missed lesion area; S2, if operation timeout warning, the heat map marks the key anatomical structure guide path.

Images