Mitral valve prolapse evaluation method and system based on three-dimensional esophageal ultrasonic 3D printing

Abstract

The invention discloses a mitral valve prolapse evaluation method and system based on three-dimensional esophageal ultrasound 3D printing, and the method comprises the steps: carrying out the denoising and enhancement processing of original image data through employing a self-adaptive multi-scale filtering algorithm according to the original image data, obtained in real time, of the three-dimensional esophageal ultrasound of a current patient, obtaining a processed ultrasonic image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism, boundary features of the mitral valve in the ultrasonic image data are automatically recognized and extracted, and an initial valve model is generated; based on the initial valve model, constructing a dynamic three-dimensional valve model capable of truly reflecting the motion state of the valve; and according to the dynamic three-dimensional valve model, performing intelligent evaluation on the mitral valve prolapse degree by combining a 3D printing technology. By utilizing the embodiment of the invention, the evaluation accuracy and efficiency of mitral valve prolapse can be improved based on the combination of the intelligent processing of the three-dimensional esophageal ultrasonic data and the three-dimensional printing technology.

Description

Technical Field

[0001] The present invention belongs to the field of image processing technology, and in particular to a method and system for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing. Background Art

[0002] Mitral valve prolapse (MVP) is a common heart valve disease characterized by abnormal protrusion of the mitral valve into the left atrium during cardiac contraction. This can lead to a range of complications, including heart failure, atrial fibrillation, and infective endocarditis. With the advancement of cardiac ultrasound technology, three-dimensional transesophageal echocardiography (TEE) has become an important tool for assessing heart valve structure and function. Compared with traditional two-dimensional ultrasound, 3D transesophageal echocardiography (TEE) provides more detailed information on valve anatomy and motion dynamics, providing clinicians with more accurate diagnostic evidence.

[0003] However, current assessment methods still have some limitations. First, traditional two-dimensional ultrasound images are easily affected by the physician's subjective judgment when calibrating valve boundaries, resulting in inconsistent results. Second, in clinical practice, the complexity of valve structure makes the manual identification and feature extraction process time-consuming and error-prone. Although image processing technologies based on machine learning and deep learning have received widespread attention and application in recent years, the relevant technologies are not yet mature in the automatic processing and analysis of three-dimensional ultrasound images, and cannot fully realize efficient and accurate valve assessment. Summary of the Invention

[0004] The purpose of the present invention is to provide a mitral valve prolapse assessment method and system based on three-dimensional esophageal ultrasound 3D printing to address the deficiencies in the existing technology. It can improve the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology.

[0005] One embodiment of the present application provides a method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing, the method comprising: Based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to perform denoising and enhancement processing on the original image data to obtain a processed ultrasound image data set; Using a deep convolutional neural network combined with a self-supervised learning mechanism, automatically identifying and extracting boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; Based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; Based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed.

[0006] Optionally, the method of using a deep convolutional neural network in combination with a self-supervised learning mechanism to automatically identify and extract boundary features of the mitral valve in the ultrasound image data to generate an initial valve model includes: Collecting historical three-dimensional esophageal ultrasound image datasets of different patients, wherein the historical three-dimensional esophageal ultrasound image datasets include images of normal mitral valves and mitral valves with varying degrees of prolapse; Constructing a deep convolutional neural network using a U-Net structure, using the historical three-dimensional esophageal ultrasound image dataset and combining it with a self-supervised learning strategy to train the deep convolutional neural network, thereby obtaining a trained deep convolutional neural network for boundary recognition; Inputting the ultrasound image data into a trained deep convolutional neural network, extracting a multi-level feature map through the convolution layer, wherein the feature map can capture multi-level information of the mitral valve boundary; Post-processing the extracted feature map, applying a global threshold or an adaptive threshold algorithm to convert the feature map into a binary image, wherein the binary image includes contour boundary information of the anterior and posterior lobes of the mitral valve; A two-dimensional boundary point cloud is generated according to the boundary information in the binary image, and the two-dimensional boundary point cloud is converted into an initial three-dimensional valve model through a three-dimensional reconstruction technology.

[0007] Optionally, constructing a dynamic three-dimensional valve model that can truly reflect the motion state of the valve based on the initial valve model includes: Based on the initial valve model, a deformation model algorithm and topology optimization technology are used to adjust the valve shape to generate a three-dimensional valve model that conforms to biomechanical characteristics; Based on the three-dimensional valve model, mechanical modeling is performed according to the stress distribution characteristics of the valve tissue to obtain a dynamic three-dimensional valve model that can truly reflect the movement state of the valve.

[0008] Optionally, the intelligent assessment of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology includes: Based on the dynamic three-dimensional valve model, a finite element analysis algorithm is applied to perform dynamic simulation to simulate the motion state of the valve under different cardiac cycles; Determining evaluation indicators based on valve motion simulation results, wherein the evaluation indicators include valve opening and closing angle, valve displacement, and stress distribution; A pre-trained machine learning model is used to quantitatively score the valve status according to the evaluation indicators. The model is trained to enable it to identify the characteristic differences between normal valves and valves with different degrees of prolapse, and output the corresponding prolapse grades. A mitral valve physical model corresponding to the dynamic three-dimensional valve model is printed using 3D printing technology to verify mitral valve prolapse and its prolapse grade.

