Training method and system for dynamic three-dimensional ultrasound contrast of uterine fallopian tube

Abstract

The invention discloses a uterine fallopian tube dynamic three-dimensional ultrasound contrast training method and system, and the method comprises the steps: carrying out the intelligent analysis of data through an artificial intelligence algorithm according to the clinical data and historical diseases of a patient, screening out a patient group with specific pathological features, and generating a corresponding training case library; according to the training case library, performing dynamic simulation on the anatomical structures of the uterus and the fallopian tube by applying a three-dimensional modeling technology based on a virtual reality technology; the dynamic ultrasonic image and the deep learning algorithm are combined, the ultrasonic operation technique and image quality of trainees are analyzed in real time, a training feedback report is automatically generated, operation steps needing to be improved and corresponding technical key points are pointed out, and corresponding technical guidance is provided. According to the embodiment of the invention, high efficiency, systematicness and intelligence of ultrasonic training can be realized, and powerful support is provided for improving the skill level of medical image professionals.

Description

Technical Field

[0001] The present invention belongs to the field of ultrasound technology, and in particular to a training method and system for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes. Background Art

[0002] With advances in medical technology, ultrasound imaging is increasingly used in gynecology. Dynamic contrast-enhanced ultrasound (DACUS) has become a crucial diagnostic tool, particularly in the examination and evaluation of the uterus and fallopian tubes. DACUS offers real-time visualization of the physiological functions and pathological changes of the uterus and fallopian tubes, offering advantages such as being non-invasive, safe, and highly timely. However, the quality of ultrasound images and the operator's skill level directly impact the accuracy and effectiveness of the diagnosis.

[0003] Currently, many medical institutions still use traditional teaching methods for ultrasound examination training. These methods often rely on limited case data and simulation models, failing to fully integrate actual operations with clinical pathological characteristics, resulting in a lack of in-depth understanding of anatomy and pathological conditions. Furthermore, traditional training methods often rely on a master-apprentice model, lacking systematicity and objectivity, and are unable to meet the high demands of modern medicine for professional skills and practical proficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide a training method and system for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes to address the deficiencies in the existing technology, achieve high efficiency, systematicness and intelligence in ultrasound training, and provide strong support for improving the skill level of medical imaging professionals.

[0005] One embodiment of the present application provides a training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes, the method comprising: Based on the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; Based on the training case library, a three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; Combining dynamic ultrasound imaging with deep learning algorithms, it analyzes the trainees' ultrasound operation techniques and image quality in real time, automatically generates training feedback reports, points out the operation steps that need to be improved and their corresponding technical points, and provides corresponding technical guidance.

[0006] Optionally, the method uses artificial intelligence algorithms to intelligently analyze the data based on the patient's clinical information and historical symptoms, screens out patient groups with specific pathological characteristics, and generates a corresponding training case library, including: The patient's clinical information and historical disease data are standardized, and a feature selection algorithm is used to screen specific key features related to pathological characteristics. Unsupervised learning is performed on these key features to generate low-dimensional feature representations in a high-dimensional space. Build a classification model based on graph neural networks, map low-dimensional feature representations to graph nodes, and use edges to represent the relationships between different features. Graph convolutional networks are used for feature propagation and node classification to identify patient groups with specific pathological characteristics. Obtain clinical information, imaging data, and operation records corresponding to patient groups with specific pathological characteristics as the corresponding training case library.

[0007] Optionally, the training case library is used to dynamically simulate the anatomical structure of the uterus and fallopian tube using three-dimensional modeling technology based on virtual reality technology to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound, including: Analyzing the data in the training case library to determine the key anatomical features and dynamic performance of each case; Using key anatomical features, corresponding structural data is extracted from the training case library to construct a dynamic 3D model of the uterus and fallopian tube with realistic anatomical features; The dynamic three-dimensional model is imported into the virtual reality environment, and dynamic animation sequences are designed to simulate physiological processes and pathological states based on the characteristics of each training case.

[0008] Optionally, the system combines dynamic ultrasound imaging with deep learning algorithms to analyze the trainee's ultrasound operation techniques and image quality in real time, automatically generates a training feedback report, points out the operation steps that need to be improved and their corresponding technical points, and provides corresponding technical guidance, including: During the training process, the trainees' ultrasound image data of dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes are captured in real time. The ultrasound image data includes multi-angle and multi-level ultrasound image sequences. A deep learning model for ultrasound image analysis, based on a convolutional neural network, was constructed. The model was trained using a library of labeled training cases, covering various anatomical structures under normal and pathological conditions. Using deep learning models to process real-time captured ultrasound image data and perform data stream analysis, including operational correctness determination and real-time image quality monitoring; Based on the real-time analysis results, a training feedback report is automatically generated, which includes: operation technique evaluation, improvement suggestions and image quality feedback.

[0009] Another embodiment of the present application provides a training system for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes, the system comprising: The screening module is used to intelligently analyze data based on the patient's clinical information and historical symptoms using artificial intelligence algorithms to screen out patient groups with specific pathological characteristics and generate a corresponding training case library; A simulation module is used to dynamically simulate the anatomical structures of the uterus and fallopian tubes based on the training case library using three-dimensional modeling technology based on virtual reality technology, so as to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; The analysis module is used to combine dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generate training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

[0010] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when run.

[0011] Yet another embodiment of the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the above methods.