[0009] Another embodiment of the present application provides a mitral valve prolapse assessment system based on three-dimensional esophageal ultrasound 3D printing, the system comprising: A processing module is used to perform denoising and enhancement processing on the original image data of the current patient's three-dimensional esophageal ultrasound acquired in real time using an adaptive multi-scale filtering algorithm to obtain a processed ultrasound image data set; an identification module, configured to automatically identify and extract boundary features of the mitral valve in the ultrasound image data using a deep convolutional neural network combined with a self-supervised learning mechanism, and generate an initial valve model; A construction module, configured to construct a dynamic three-dimensional valve model based on the initial valve model, which can truly reflect the movement state of the valve; An evaluation module is used to intelligently evaluate the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

[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 mitral valve prolapse assessment method based on three-dimensional esophageal ultrasound 3D printing. According to the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, the original image data is denoised and enhanced using an adaptive multi-scale filtering algorithm to obtain a processed ultrasound image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed, thereby improving the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A hardware structure block diagram of a computer terminal for a method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing provided in an embodiment of the present invention; Figure 2 A schematic flow chart of a method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing provided in an embodiment of the present invention; Figure 3 A schematic structural diagram of a mitral valve prolapse assessment system based on three-dimensional esophageal ultrasound 3D printing provided in 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 method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing. 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 method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing 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, which, when executed, can cause the processor to perform any one of the mitral valve prolapse assessment methods based on three-dimensional esophageal ultrasound 3D printing.

[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 running the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any one of the mitral valve prolapse assessment methods based on three-dimensional esophageal ultrasound 3D printing.

[0020] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1The 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 method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing, which may include the following steps: S201, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, using an adaptive multi-scale filtering algorithm to perform denoising and enhancement processing on the original image data to obtain a processed ultrasound image dataset; In the evaluation method of the present invention, the first step is to obtain the patient's three-dimensional esophageal ultrasound image data in real time, and use an adaptive multi-scale filtering algorithm to perform denoising and enhancement processing. This process aims to improve the quality of ultrasound images, making subsequent analysis and evaluation more accurate. The adaptive multi-scale filtering algorithm can effectively remove background noise while retaining detailed information of important structures such as the mitral valve by dynamically adjusting and processing the noise in different areas according to different scales and frequency characteristics. For example, in some clinical application scenarios, the patient's heart may cause blur or noise in the ultrasound image due to factors such as movement or breathing, and the adaptive multi-scale filtering can carefully adjust the filter parameters according to the grayscale changes of the image, thereby retaining the clarity of the mitral valve boundary while effectively reducing the interference signal in the image. Finally, the processed image data set will be used in subsequent steps to ensure the accurate construction and evaluation of the model.

[0023] By acquiring and processing three-dimensional esophageal ultrasound data in real time, this technical step can significantly improve the accuracy of mitral valve prolapse assessment. High-quality ultrasound image data lays a solid foundation for subsequent boundary recognition and dynamic model construction. At the same time, the introduction of an adaptive multi-scale filtering algorithm can reduce the impact of human factors on image quality, giving the system a higher level of automation and reliability. This not only improves the doctor's understanding and judgment of valve status during the assessment process, but also provides an accurate basis for timely treatment measures, thereby improving the patient's health management level.

[0024] To achieve this, the system first collects the patient's three-dimensional imaging data in real time using an esophageal ultrasound probe. This data contains rich cardiac anatomical information and dynamic changes. Next, the system applies an adaptive multi-scale filtering algorithm to process these raw ultrasound images. Specifically, the algorithm dynamically selects appropriate filtering parameters based on the local characteristics of the image. For example, in certain areas of the image, there may be a lot of background noise. In this case, the filter will increase its strength to strengthen noise suppression, while in other areas, the filter strength may be reduced to retain more detailed information.

[0025] After adaptive processing, image resolution is significantly improved, making details of the mitral valve and surrounding structures more clearly visible. This processed data is then integrated into an image dataset for subsequent deep learning model training and feature extraction. For example, suppose a patient exhibits atypical mitral valve motion during ultrasound monitoring. After processing using adaptive multi-scale filtering technology, subtle changes in the valve boundary can be clearly captured. This change is crucial for subsequent model recognition and dynamic modeling, providing physicians with accurate assessment information.

[0026] After this series of steps, the resulting processed dataset can be used to train a deep convolutional neural network to accurately identify the boundary features of the mitral valve. This in turn generates an initial valve model and a dynamic 3D valve model, providing the necessary support for detailed assessment of mitral valve prolapse. This method can effectively improve the diagnostic capabilities of mitral valve disease and the scientific nature of treatment decisions.

[0027] S202, using a deep convolutional neural network combined with a self-supervised learning mechanism to automatically identify and extract boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; In this invention, the use of a deep convolutional neural network (DCNN) combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve from ultrasound image data is an important step in assessing mitral valve prolapse. By employing a DCNN, the system is able to perform multi-level feature extraction on the input ultrasound image data. This process enables the network to gradually learn from simple low-level features (such as edges and corners) to more complex high-level features (such as the shape and structure of the valve). The introduction of a self-supervised learning strategy eliminates the need for the network to rely on large amounts of labeled data during training. Instead, the network trains itself by self-generating labels on unlabeled data, thereby improving the model's generalization ability and the accuracy of boundary recognition. Ultimately, this process generates an initial valve model, providing a foundation for subsequent dynamic modeling and evaluation.

[0028] The automated recognition and extraction process in this step significantly improves the efficiency and accuracy of mitral valve boundary identification. Traditional manual labeling methods are often influenced by the physician's subjective experience, which can easily lead to inconsistencies and errors. Deep learning algorithms, particularly those combined with self-supervised learning strategies, can provide more consistent and objective results, reducing errors caused by manual operation. Furthermore, the generated initial valve model not only lays the foundation for subsequent dynamic analysis but also provides clinicians with a more reliable basis for evaluation, ultimately improving the diagnosis and management of mitral valve prolapse.

[0029] Specifically, a historical three-dimensional esophageal ultrasound image dataset of different patients may be collected, wherein the historical three-dimensional esophageal ultrasound image dataset includes images of a normal mitral valve and mitral valves with varying degrees of prolapse; In the first step, the system collects historical 3D esophageal ultrasound imaging data from a diverse group of patients. This dataset includes images of normal mitral valves and various degrees of mitral valve prolapse, ensuring data diversity and comprehensiveness. By sourcing data from multiple medical centers, it encompasses patients of varying races, ages, and genders, providing a rich sample for subsequent model training.

[0030] This step ensures a comprehensive and diverse dataset for deep learning model training, effectively improving the model's adaptability and recognition accuracy for different mitral valve types. This rich historical data makes the model more robust in practical applications, providing a reliable basis for the assessment of mitral valve prolapse.

[0031] In the first step, the system's main task is to collect a diverse historical three-dimensional esophageal ultrasound image dataset to ensure the comprehensiveness and representativeness of model training. Specifically, the system needs to cooperate with multiple medical institutions to obtain patient ultrasound image data from hospital databases. These data sources include those from cardiovascular centers, ultrasound departments, and research hospitals, etc., with the aim of ensuring the diversity of samples, covering different genders, age groups, and multiple potential disease types. Each dataset should have complete patient information, including medical history, age, weight, gender, and specific diagnosis information, to facilitate subsequent analysis and model performance evaluation.

[0032] To improve the quality of the dataset, the team needs to follow strict data screening criteria. Only ultrasound images with high technical quality and clear resolution should be selected, avoiding images that are blurry or noisy. In addition, known normal mitral valve images and images of varying degrees of prolapse need to be balanced to ensure that the model can see a variety of different situations during training. The team also needs to take into account the differences in images produced by different medical devices. For example, the imaging parameters and imaging techniques of different ultrasound instruments may be different, which should be standardized when constructing the dataset.

[0033] Finally, the collected data needs to be organized and labeled. This process involves a professional physician reviewing each image to ensure its accuracy and relevance. During this review, the physician identifies the mitral valve structure and status in each image. This information is then used as training labels for the subsequent machine learning model. This series of steps ultimately creates a high-quality, diverse dataset, laying a solid foundation for subsequent model training.

[0034] Constructing a deep convolutional neural network using a U-Net structure, using the historical three-dimensional esophageal ultrasound image dataset and combining it with a self-supervised learning strategy to train the deep convolutional neural network, thereby obtaining a trained deep convolutional neural network for boundary recognition; A deep convolutional neural network (DCNN) was constructed using the U-Net architecture, a network architecture well-suited for medical image segmentation tasks. The model was trained using the historical 3D image dataset collected in the first step. Incorporating a self-supervised learning strategy, the network effectively learns by self-generating labels even in the absence of complete annotations. During training, the network continuously adjusts its parameters to accurately identify the mitral valve boundary.

[0035] The combination of a U-Net architecture and self-supervised learning not only improves the model's ability to identify complex structures, but also reduces reliance on manual annotation, thereby reducing labor costs and time consumption. As the network continues to train, the model's accuracy gradually improves, ultimately generating high-quality mitral valve boundary recognition results, providing important support for subsequent analysis.

[0036] In the second step, building a deep convolutional neural network with a U-Net structure is crucial for achieving boundary recognition. U-Net is a network architecture widely used in medical image segmentation tasks, designed to process feature information at various scales. During its construction, by setting appropriate convolutional, pooling, and upsampling layers, U-Net can effectively extract multiple layers of information, from low-level to high-level features. Furthermore, U-Net's skip connections mechanism preserves high-resolution features from the encoder, resulting in more accurate segmentation results during the reconstruction phase.

[0037] Next, the team will use a self-supervised learning strategy to train the constructed U-Net model. Self-supervised learning is a relatively advanced training method that allows the model to train without labeled data by constructing its own learning objectives. Specifically, the model will self-generate feature labels during each training cycle, allowing for self-correction and optimization. This strategy enables effective learning from unlabeled data, reducing the need for manual labeling and improving model training efficiency and accuracy.

[0038] During model training, the team also needed to set an appropriate loss function and optimizer. Common loss functions include cross-entropy loss and the Dice coefficient, which effectively evaluate the model's segmentation performance. Using a backpropagation algorithm and an optimizer (such as Adam or SGD), the network weights were gradually adjusted to converge the model to the optimal solution. After several rounds of training and validation, a fully trained, high-performance deep convolutional neural network for edge recognition was ultimately achieved.

[0039] Inputting the ultrasound image data into a trained deep convolutional neural network, extracting a multi-level feature map through the convolution layer, wherein the feature map can capture multi-level information of the mitral valve boundary; In this step, a trained deep convolutional neural network receives new ultrasound image data and processes it through multiple convolutional layers. Each convolutional layer extracts features of varying granularity, enabling the network to capture rich boundary information. For example, low-level features may capture edge information, while higher-level features may learn the overall shape and motion of the valve.

[0040] Extracting multi-layer features is key to achieving accurate boundary recognition, as this process ensures the network has a deep understanding of the various characteristics of the mitral valve. By extracting this rich feature information, the accuracy of model recognition can be effectively improved, laying a solid foundation for mitral valve model generation and subsequent analysis.

[0041] In the third step, the ultrasound image data is fed into a trained deep convolutional neural network (DCNN), where the network's convolutional layers are used to extract multi-layer feature maps. This process first involves preprocessing the input image, such as normalizing and rescaling, to ensure that the image fits the network's input specifications. Normalizing the ultrasound images effectively eliminates contrast differences between images and improves the model's robustness.

[0042] As ultrasound image data passes through the network's convolutional layers, the network gradually extracts features at each level. The model's lower convolutional layers typically learn simple features, such as edges and textures. As the number of network layers increases, higher convolutional layers are able to capture more complex features, including valve shape, motion, and other anatomical features. During this process, the output of each convolutional layer forms a feature map that comprehensively captures the multi-level information of the mitral valve in the ultrasound image.