[0012] Compared with the existing technology, the present invention provides a training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes. According to the patient's clinical information and historical symptoms, an artificial intelligence algorithm is used to perform intelligent data analysis, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; based on the training case library, a three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes; combining dynamic ultrasound imaging with deep learning algorithms, the ultrasound operation techniques and image quality of the trainees are analyzed in real time, and a training feedback report is automatically generated to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance, thereby achieving efficient, systematic and intelligent ultrasound training, and providing strong support for improving the skill level of medical imaging professionals. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A hardware structure block diagram of a computer terminal for a training method for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes provided in an embodiment of the present invention; Figure 2 A schematic flow chart of a training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes provided in an embodiment of the present invention; Figure 3 A schematic structural diagram of a training system for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limiting the present invention.

[0015] The embodiment of the present invention first provides a training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes. The method can be applied to electronic devices, such as computer terminals, specifically ordinary computers.

[0016] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for a training method of uterine and fallopian tube dynamic three-dimensional ultrasound imaging provided by an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0017] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes.

[0018] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0019] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any training method for dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes.

[0020] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0021] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0022] See also Figure 2 The embodiment of the present invention provides a training method for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes, which may include the following steps: S201, based on the patient's clinical information and historical symptoms, use artificial intelligence algorithms to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; In this step, through intelligent analysis of the patient's clinical data and historical symptoms, artificial intelligence algorithms are used to screen out patient groups with specific pathological characteristics to generate a corresponding training case library. Specifically, first, patient data from the hospital's information system is collected, including age, medical history, imaging data, and previous treatment records. Then, standardization is applied to preprocess these data to eliminate data inconsistencies. Next, feature selection algorithms are used to identify key features related to specific pathological characteristics. These features will be used in subsequent unsupervised learning processes to generate low-dimensional feature representations in high-dimensional space. Finally, a classification model based on graph neural networks is constructed to convert the low-dimensional feature representation into graph nodes, and feature propagation and node classification are performed based on the relationship between nodes. Finally, patient groups with specific pathological characteristics are identified to provide data support for subsequent training cases.

[0023] The significance of this step lies in effectively gathering relevant cases to form a high-quality training case library, thereby providing a scientific basis for medical training. Through intelligent analysis using artificial intelligence, patient data with typical pathological features can be quickly and accurately screened, which not only saves medical staff time but also ensures that the selected cases are clinically representative and important. Furthermore, the resulting training case library will cover a diverse range of cases, allowing trainees to receive more comprehensive education and practice, improve their ultrasound imaging skills, and ultimately enhance the accuracy and efficiency of clinical diagnosis.

[0024] Specifically, the patient's clinical information and historical disease data can be standardized, and a feature selection algorithm can be used to screen specific key features related to pathological characteristics. Unsupervised learning can be performed on the specific key features to generate low-dimensional feature representations in a high-dimensional space. In this step, collected patient clinical and historical data are first standardized to ensure consistency and comparability. This includes normalizing numerical data and one-hot encoding categorical variables. Next, feature selection algorithms (such as random forest or LASSO regression) are used to identify key features that are highly correlated with specific pathological characteristics. These features are then used in unsupervised learning to generate a low-dimensional feature representation of each patient in a high-dimensional space, thereby revealing underlying pathological patterns.

[0025] This process is crucial for building an effective training case library. Through standardization and feature selection, we can effectively remove noise and focus on important pathological features, thereby improving the accuracy and efficiency of the model. The resulting low-dimensional feature representation can help us gain a deeper understanding of the correlations between patient data, laying the foundation for subsequent model development and ensuring the quality and practicality of the training case library.

[0026] When standardizing patient clinical data and historical medical records, ensuring data integrity and consistency is paramount. This typically involves processing multiple data types, such as numerical, categorical, and textual data. Numerical data, such as age and blood pressure, is typically normalized using z-scores to transform the data into a standard normal distribution, eliminating the influence of varying dimensions on the results. Categorical data, such as gender and disease type, is converted into binary vectors using one-hot encoding to prevent the model from misinterpreting the sequential relationships between categories. Textual data can be cleaned and segmented using natural language processing tools, laying a solid foundation for subsequent feature selection.

[0027] Next, the implementation of the feature selection algorithm is a critical step. In order to screen out key features that are highly correlated with specific pathological characteristics, a variety of algorithms can be used, such as random forest, LASSO regression, or mutual information method. These methods can evaluate the importance of each feature and automatically remove redundant features that contribute little to model prediction. Cross-validation is used to evaluate the effectiveness of feature selection to ensure that the selected features maintain high classification accuracy on different data sets. Therefore, this process is not only a data processing, but also a pursuit of optimizing model performance to ensure that the feature set used as model input is highly representative and effective.

[0028] Finally, generating low-dimensional feature representations is a core component of unsupervised learning. Dimensionality reduction techniques such as principal component analysis (PCA), t-SNE, or Unified Mapping (UMAP) can be used to map the selected high-dimensional features into a low-dimensional space. In practice, PCA uses linear transformations to identify the principal components of features, compressing the data into a lower-dimensional representation and preserving the original information to the greatest extent possible. t-SNE and UMAP are more suitable for processing nonlinear data, ensuring good separability of different data categories in a low-dimensional space. The resulting low-dimensional feature representation not only provides a rich information foundation for subsequent model training but also visualizes the underlying structure in patient data, revealing important pathological patterns.

[0029] Build a classification model based on graph neural networks, map low-dimensional feature representations to graph nodes, and use edges to represent the relationships between different features. Graph convolutional networks are used for feature propagation and node classification to identify patient groups with specific pathological characteristics. In this step, a graph neural network (GNN) model is first constructed to map the low-dimensional feature representations generated in step 1 into graph nodes. In this graph, nodes represent the feature representations of each patient, while edges represent the relationships between different features. Next, a graph convolutional network (GCN) algorithm is used to propagate the features of the nodes in the graph, updating each node's feature representation with information from neighboring nodes to obtain learned patient characteristics. Finally, through node classification, patient groups with specific pathological characteristics are identified.

[0030] By introducing graph neural networks (GNNs), this process effectively captures the relationships and interactions between patient characteristics, thereby enhancing the ability to identify pathological features in the case library. Compared to traditional machine learning methods, GNNs can process complex, non-Euclidean data, helping to identify potential pathological patterns and providing a more accurate and targeted case foundation for subsequent training.

[0031] The first step in building a graph neural network (GNN) model is to map low-dimensional feature representations into graph nodes and establish edge relationships between nodes. This process is typically based on feature similarity assessment. Specifically, we can use Euclidean distance or cosine similarity to measure the similarity between each pair of patients. Edges between nodes are connected by defining a threshold, and the weight of the edge can be set to the inverse of the similarity, making connections with high similarity closer. At the same time, to process large-scale data, a sparse graph representation can be used, retaining only edges above a certain similarity threshold to improve computational efficiency. This graph structure not only helps capture the complex relationships between patient features but also lays a good foundation for subsequent feature propagation.

[0032] The core of graph neural networks is feature propagation using graph convolutional networks (GCNs). GCNs aggregate information about neighboring nodes through multiple layers of graph convolutional layers, updating the features of each node. Implementation requires first defining the aggregation function for each layer of graph convolution. Simple operations such as summation, averaging, or max pooling can be used to integrate features from neighboring nodes. As the number of layers increases, the model is able to gradually incorporate information from more distant neighbors, thereby fully capturing the contextual features of the nodes. During training, a backpropagation algorithm is used to adjust the parameters of the graph convolutional network to optimize the classification accuracy of node features, laying a solid foundation for subsequent identification of patient groups with specific pathological characteristics.

[0033] Finally, after model training is complete, the graph neural network needs to be validated and tested to ensure its ability to accurately identify patients with specific pathological features. K-fold cross-validation can be used to evaluate the model's performance on different subsets to ensure its generalization. During testing, new patient data is input, and the trained graph convolutional network is used to make predictions, outputting the category of each node (patient). By comparing the predicted results with the true labels, the model's classification performance can be evaluated and model parameters can be further tuned as needed. Ultimately, the constructed graph neural network not only achieves effective identification of patient groups but also lays a solid data foundation for the subsequent establishment of a clinical training case library.

[0034] Obtain clinical information, imaging data, and operation records corresponding to patient groups with specific pathological characteristics as the corresponding training case library.

[0035] In this step, clinical information, imaging data, and operation records are collected from the patient population with specific pathological characteristics identified in the previous stage. This includes the patient's basic vital signs, medical records, relevant ultrasound imaging data, and physician operation logs. Through systematic data organization and storage, this information can be easily used for subsequent training and research.

[0036] This step ensures the integrity and practicality of the training case library. By integrating real patient information and imaging data, it provides trainees with a rich learning resource, allowing them to encounter real pathological conditions in a simulated learning environment, thereby improving the effectiveness and relevance of learning and laying a solid foundation for future clinical practice.

[0037] In this step, detailed clinical information must first be extracted from the patient population identified with specific pathological characteristics. This data extraction can be automated through tight integration with hospital information management systems (HIS) and electronic health record (EHR) systems. This process includes obtaining basic patient information (such as name, age, and gender), medical records (chief complaint, current medical history, past medical history, etc.), and detailed records of treatment processes. During this information extraction process, SQL can be used to query the database, ensuring efficient and accurate retrieval of qualified patient information to form a preliminary dataset.

[0038] Next, collecting imaging data is a crucial step in establishing a training case library. Imaging data is typically stored in a picture archiving and communication system (PACS). During this phase, ultrasound, CT, or MRI images relevant to the selected patients can be accessed through an interface with the PACS system. At the same time, standardized imaging data processing is essential to ensure that images acquired with different devices and at different time points can be compared and analyzed on the same platform. This not only provides rich visual information for subsequent case analysis but also helps medical trainees practice and learn based on real-world cases.

[0039] Finally, all collected clinical information, imaging data, and procedure records should be integrated into a structured training case library. Within this library, a database structure should be designed to facilitate retrieval and query, allowing physicians and trainers to quickly access the required information. Data management tools such as MongoDB or PostgreSQL can be used to build the database, and a user-friendly interface should be designed to allow users to filter based on specific criteria (such as pathological characteristics and treatment records). Furthermore, compliance with patient privacy regulations must be ensured, and data must be used and shared in a legal and compliant manner. Ultimately, this training case library will become a valuable resource for subsequent medical training, helping to enhance the professional skills of medical personnel and ultimately improve patient outcomes.

[0040] S202, based on the training case library, using three-dimensional modeling technology based on virtual reality technology, dynamically simulate the anatomical structure of the uterus and fallopian tube to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; Based on the training case library, 3D modeling technology based on virtual reality technology can be used to dynamically simulate the anatomical structures of the uterus and fallopian tubes. This process first requires an in-depth analysis of the data in the training case library to identify the key anatomical features and dynamic manifestations of each case. Using these key features, the corresponding structural data is extracted from the case library to construct dynamic 3D models with realistic anatomical features. These dynamic 3D models are then imported into a virtual reality environment, providing trainees with realistic operational scenarios. Furthermore, dynamic animation sequences are designed for each training case to vividly simulate physiological processes and pathological conditions, helping trainees intuitively and immersively understand and familiarize themselves with the anatomical changes of different cases under ultrasound in a virtual environment.