[0043] Generating feature maps and extracting multi-level information are crucial for building an accurate model, as they directly impact the accuracy of subsequent boundary recognition. Through multi-layer convolutional filtering and enhancement, the network is able to capture the detailed features of the mitral valve. These details are crucial for boundary recognition and model reconstruction in subsequent steps. Therefore, the feature maps extracted at this stage provide a rich foundation for subsequent post-processing and analysis.

[0044] Post-processing the extracted feature map, applying a global threshold or an adaptive threshold algorithm to convert the feature map into a binary image, wherein the binary image includes contour boundary information of the anterior and posterior lobes of the mitral valve; The extracted feature map is post-processed using a global threshold or adaptive threshold algorithm. By setting a reasonable threshold, the important areas in the feature map are separated from the background to generate a binary image. In the binary image, the contour boundaries of the anterior and posterior lobes are clearly visible, which facilitates subsequent processing. This post-processing step is crucial because it converts complex feature information into an easy-to-understand binary image, making the contour information of the anterior and posterior lobes of the mitral valve clear and easy to analyze. Through clear boundary representation, the structural characteristics of the valve can be more intuitively reflected, laying the foundation for subsequent three-dimensional reconstruction and dynamic model generation. This not only improves the efficiency of subsequent steps, but also reduces the possibility of misidentification due to blurred boundaries.

[0045] In the fourth step, the system post-processes the extracted feature map to convert it into a binary image. This process first requires selecting an appropriate thresholding algorithm. Global thresholding is a commonly used technique. A fixed threshold is set and each pixel in the feature map is compared against it. If the threshold is exceeded, the pixel is marked as foreground (i.e., the mitral valve portion); otherwise, it is marked as background. Global thresholding is suitable for situations with large contrast differences, but in some cases, such as uneven image quality or complex structures, adaptive thresholding may be more effective. Adaptive thresholding adjusts the threshold based on the brightness information of local image regions. This method uses local statistical information (such as the mean and standard deviation) to determine the threshold for each region. This allows for accurate separation of the front and back lobe boundaries even under highly varying lighting conditions. This process effectively reduces false positives caused by uneven lighting or other interfering factors, thereby improving the quality of the binary image.

[0046] After applying the thresholding algorithm, the resulting binary image will clearly highlight the contours of the anterior and posterior mitral valve leaflets. Post-processing can further include morphological operations, such as dilation and erosion, to clean up noise and irregularities in the image. These morphological operations enhance the boundaries of the mitral valve and provide a clearer outline, thus laying a good foundation for subsequent analysis. Through these methods, the resulting binary image accurately reflects the mitral valve's anatomy and provides a clear input for dynamic model generation.

[0047] Through precise thresholding and morphological adjustments, the resulting binary image not only clearly visualizes the contours of the anterior and posterior mitral valve leaflets, but also provides a structured data source for subsequent 3D reconstruction techniques. This allows the physician or system to visually visualize the valve's anatomical features, providing visual support for subsequent clinical analysis and surgical planning.

[0048] A two-dimensional boundary point cloud is generated according to the boundary information in the binary image, and the two-dimensional boundary point cloud is converted into an initial three-dimensional valve model through a three-dimensional reconstruction technology.

[0049] The system uses the boundary information in the binary image generated in the previous step to extract the corresponding boundary points. These boundary points form a two-dimensional point cloud, which is converted into an initial three-dimensional valve model through three-dimensional reconstruction technology (such as surface-based reconstruction or voxel reconstruction methods). During the reconstruction process, the system takes into account the shape and biomechanical properties of the valve to generate a three-dimensional model that can truly reflect the valve structure. The generated initial three-dimensional valve model is the basis for subsequent dynamic analysis and evaluation. This model not only provides clear three-dimensional visual effects, but can also be used for quantitative analysis and visualization, helping doctors better understand the anatomical structure of the valve and possible pathological conditions. At the same time, the generation of three-dimensional models provides strong data support for personalized medical treatment and surgical planning, improving the accuracy and effectiveness of clinical decision-making.

[0050] In the fifth step, the system generates a 2D boundary point cloud based on the boundary information in the binary image. This process first requires identifying the boundary contours in the binary image. Typically, edge detection algorithms (such as Canny edge detection or Hough transform) are used to accurately extract the contour information. These contour points are converted into a point cloud dataset representing the boundaries of the front and back lobes on a 2D plane.

[0051] Next, the generated 2D boundary point cloud is used as input for 3D reconstruction techniques to construct an initial 3D valve model. This step can be accomplished using a variety of methods, such as surface-based reconstruction, voxel reconstruction, or polygonal mesh construction. Surface-based reconstruction methods use interpolation techniques to generate a continuous surface between boundary points, forming a smooth 3D model. Voxel reconstruction, on the other hand, represents the valve structure by creating a voxel grid in 3D space, making it suitable for processing complex shapes.

[0052] During 3D reconstruction, the system considers the valve's biomechanical properties and clinical data to ensure the generated 3D model is not only accurate in shape but also reflects the valve's actual function. This initial 3D valve model can be further used for dynamic analysis, computational fluid dynamics simulations, or clinical surgical planning. The resulting 3D model will effectively support subsequent scientific research and clinical applications.

[0053] S203, constructing a dynamic three-dimensional valve model that can truly reflect the movement state of the valve based on the initial valve model; In this method, constructing a dynamic three-dimensional valve model based on the initial valve model that can truly reflect the motion state of the valve is a crucial step in the entire evaluation process. This process mainly relies on biomechanical principles and aims to simulate the various morphological changes that the mitral valve undergoes during cardiac contraction and relaxation. By adopting deformation model algorithms and topology optimization techniques, the system can adjust the three-dimensional shape of the valve so that it not only conforms to the biological structure in static state, but also presents the actual anatomical characteristics and functional state of the valve in dynamic changes. This process involves modeling the complex mechanical properties of valve tissue, which further improves the adaptability of the model to clinical conditions.