[0041] This dynamic simulation training method, based on virtual reality technology, greatly improves the interactivity and effectiveness of training. Through realistic dynamic operation scenarios, trainees can not only deepen their understanding of anatomical structures, but also practice and reflect on the operation process, thereby improving their clinical skills. At the same time, virtual reality technology can create a safer training environment, reduce the risks that may arise when operating on real patients, and enable trainees to perform multiple operation exercises without pressure. Ultimately, this training method provides important support for improving the professional capabilities and clinical level of medical personnel, helping to improve the quality of diagnosis and treatment for patients.

[0042] Specifically, the data in the training case library can be analyzed to determine the key anatomical features and dynamic performance of each case; This step first requires a systematic analysis of the case data in the training case library to identify the key anatomical features of each case. This involves carefully reviewing the patient's ultrasound imaging data to extract anatomical features related to the uterus and fallopian tubes, such as uterine morphology, the orientation of the fallopian tubes, and their diameter. Furthermore, the dynamic changes in each anatomical structure under different physiological or pathological conditions must be analyzed, such as the dilation and contraction of the fallopian tubes during ovulation. Therefore, this systematic data analysis ensures that the foundational data support for subsequent modeling and dynamic simulation is provided.

[0043] Identifying the key anatomical features and dynamic manifestations of each case helps ensure the realism and accuracy of the dynamic 3D model. This analysis not only provides precise parameters for subsequent 3D modeling but also enhances trainers' ability to understand and identify anatomical structures and variations. This process provides the necessary scientific basis for understanding physiological changes and pathological manifestations, ensuring that the resulting VR training is clinically relevant and ultimately improves training effectiveness.

[0044] The first step in data analysis is to comprehensively collect and organize relevant data from the training case library. This includes patient clinical information, ultrasound images, and previous medical records. During this process, data processing tools (such as Pandas or NumPy) can be used to clean the data to remove duplicate or missing data and ensure data completeness and accuracy. Subsequently, to facilitate subsequent analysis, a unified data format and structure is established, ensuring that all case data can be managed and retrieved within the same database. This stage focuses on building a clean, high-quality dataset, laying a solid foundation for subsequent key feature extraction.

[0045] Next, medical image processing techniques are used to analyze the ultrasound image data and extract key anatomical features. Image segmentation techniques (such as threshold segmentation and edge detection) are first applied to isolate the uterine and fallopian tube structures from the ultrasound image. Next, feature extraction algorithms (such as HOG or SIFT) are used to analyze the extracted structures and obtain their geometric features, such as area, perimeter, and shape. Furthermore, dynamic features displayed in the images are quantitatively analyzed, such as the contraction and expansion of the uterus during different physiological cycles. This step not only includes qualitative analysis but also incorporates quantitative data to fully capture the changes in anatomical features.

[0046] Finally, summarize the analysis results to identify key anatomical features and their dynamic behavior. This can be achieved by constructing a data visualization dashboard, using tools such as Matplotlib or Seaborn to present the extracted key information in graphical form. This summary document should detail the anatomical features, dynamic behavior, and changes in each training case under different pathological conditions. This documentation will also provide essential reference data for subsequent dynamic 3D modeling, ensuring that the final model accurately reflects the anatomical changes and physiological processes of different cases.

[0047] Using key anatomical features, corresponding structural data is extracted from the training case library to construct a dynamic 3D model of the uterus and fallopian tube with realistic anatomical features; In this step, based on the key anatomical features identified in the first analysis step, the corresponding structural data is extracted and a dynamic three-dimensional model is constructed. First, using computer graphics methods, the extracted anatomical feature data is converted into a three-dimensional geometric model to ensure that the model's anatomical structure truly reflects human physiology. Second, dynamic elements such as uterine contractions and fallopian tube movement are added, and physical simulation technology is used to ensure that the model can dynamically reflect physiological processes. The model must not only accurately display the anatomical features of different cases but also be able to be appropriately adjusted according to different operational requirements to provide trainees with a diverse learning experience.

[0048] The dynamic 3D model provides trainees with an intuitive learning resource, enabling them to better understand the structure and dynamic changes of the uterus and fallopian tubes. This accurate 3D model allows trainees to conduct simulated training in a virtual environment, increasing their proficiency and confidence in handling real-world cases. Furthermore, by simulating dynamic physiological processes, trainees can integrate theoretical knowledge with practical skills during operation, significantly enhancing training effectiveness.

[0049] This step first requires extracting the structural data required for key anatomical features from the analyzed training case library. To do this, database query tools (such as SQL) can be used to select specific cases and extract annotated key point data, anatomical structure dimensions, and other information. The extracted data is then organized into a format compatible with 3D modeling software, such as CSV or JSON. This ensures that this data accurately represents the coordinates, shapes, and relationships of the anatomical structures in 3D space, facilitating subsequent model construction.

[0050] Next, use 3D modeling software (such as Blender, Maya, or 3ds Max) to begin building a dynamic 3D model. Using the pre-coded data, key anatomical features are converted into 3D geometry. During this process, the model is fine-tuned using the software's modeling tools to ensure its structure accurately reflects physiological characteristics. For example, by setting appropriate boundary conditions and physical properties, the model's performance during dynamic simulations becomes more realistic. Furthermore, by adding materials and textures, the model's visual fidelity to the actual anatomy improves training immersion and learning outcomes.

[0051] Finally, the constructed dynamic 3D model undergoes physical simulation to achieve dynamic performance. For example, a physics engine (such as Unity or Unreal Engine) is used to add dynamic features to the model, enabling the uterus and fallopian tubes to simulate physiological processes such as contraction, expansion, and fluid movement. This requires not only programming the model but also setting up corresponding animation sequences to trigger the model's dynamic performance under specific conditions. This series of operations ultimately results in an accurate and dynamically expressive 3D model, laying the foundation for subsequent virtual reality training.

[0052] The dynamic three-dimensional model is imported into the virtual reality environment, and dynamic animation sequences are designed to simulate physiological processes and pathological states based on the characteristics of each training case.

[0053] In this phase, the previously constructed dynamic 3D model is imported into a virtual reality (VR) environment to create an immersive learning experience for trainees. This requires installing the necessary software tools and equipment within the VR environment, as well as configuring the environment accordingly to ensure the model's smooth operation. Next, specific dynamic animation sequences are designed based on the characteristics of each case. These animations realistically reflect physiological processes (such as uterine contractions and egg release) and pathological conditions (such as ectopic pregnancy and fallopian tube obstruction). Through this vivid simulation, trainees can visually understand the anatomical changes associated with different pathological conditions.

[0054] Combining dynamic models with a virtual reality environment not only makes training more engaging but also significantly improves learning efficiency. Through realistic simulations, trainees can practice repeatedly in a simulated environment, deepening their understanding and retention of case studies and enhancing their practical skills. This form of training can effectively reduce operational errors in real-world clinical scenarios and significantly enhance the safety and effectiveness of medical services.

[0055] Before implementing a VR environment, existing 3D models must first be further optimized and adapted to ensure they are supported by the VR system. This typically requires converting the models into formats suitable for VR platforms (such as FBX and GLTF). This process involves simplifying the model's polygon count to improve rendering efficiency and reduce latency. Furthermore, ensuring that the model's textures and materials display correctly in VR requires appropriate mapping and material settings to ensure realistic visuals.

[0056] Next, the optimized 3D model is imported into a virtual reality development platform (such as Unity or Unreal Engine). At this point, the appropriate environment needs to be set up, including lighting, background, and camera perspective, to create an immersive experience suitable for trainees. At the same time, targeted dynamic animation sequences are designed based on the characteristics of different training cases to simulate physiological processes and pathological conditions. For example, an input condition can be set so that when a pathological condition (such as fallopian tube obstruction) is simulated, the system automatically displays the corresponding anatomical changes and physiological reactions. This stage focuses on interactive design, allowing trainees to manipulate the model through gestures, controllers, or other input devices to achieve better learning outcomes.

[0057] Finally, comprehensive testing and feedback were conducted to ensure that the interaction between the VR environment and the dynamic model was responsive and accurate. A series of user tests were conducted to gather feedback from participants and assess the effectiveness of the VR training system in real-world applications. This foundation allowed for necessary adjustments to ensure a smooth and intuitive user experience, ultimately creating a highly interactive and realistic VR training environment.

[0058] S203 combines dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generating training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

[0059] The real-time analysis process, combining dynamic ultrasound imaging with deep learning algorithms, aims to assess trainees' operational techniques and image quality when performing dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes. First, during training, the system captures the trainee's ultrasound image data in real time. This data includes multi-angle, multi-layer ultrasound image sequences obtained by the ultrasound equipment. Subsequently, a deep learning model is used to rapidly process and analyze the captured image data. The model determines the correctness of the operation based on preset standards and evaluates the image quality. This system provides timely feedback on the trainee's operational performance, identifies operational steps that may require improvement, and provides guidance on corresponding technical points, thereby helping trainees better understand and master the techniques and key points of ultrasound angiography.

[0060] This real-time analysis mechanism not only improves the effectiveness of the training process but also significantly enhances the learning experience for trainees. Through immediate feedback and assessment, trainees can quickly identify deficiencies in their own procedures and implement targeted improvements. This dynamic guidance significantly accelerates the learning curve, enabling trainees to learn by doing rather than relying on traditional passive learning methods. Ultimately, this will provide strong support for improving medical staff's clinical skills and operational confidence, thereby enhancing patient diagnosis and treatment safety and effectiveness.

[0061] Specifically, during the training process, the ultrasound image data of the trainees performing dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes can be captured in real time, and the ultrasound image data includes ultrasound image sequences of multiple angles and multiple planes; To ensure authentic teaching during training, a system for capturing ultrasound image data in real time is first necessary. This can be achieved through integration with the ultrasound equipment's interface, enabling instant data transmission. Ultrasound image data acquisition should be multi-angle and multi-dimensional, meaning that images must be captured from various perspectives and sections during the scan. For example, the ultrasound probe can be moved from various positions (e.g., abdominal and vaginal) to obtain a comprehensive view. This system can collect a large amount of image data, representing the results obtained by trainees during actual operation, providing a direct basis for subsequent analysis.

[0062] Real-time ultrasound image data capture provides authentic operational feedback, enabling trainees to conduct dynamic image analysis. This approach not only captures every detail of the operation, ensuring data accuracy, but also provides trainees with more case studies for learning and analysis. This allows trainees to promptly apply theoretical knowledge learned in real-world operations, thereby improving their practical skills and ultimately achieving effective learning objectives.

[0063] First, to achieve real-time capture of ultrasound image data, an effective connection between the ultrasound device and the computer system must be ensured. This is typically achieved via a USB or network interface, with the appropriate drivers and software installed on the computer. The system must have the necessary data processing capabilities to quickly receive large amounts of image data. This process involves technicians conducting thorough system testing before training begins to ensure smooth and undelayed data transmission. Furthermore, a data storage solution must be established to ensure that captured image data can be stored and backed up in real time to prevent data loss.

[0064] Next, trainees should familiarize themselves with the ultrasound probe's operation and techniques when performing ultrasound examinations. During this time, operators should regularly examine the patient from different angles and sections to obtain comprehensive ultrasound images. This process can be achieved through the establishment of standardized operating procedures, ensuring that each trainee follows specific steps when performing dynamic 3D ultrasound imaging to ensure the diversity and integrity of the imaging data. For example, different scanning positions and techniques can be set to capture the entire view of the uterus and fallopian tubes, thereby generating a rich image library for subsequent analysis.