[0054] The significance of constructing a dynamic three-dimensional valve model lies in that it provides clinicians with a more intuitive tool to assess mitral valve prolapse and its severity. By realistically simulating the movement of the valve during different cardiac cycles, doctors can more accurately identify valve dysfunction. The dynamic model not only reflects the changes in the valve under specific conditions, but also provides data support for future surgeries or treatment plans, thereby improving the effectiveness and safety of personalized treatment for patients. In addition, the application of this technology has opened up new horizons for the development of medical imaging and helped to promote in-depth research in the field of cardiovascular disease.

[0055] Specifically, based on the initial valve model, the valve shape can be adjusted using a deformation model algorithm and topology optimization technology to generate a three-dimensional valve model that conforms to biomechanical characteristics; In this step, the system will use the deformation model algorithm and topology optimization technology to adjust the initial valve model to make it more consistent with the biomechanical characteristics. The deformation model algorithm can simulate the movement of the valve under physiological conditions and take into account the different stress conditions that the valve is subjected to during the cardiac cycle. The system first needs to define the boundary conditions and applied loads of the valve movement, including the contraction, relaxation of the heart and the impact of blood flow on the valve. Then, with the help of the deformation model algorithm, the morphological changes that the valve undergoes under these conditions are simulated to ensure that the generated three-dimensional valve model can exhibit real physiological characteristics.

[0056] Generating a biomechanically accurate 3D valve model is crucial for subsequent dynamic analysis. This model not only provides a precise geometric foundation for practical operations but also offers clinicians a tool for better understanding mitral valve motion. Accurate valve models can help physicians identify and assess the valve's role in cardiac function, providing data support for developing personalized treatment plans and preoperative planning, ultimately improving patient outcomes and safety.

[0057] First, the system needs to select a suitable deformation model algorithm. Common methods include finite element analysis (FEA) based on physical phenomena, which can effectively simulate the dynamic behavior of the valve under physiological conditions. At this stage, the system will set the initial conditions of the model according to the actual valve functional requirements, such as the structural parameters, dimensions and material properties of the valve. In order to ensure the accuracy of the model under different physiological conditions, the system needs to obtain relevant biomechanical data, such as the elastic modulus, yield strength and strain rate of the mitral valve and other material properties.

[0058] After establishing the initial valve model, the system applies boundary conditions and loads. These boundary conditions include the valve's anchor points, connection points, and interactions with other cardiac structures. Furthermore, dynamic loads under physiological conditions must be applied, such as the pressure waves during cardiac contraction and their impact on the valve. By precisely specifying these conditions, the system can better simulate the valve's realistic behavior under dynamic conditions.

[0059] Next, the system dynamically simulates the valve model using a deformable modeling algorithm. Based on the pre-set cardiac cycle, the model progresses through multiple time steps, simulating the valve's behavior during contraction and relaxation. This process involves not only changes in the model's shape but also the distribution of internal stresses and strains. Ultimately, the system produces a biomechanically accurate 3D valve model, providing the necessary foundation for subsequent mechanical analysis and dynamic evaluation.

[0060] Based on the three-dimensional valve model, mechanical modeling is performed according to the stress distribution characteristics of the valve tissue to obtain a dynamic three-dimensional valve model that can truly reflect the movement state of the valve.

[0061] In this step, the system will perform detailed mechanical modeling on the generated three-dimensional valve model to analyze the stress distribution characteristics of the valve under different cardiac cycles. Mechanical modeling involves defining multiple aspects such as material properties, loads, and boundary conditions. The system will collect biomechanical property data of the valve tissue, including elastic modulus, yield strength, etc., to construct an accurate material model. Then, using the finite element analysis method, the valve is dynamically simulated under various physiological conditions to simulate the movement state of the valve during cardiac contraction and relaxation, thereby obtaining key parameters such as stress and strain.

[0062] This mechanical modeling step is crucial, not only providing a realistic physiological foundation for the dynamic 3D valve model but also providing essential metrics for subsequent valve prolapse assessment. Understanding the stress distribution of the valve under different states allows physicians to better assess valve function and its impact on overall cardiac performance. Furthermore, accurately capturing changes in valve motion will facilitate the development of personalized treatment plans, ensuring optimal patient care.

[0063] In this step, the system needs to perform detailed mechanical modeling on the generated three-dimensional valve model to analyze the stress distribution characteristics of the valve under different cardiac cycles. First, the system will determine the material properties of the valve tissue based on biomechanical properties, such as nonlinear elasticity and viscoelasticity. To this end, the system needs to consult relevant literature to obtain experimental data on the mitral valve and other heart valves, including their mechanical properties under physiological and pathological conditions. Next, these material parameters will be input into finite element analysis software (such as ANSYS or COMSOL) to accurately simulate the mechanical performance of the valve.

[0064] After setting the material properties, the system proceeds to setting boundary conditions and loads. During this stage, the system defines how the valve interacts with other cardiac structures, such as its connection points and support structures. This includes simulating the effects of blood flow on the valve, specifically the fluid dynamic loads applied to the valve during systole and diastole. Furthermore, to ensure a comprehensive model, the system considers the effects of temperature changes and physiological variations on the valve material properties.

[0065] Finally, the system will run a finite element analysis to obtain the dynamic response data of the valve under different physiological conditions. Through this process, the system can calculate key parameters such as the stress distribution, strain, and displacement of the valve during contraction and relaxation. The output of the model will include graphical representations of the stress field and strain field. Using this data, the system can comprehensively evaluate the motion state of the valve and provide solid data support for subsequent mitral valve prolapse assessment and clinical application. These results can not only be used to evaluate the functional status of the valve, but also provide a theoretical basis for future surgical planning and treatment options.