[0065] Finally, after each image capture, the system needs to automatically organize and annotate the captured ultrasound image data. This process can be achieved through machine learning technology. For example, the system can be equipped with a custom algorithm to automatically categorize and store images based on their characteristics and location. This not only improves data management efficiency but also provides a convenient data foundation for subsequent analysis. This real-time capture and automatic annotation method ensures that trainees have sufficient real-world data to support their operations, facilitating subsequent performance analysis and evaluation.

[0066] A deep learning model for ultrasound image analysis, based on a convolutional neural network, was constructed. The model was trained using a library of labeled training cases, covering various anatomical structures under normal and pathological conditions. This step begins with the creation of a deep learning model for ultrasound image analysis, typically using a convolutional neural network (CNN) as the underlying architecture. The data required to train this model comes from a library of annotated training cases, covering both normal and pathological anatomical structures. To build the deep learning model, the annotated data in the case library must first be preprocessed, including data cleaning, augmentation, and formatting to improve model training. Cross-validation techniques are then used to evaluate the model's performance to ensure its generalizability and adaptability to diverse case studies.

[0067] Building deep learning models can significantly improve the efficiency and accuracy of ultrasound image analysis. These models can automatically identify key features in images by learning from massive amounts of annotated data and provide rapid, real-time feedback. Compared to manual analysis, deep learning models not only reduce human error but also process complex imaging data, improving the accuracy of image quality and correctness of procedures. Ultimately, this will significantly enhance the operational skills of trained personnel, ensuring more efficient and accurate ultrasound examinations in practice.

[0068] The first step in building a deep learning model is to prepare training data. Appropriate ultrasound images are selected as training data by analyzing a library of labeled training cases. This data should cover a wide range of anatomical structures, both normal and pathological, to ensure the model can learn a rich set of features. To enhance the diversity of the training set, data augmentation techniques such as rotation, flipping, and deformation are employed to generate a wider range of image variants. This processed data provides a solid foundation for subsequent model training, ensuring enhanced generalization.

[0069] Next, a convolutional neural network (CNN) is constructed using a deep learning framework such as TensorFlow or PyTorch. The network architecture typically includes multiple convolutional layers, pooling layers, and fully connected layers to extract image features layer by layer. When building the model, the parameters of each layer must be carefully selected to ensure that the network can effectively capture detailed features in the image. Furthermore, the model is evaluated using cross-validation to ensure balanced performance on both training and test data. The key to this step is to adjust the model parameters through multiple iterations to eliminate overfitting and improve accuracy.

[0070] Finally, after model training is complete, it needs to be evaluated and optimized. This is typically done by testing the model on a validation set to ensure its generalization ability to unseen data. The evaluation results serve as the basis for further optimization, which may require adjusting the learning rate, adding regularization, and other measures to improve performance. Throughout this process, the model is ensured to be able to quickly and stably process real-time ultrasound imaging data, ensuring its future application. Through the careful design and implementation of these steps, an efficient and accurate deep learning model is constructed, laying the foundation for ultrasound image analysis.

[0071] Using deep learning models to process real-time captured ultrasound image data and perform data stream analysis, including operational correctness determination and real-time image quality monitoring; In this step, the constructed deep learning model is used to process the real-time captured ultrasound image data. By inputting real-time data into the model, the system can instantly analyze the image content and determine the correctness of the operation and the quality of the image. For example, the model can identify the presence of quality issues such as artifacts and blur in the image and issue a timely alarm. At the same time, the system can also determine whether the trainee's operation complies with the standard operating procedures by comparing them with the standard operating procedures and provide improvement suggestions. This process ensures that real-time feedback during training can help trainees make adjustments and optimizations, thereby improving learning outcomes.

[0072] Through real-time data stream analysis powered by deep learning models, trainees receive immediate feedback on their operational performance and image quality. This immediacy allows trainees to quickly correct errors during actual operations, thereby improving their skills. Furthermore, real-time monitoring of image quality helps ensure the reliability and validity of acquired data in clinical applications, providing strong support for subsequent diagnosis and decision-making. This established process ensures that training is not only scientifically sound but also achieves effective learning outcomes, further enhancing the professional skills of medical personnel.

[0073] In this step, the system first needs to input the real-time ultrasound image data stream into the deep learning model. This process involves formatting and standardizing the data to ensure that the image data can be correctly recognized and processed by the model. For example, the image needs to be adjusted to a specific pixel dimension and undergo necessary preprocessing (such as normalization) to improve the model's recognition efficiency. The data stream analysis system should have fast processing capabilities so that it can be forwarded to the deep learning framework in real time to ensure that the analysis process does not introduce significant delays.

[0074] Next, the deep learning model analyzes the input image data in real time to determine the correctness of the operation and the quality of the image. At this stage, the model not only identifies the anatomical features in the image but also detects image quality, such as image clarity, contrast, and artifacts. The analysis results will generate an evaluation report indicating the operator's performance in the current operation, such as whether best practices are being followed and whether there are any operating techniques that need adjustment. At the same time, the report will provide detailed feedback on image quality, providing comprehensive guidance for trainers.

[0075] Finally, based on the results of real-time analysis, the system needs to promptly update and adjust analysis parameters to ensure the accuracy and rationality of the assessment. This process can be achieved through the online learning mechanism of machine learning, where the model can self-adjust and optimize based on the continuous input of new data. This real-time feedback mechanism not only enhances the interactivity of training but also promotes the rapid improvement of trainees' skills. Through this continuous monitoring and feedback, trainees can receive real-time guidance in a real-world operational environment, thereby continuously improving their skills.