[0066] S204: Perform intelligent assessment of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

[0067] In this evaluation method, the computer system constructs a dynamic three-dimensional valve model to accurately simulate the movement behavior of the mitral valve under different physiological states. By combining 3D printing technology, the system can convert this computer-generated dynamic model into a solid three-dimensional physical model. This process includes the use of finite element analysis algorithms to evaluate the dynamic performance of the valve during the heart's contraction and relaxation cycles, which is specifically reflected in the simulation of the valve's movement state. Based on the output results of the dynamic model, the computing system can evaluate the state of the valve from multiple dimensions, such as key indicators such as valve opening and closing angles, displacement, and stress distribution. These biomechanical parameters can provide clinicians with valuable information to help formulate personalized treatment plans.

[0068] The use of dynamic three-dimensional valve models combined with 3D printing technology for intelligent assessment can provide important support for the diagnosis and treatment of mitral valve prolapse. By accurately calculating the valve's motion state, the computing system can identify different degrees of valve prolapse, thereby improving the patient's treatment plan and enhancing treatment effectiveness. The physical model allows the medical team to intuitively observe and analyze the valve structure before surgery, thereby reducing surgical risks and improving surgical success rates. In addition, by outputting quantitative scores and prolapse grades, doctors can more accurately assess changes in the patient's condition, promptly adjust treatment strategies, and ultimately achieve personalized medical services.

[0069] Specifically, a finite element analysis algorithm may be applied to perform dynamic simulation based on the dynamic three-dimensional valve model to simulate the motion state of the valve in different cardiac cycles; During this stage, the computer system uses finite element analysis algorithms to simulate the dynamic behavior of the mitral valve during cardiac contraction and relaxation. The system first imports the dynamic three-dimensional valve model into the finite element analysis software and defines the required material properties and geometric characteristics for the model, including biomechanical properties such as the thickness and elastic modulus of the valve. Then, the physiological loads and boundary conditions applied to the valve are configured to reflect the actual working environment of the heart. The computer system calculates and simulates the dynamic performance of the valve at different time points, and outputs time-related displacement, stress and strain data for further analysis. This dynamic simulation can capture subtle changes in the valve throughout the cardiac cycle, providing a quantitative data basis for subsequent evaluation.

[0070] Through dynamic simulation, the computer system not only provides a deep understanding of the valve's motion characteristics and biomechanical responses under different physiological conditions, but also provides a crucial basis for evaluating mitral valve function. The simulation results will help clinicians identify valve dysfunction and provide scientific data support for surgical planning and treatment options. This computer-assisted simulation method can effectively reduce human error and improve the objectivity and accuracy of assessments.

[0071] This step begins with a computer system acquiring a high-quality, dynamic, three-dimensional model of the valve. Typically, this model is derived from medical imaging data (such as CT or MRI scans), which is processed and reconstructed to create an accurate geometric form. Using computer-aided design (CAD) software, the computer system optimizes the model to ensure it meets the requirements of finite element analysis (FEA), including simplifying unnecessary details to reduce the computational burden.

[0072] Next, the computer system imports the dynamic 3D valve model into finite element analysis software (such as ANSYS or COMSOL). Within the software, researchers define the model's material properties, including the valve's elastic modulus, viscoelasticity, and Poisson's ratio. These biomechanical properties are parameterized based on previous experimental data or existing literature. The system then applies physiological loads, such as pressure during cardiac systole and diastole, to the model, along with valve constraints (such as fixed connection points). These conditions ensure the physiological relevance of the simulation results.

[0073] Finally, the computer system sets an appropriate time step to simulate the dynamic behavior of the valve during cardiac motion. After running the simulation, the system generates multiple frames of output, including displacement, stress field distribution, and deformation of the valve during systole and diastole. The simulation results are recorded as graphs and data files for subsequent analysis and quantitative evaluation of valve performance.

[0074] Determining evaluation indicators based on valve motion simulation results, wherein the evaluation indicators include valve opening and closing angle, valve displacement, and stress distribution; After the simulation is complete, the computer system post-processes the finite element analysis results to extract key indicators for evaluating mitral valve function. These indicators are designed to quantify the valve's state during the cardiac cycle, primarily including valve opening and closing angles, displacement, and stress distribution. The system automatically analyzes the simulation results using data processing algorithms, generating corresponding evaluation indicators and displaying them graphically for intuitive understanding and further analysis by the physician.

[0075] Through these evaluation indicators, the computer system can provide important information about the functional status of the valve, providing a scientific basis for clinicians to assess the severity of mitral valve prolapse. Changes in opening and closing angles and displacement directly reflect valve dysfunction, while analysis of stress distribution further reveals the biomechanical response of the valve under different load conditions. Through quantitative analysis, clinicians can make more accurate judgments during the evaluation process, thereby providing more personalized treatment plans for patients.

[0076] After collecting the simulation results, the computer system uses data processing algorithms to extract key information. First, the system analyzes the displacement data of the valve during contraction and relaxation, calculating the valve's opening and closing angles through geometric analysis. This process involves processing the displacement vectors, including determining the valve's baseline position and calculating the angular changes at different time points. The system then generates a graph of the opening and closing angles, providing a visual representation of the valve's motion.

[0077] Next, a computer system post-processes the stress distribution data to generate a visualization of the stress field. This process typically uses graphics processing software to convert the stress data from the finite element analysis results into a color map that reflects the stress variations in different regions. The system uses an algorithm to calculate the maximum and minimum stress values ​​at each point and analyzes areas of stress concentration based on biomechanical theory. This information is crucial for assessing the structural integrity of the valve.

[0078] Finally, the system compiles the extracted evaluation metrics into a report containing various charts and data tables. The report includes detailed information on valve opening and closing angles, displacement changes, and stress distribution, along with a discussion of the biomechanical analysis. This information will provide crucial insights for subsequent machine learning model training and clinical evaluation, ensuring the scientific and accurate nature of the evaluation process.