[0076] Based on the real-time analysis results, a training feedback report is automatically generated, which includes: operation technique evaluation, improvement suggestions and image quality feedback.

[0077] In this step, the system automatically generates training feedback reports based on real-time analysis results. These reports will include a comprehensive evaluation of the trainer's operating techniques, improvement suggestions for identified deficiencies, and feedback on ultrasound image quality. By automatically generating reports, trainers' evaluation time can be effectively saved, ensuring that they can quickly obtain the most needed information. The report can also include charts and data visualizations to help trainers more intuitively understand their own performance and problems. This systematic feedback mechanism will provide trainees with precise learning guidance and further improve their ultrasound angiography skills.

[0078] Automatically generated training feedback reports not only encourage trainees to reflect on and learn from their own procedures but also provide a basis for subsequent personalized training. By clearly outlining improvement suggestions and providing quantified image quality feedback, trainees can more clearly understand their progress and shortcomings, enabling them to make targeted technical improvements. The establishment of such a feedback mechanism contributes to the systematization and scientific nature of training, improving not only the effectiveness of training but also the satisfaction and confidence of trainees.

[0079] After synthesizing the real-time analysis results, the system needs to establish an automated report generation mechanism. First, the system extracts relevant information from the real-time analysis data stream, including operator technique assessments, image quality monitoring results, and corresponding improvement suggestions. By writing data processing scripts, this information is integrated into a structured data format to facilitate subsequent processing and presentation. During this process, to ensure accurate communication of information, the system uses natural language processing technology to convert technical data into easy-to-understand text descriptions. This allows trainees to easily understand the evaluation results and reflect on their own operations.

[0080] The system then automatically generates a feedback report that includes an assessment of operating techniques, image quality feedback, and suggestions for improvement for identified issues. To enhance the readability and intuitiveness of the report, data visualization tools such as charts and graphs are used to visualize key data. For example, a bar chart can display the accuracy of different operating steps, while a heat map can provide detailed feedback on image quality, allowing trainees to easily understand the report. This visual report not only improves the efficiency of information transmission but also enhances the learning experience for trainees.

[0081] Finally, the generated feedback report will be automatically sent to each trainee via email or the training management system. The system should also include an archiving mechanism to store the reports in a database for subsequent tracking and evaluation. Trainees can access historical feedback reports at any time for self-comparison and progress tracking. Furthermore, the system can regularly summarize and analyze all reports to identify common issues and common training needs, providing data support for improving training courses and content. This continuous feedback and evaluation mechanism helps continuously optimize training outcomes and enhance medical staff's clinical skills.

[0082] It can be seen that according to the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to perform intelligent analysis of the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; based on the training case library, three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes; combined with dynamic ultrasound imaging and deep learning algorithms, the trainees' ultrasound operation techniques and image quality are analyzed in real time, and training feedback reports are automatically generated to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance, so as to achieve efficient, systematic and intelligent ultrasound training, and provide strong support for improving the skill level of medical imaging professionals.

[0083] Another embodiment of the present invention provides a training system for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes, see Figure 3 , the system may include: The screening module 301 is used to perform intelligent data analysis based on the patient's clinical information and historical symptoms using artificial intelligence algorithms to screen out patient groups with specific pathological characteristics and generate a corresponding training case library; The simulation module 302 is configured to dynamically simulate the anatomical structures of the uterus and fallopian tubes using a three-dimensional modeling technique based on virtual reality technology based on the training case library, thereby providing a realistic operation scenario and helping trainees become familiar with the anatomical changes of different cases under ultrasound. The analysis module 303 is used to combine dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generate training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

[0084] It can be seen that according to the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to perform intelligent analysis of the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; based on the training case library, three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes; combined with dynamic ultrasound imaging and deep learning algorithms, the trainees' ultrasound operation techniques and image quality are analyzed in real time, and training feedback reports are automatically generated to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance, so as to achieve efficient, systematic and intelligent ultrasound training, and provide strong support for improving the skill level of medical imaging professionals.

[0085] An embodiment of the present invention further provides a storage medium storing a computer program, wherein the computer program is configured to execute the steps of any one of the above method embodiments when running.

[0086] Specifically, in this embodiment, the above-mentioned storage medium may be configured to store a computer program for performing the following steps: S201, based on the patient's clinical information and historical symptoms, use artificial intelligence algorithms to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; S202, based on the training case library, using three-dimensional modeling technology based on virtual reality technology, dynamically simulate the anatomical structure of the uterus and fallopian tube to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; S203 combines dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generating training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

[0087] It can be seen that according to the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to perform intelligent analysis of the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; based on the training case library, three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes; combined with dynamic ultrasound imaging and deep learning algorithms, the trainees' ultrasound operation techniques and image quality are analyzed in real time, and training feedback reports are automatically generated to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance, so as to achieve efficient, systematic and intelligent ultrasound training, and provide strong support for improving the skill level of medical imaging professionals.

[0088] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any one of the above method embodiments.

[0089] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0090] Specifically, in this embodiment, the processor may be configured to execute the following steps through a computer program: S201, based on the patient's clinical information and historical symptoms, use artificial intelligence algorithms to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; S202, based on the training case library, using three-dimensional modeling technology based on virtual reality technology, dynamically simulate the anatomical structure of the uterus and fallopian tube to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; S203 combines dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generating training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

[0091] It can be seen that according to the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to perform intelligent analysis of the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; based on the training case library, three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes; combined with dynamic ultrasound imaging and deep learning algorithms, the trainees' ultrasound operation techniques and image quality are analyzed in real time, and training feedback reports are automatically generated to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance, so as to achieve efficient, systematic and intelligent ultrasound training, and provide strong support for improving the skill level of medical imaging professionals.