[0079] A pre-trained machine learning model is used to quantitatively score the valve status according to the evaluation indicators. The model is trained to enable it to identify the characteristic differences between normal valves and valves with different degrees of prolapse, and output the corresponding prolapse grades. A mitral valve physical model corresponding to the dynamic three-dimensional valve model is printed using 3D printing technology to verify mitral valve prolapse and its prolapse grade.

[0080] At this stage, the computer system will apply a trained machine learning model to quantitatively analyze the evaluation indicators. Through machine learning training on previous case data, the model is able to identify the characteristic differences between normal valves and valves with varying degrees of prolapse. The system will input the evaluation indicators obtained from step 2, quickly analyze, and output the corresponding prolapse grade and score, thereby providing clinicians with effective diagnostic basis. Through the machine learning analysis of the computer system, the evaluation process becomes highly automated, especially in terms of quantifying valve dysfunction. The machine learning model can provide objective scoring and evaluation, thereby reducing errors caused by human factors. This method can improve the efficiency of doctors' judgments and promote the accurate identification of varying degrees of mitral valve prolapse, providing support for subsequent clinical intervention.

[0081] After completing the aforementioned evaluation and quantitative scoring, a computer system will use 3D printing technology to convert the dynamic three-dimensional valve model into a physical model. This process typically involves inputting the computer-generated model file into a printer and printing it using appropriate biocompatible materials to produce a highly accurate model that conforms to the biological structure. This physical model faithfully reproduces the valve's morphology and functional characteristics and can be used for clinical validation. The printed physical model allows the medical team to visually observe and analyze the structure and kinematic characteristics of the mitral valve, thereby enhancing their understanding of the extent of prolapse. This physical model can not only be used for pre-operative rehearsal and planning, but also provide visual support during communication with the patient, helping them better understand their condition and the anticipated surgical procedure. Furthermore, this verification method will effectively improve the skills of medical professionals and the success rate of surgeries.

[0082] In this step, the computer system applies a pre-trained machine learning model to quantitatively analyze the evaluation metrics. First, the system collects and organizes historical case data, including evaluation metrics for normal valves and valves with varying degrees of prolapse. The dataset should include a variety of variables, such as valve opening and closing angles, displacement, and stress distribution, to ensure comprehensive and accurate model training. Data cleaning and processing are key to this process, ensuring that the data input into the model is accurate and consistent. Next, the computer system applies a feature selection algorithm to select the evaluation metrics most relevant to valve status, ensuring the model focuses on important features. The feature-selected data is then fed into the machine learning model. Common models include random forests, support vector machines, or deep learning algorithms such as neural networks. The system evaluates and optimizes the model using techniques such as cross-validation to improve its predictive power on new data. After training is complete, the system saves the model for use in real-world applications. During the actual evaluation process, the computer system inputs the newly acquired evaluation metrics into the trained model. The model rapidly analyzes this data, identifying the characteristic differences between normal valves and valves with varying degrees of prolapse, and outputs corresponding quantitative scores and prolapse grades. These results are compiled into visual reports for clinicians to reference, helping them make more accurate diagnostic and treatment decisions. Combined with intelligent assessments powered by machine learning, this ensures the efficiency and accuracy of the entire evaluation process.

[0083] A physical model of the dynamic 3D valve model is generated using 3D printing technology. First, based on the model's assessment results (including prolapse grade), the system automatically selects appropriate printing materials and printing parameters to ensure the biocompatibility and structural stability of the physical model. This ensures that the printed model faithfully simulates the valve's physiological properties, facilitating subsequent clinical validation. Next, the computer system slices the dynamic 3D valve model using slicing software, breaking it into segments that the 3D printer can recognize. These segments contain height and detail information for each layer, enabling the printer to accurately reconstruct the valve's shape and functional properties. During this process, the computer system also optimizes the model to reduce support structures and improve printing efficiency. Finally, after printing, the computer system performs necessary post-processing steps on the physical model, including removing printed supports and smoothing the surface. These physical models can be used for pre-operative verification and planning. Physicians can directly observe the models to better understand the valve structure and function, providing valuable reference for surgical strategy development. The physical models generated through 3D printing technology, combined with machine learning assessment results, can provide clinicians with powerful decision support, improving treatment outcomes and patient quality of life.

[0084] It can be seen that, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to denoise and enhance the original image data to obtain a processed ultrasound image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed, thereby improving the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology.

[0085] Another embodiment of the present invention provides a mitral valve prolapse assessment system based on three-dimensional esophageal ultrasound 3D printing, see Figure 3 , the system may include: The processing module 301 is configured to perform denoising and enhancement processing on the original image data of the current patient's three-dimensional esophageal ultrasound acquired in real time using an adaptive multi-scale filtering algorithm to obtain a processed ultrasound image data set; an identification module 302 for automatically identifying and extracting boundary features of the mitral valve in the ultrasound image data using a deep convolutional neural network combined with a self-supervised learning mechanism, and generating an initial valve model; A construction module 303 is configured to construct a dynamic three-dimensional valve model based on the initial valve model, which can truly reflect the motion state of the valve; The evaluation module 304 is configured to perform an intelligent evaluation of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

[0086] It can be seen that, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to denoise and enhance the original image data to obtain a processed ultrasound image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed, thereby improving the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology.

[0087] 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.

[0088] 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 original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, using an adaptive multi-scale filtering algorithm to perform denoising and enhancement processing on the original image data to obtain a processed ultrasound image dataset; S202, using a deep convolutional neural network combined with a self-supervised learning mechanism to automatically identify and extract boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; S203, constructing a dynamic three-dimensional valve model that can truly reflect the movement state of the valve based on the initial valve model; S204: Perform intelligent assessment of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

[0089] It can be seen that, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to denoise and enhance the original image data to obtain a processed ultrasound image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed, thereby improving the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology.

[0090] 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.