[0092] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the scope of implementation of the present invention is not limited to what is shown in the drawings. Any changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which do not exceed the spirit covered by the description and drawings, should be within the scope of protection of the present invention.

Claims

1. A training method for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes, characterized in that: The method comprises: Based on the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library; Based on the training case library, a three-dimensional modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; Combining dynamic ultrasound imaging with deep learning algorithms, it analyzes the trainees' ultrasound operation techniques and image quality in real time, automatically generates training feedback reports, points out the operation steps that need to be improved and their corresponding technical points, and provides corresponding technical guidance.

2. The method according to claim 1, characterized in that Based on the patient's clinical information and historical symptoms, artificial intelligence algorithms are used to intelligently analyze the data, screen out patient groups with specific pathological characteristics, and generate a corresponding training case library, including: The patient's clinical information and historical disease data are standardized, and a feature selection algorithm is used to screen specific key features related to pathological characteristics. Unsupervised learning is performed on these key features to generate low-dimensional feature representations in a high-dimensional space. Build a classification model based on graph neural networks, map low-dimensional feature representations to graph nodes, and use edges to represent the relationships between different features. Graph convolutional networks are used for feature propagation and node classification to identify patient groups with specific pathological characteristics. Obtain clinical information, imaging data, and operation records corresponding to patient groups with specific pathological characteristics as the corresponding training case library.

3. The method according to claim 2, characterized in that Based on the training case library, the 3D modeling technology based on virtual reality technology is used to dynamically simulate the anatomical structure of the uterus and fallopian tubes to provide a realistic operation scenario and help trainers become familiar with the anatomical changes of different cases under ultrasound, including: Analyzing the data in the training case library to determine the key anatomical features and dynamic performance of each case; Using key anatomical features, corresponding structural data is extracted from the training case library to construct a dynamic 3D model of the uterus and fallopian tube with realistic anatomical features; The dynamic three-dimensional model is imported into the virtual reality environment, and dynamic animation sequences are designed to simulate physiological processes and pathological states based on the characteristics of each training case.

4. The method according to claim 3, characterized in that The system combines dynamic ultrasound imaging with deep learning algorithms to analyze trainees' ultrasound operation techniques and image quality in real time, automatically generating training feedback reports to identify operational steps that need improvement and their corresponding technical points, and providing corresponding technical guidance, including: During the training process, the trainees' ultrasound image data of dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes are captured in real time. The ultrasound image data includes multi-angle and multi-level ultrasound image sequences. A deep learning model for ultrasound image analysis, based on a convolutional neural network, was constructed. The model was trained using a library of labeled training cases, covering various anatomical structures under normal and pathological conditions. Using deep learning models to process real-time captured ultrasound image data and perform data stream analysis, including operational correctness determination and real-time image quality monitoring; Based on the real-time analysis results, a training feedback report is automatically generated, which includes: operation technique evaluation, improvement suggestions and image quality feedback.

5. A training system for dynamic three-dimensional ultrasound imaging of the uterus and fallopian tubes, characterized in that: The system comprises: The screening module is used to intelligently analyze data based on the patient's clinical information and historical symptoms using artificial intelligence algorithms to screen out patient groups with specific pathological characteristics and generate a corresponding training case library; A simulation module is used to dynamically simulate the anatomical structures of the uterus and fallopian tubes based on the training case library using three-dimensional modeling technology based on virtual reality technology, so as to provide a realistic operation scenario and help trainees become familiar with the anatomical changes of different cases under ultrasound; The analysis module is used to combine dynamic ultrasound imaging with deep learning algorithms to analyze the trainees' ultrasound operation techniques and image quality in real time, automatically generate training feedback reports to point out the operation steps that need to be improved and their corresponding technical points, and provide corresponding technical guidance.

6. The system according to claim 5, characterized in that The screening module is specifically used for: The patient's clinical information and historical disease data are standardized, and a feature selection algorithm is used to screen specific key features related to pathological characteristics. Unsupervised learning is performed on these key features to generate low-dimensional feature representations in a high-dimensional space. Build a classification model based on graph neural networks, map low-dimensional feature representations to graph nodes, and use edges to represent the relationships between different features. Graph convolutional networks are used for feature propagation and node classification to identify patient groups with specific pathological characteristics. Obtain clinical information, imaging data, and operation records corresponding to patient groups with specific pathological characteristics as the corresponding training case library.

7. The system according to claim 6, characterized in that The simulation module is specifically used for: Analyzing the data in the training case library to determine the key anatomical features and dynamic performance of each case; Using key anatomical features, corresponding structural data is extracted from the training case library to construct a dynamic 3D model of the uterus and fallopian tube with realistic anatomical features; The dynamic three-dimensional model is imported into the virtual reality environment, and dynamic animation sequences are designed to simulate physiological processes and pathological states based on the characteristics of each training case.

8. The system according to claim 7, characterized in that The analysis module is specifically used to: During the training process, the trainees' ultrasound image data of dynamic three-dimensional ultrasound angiography of the uterus and fallopian tubes are captured in real time. The ultrasound image data includes multi-angle and multi-level ultrasound image sequences. A deep learning model for ultrasound image analysis, based on a convolutional neural network, was constructed. The model was trained using a library of labeled training cases, covering various anatomical structures under normal and pathological conditions. Using deep learning models to process real-time captured ultrasound image data and perform data stream analysis, including operational correctness determination and real-time image quality monitoring; Based on the real-time analysis results, a training feedback report is automatically generated, which includes: operation technique evaluation, improvement suggestions and image quality feedback.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when run.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images