[0091] 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.

[0092] Specifically, in this embodiment, the processor may be configured to execute the following steps through a computer program: S201, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, using an adaptive multi-scale filtering algorithm to perform denoising and enhancement processing on the original image data to obtain a processed ultrasound image dataset; S202, using a deep convolutional neural network combined with a self-supervised learning mechanism to automatically identify and extract boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; S203, constructing a dynamic three-dimensional valve model that can truly reflect the movement state of the valve based on the initial valve model; S204: Perform intelligent assessment of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

[0093] It can be seen that, based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to denoise and enhance the original image data to obtain a processed ultrasound image data set; a deep convolutional neural network is combined with a self-supervised learning mechanism to automatically identify and extract the boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed, thereby improving the accuracy and efficiency of mitral valve prolapse assessment based on the combination of intelligent processing of three-dimensional esophageal ultrasound data and three-dimensional printing technology.

[0094] 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 method for evaluating mitral valve prolapse based on three-dimensional esophageal ultrasound 3D printing, characterized in that: The method comprises: Based on the original three-dimensional esophageal ultrasound image data of the current patient acquired in real time, an adaptive multi-scale filtering algorithm is used to perform denoising and enhancement processing on the original image data to obtain a processed ultrasound image data set; Using a deep convolutional neural network combined with a self-supervised learning mechanism, automatically identifying and extracting boundary features of the mitral valve in the ultrasound image data to generate an initial valve model; Based on the initial valve model, a dynamic three-dimensional valve model that can truly reflect the movement state of the valve is constructed; Based on the dynamic three-dimensional valve model, combined with 3D printing technology, an intelligent assessment of the degree of mitral valve prolapse is performed.

2. The method according to claim 1, characterized in that The method of using a deep convolutional neural network in combination with a self-supervised learning mechanism to automatically identify and extract boundary features of the mitral valve in the ultrasound image data and generate an initial valve model includes: Collecting historical three-dimensional esophageal ultrasound image datasets of different patients, wherein the historical three-dimensional esophageal ultrasound image datasets include images of normal mitral valves and mitral valves with varying degrees of prolapse; Constructing a deep convolutional neural network using a U-Net structure, using the historical three-dimensional esophageal ultrasound image dataset and combining it with a self-supervised learning strategy to train the deep convolutional neural network, thereby obtaining a trained deep convolutional neural network for boundary recognition; Inputting the ultrasound image data into a trained deep convolutional neural network, extracting a multi-level feature map through the convolution layer, wherein the feature map can capture multi-level information of the mitral valve boundary; Post-processing the extracted feature map, applying a global threshold or an adaptive threshold algorithm to convert the feature map into a binary image, wherein the binary image includes contour boundary information of the anterior and posterior lobes of the mitral valve; A two-dimensional boundary point cloud is generated according to the boundary information in the binary image, and the two-dimensional boundary point cloud is converted into an initial three-dimensional valve model through a three-dimensional reconstruction technology.

3. The method according to claim 2, characterized in that The step of constructing a dynamic three-dimensional valve model based on the initial valve model that can truly reflect the valve motion state includes: Based on the initial valve model, a deformation model algorithm and topology optimization technology are used to adjust the valve shape to generate a three-dimensional valve model that conforms to biomechanical characteristics; Based on the three-dimensional valve model, mechanical modeling is performed according to the stress distribution characteristics of the valve tissue to obtain a dynamic three-dimensional valve model that can truly reflect the movement state of the valve.

4. The method according to claim 3, characterized in that The intelligent assessment of the degree of mitral valve prolapse based on the dynamic three-dimensional valve model and combined with 3D printing technology includes: Based on the dynamic three-dimensional valve model, a finite element analysis algorithm is applied to perform dynamic simulation to simulate the motion state of the valve under different cardiac cycles; Determining evaluation indicators based on valve motion simulation results, wherein the evaluation indicators include valve opening and closing angle, valve displacement, and stress distribution; A pre-trained machine learning model is used to quantitatively score the valve status according to the evaluation indicators. The model is trained to enable it to identify the characteristic differences between normal valves and valves with different degrees of prolapse, and output the corresponding prolapse grades. A mitral valve physical model corresponding to the dynamic three-dimensional valve model is printed using 3D printing technology to verify mitral valve prolapse and its prolapse grade.

5. A mitral valve prolapse assessment system based on three-dimensional esophageal ultrasound 3D printing, characterized in that: The system comprises: A processing module is used to perform denoising and enhancement processing on the original image data of the current patient's three-dimensional esophageal ultrasound acquired in real time using an adaptive multi-scale filtering algorithm to obtain a processed ultrasound image data set; an identification module, configured to automatically identify and extract boundary features of the mitral valve in the ultrasound image data using a deep convolutional neural network combined with a self-supervised learning mechanism, and generate an initial valve model; A construction module, configured to construct a dynamic three-dimensional valve model based on the initial valve model, which can truly reflect the movement state of the valve; An evaluation module is used to intelligently evaluate the degree of mitral valve prolapse based on the dynamic three-dimensional valve model in combination with 3D printing technology.

6. The system according to claim 5, characterized in that The identification module is specifically used to: Collecting historical three-dimensional esophageal ultrasound image datasets of different patients, wherein the historical three-dimensional esophageal ultrasound image datasets include images of normal mitral valves and mitral valves with varying degrees of prolapse; Constructing a deep convolutional neural network using a U-Net structure, using the historical three-dimensional esophageal ultrasound image dataset and combining it with a self-supervised learning strategy to train the deep convolutional neural network, thereby obtaining a trained deep convolutional neural network for boundary recognition; Inputting the ultrasound image data into a trained deep convolutional neural network, extracting a multi-level feature map through the convolution layer, wherein the feature map can capture multi-level information of the mitral valve boundary; Post-processing the extracted feature map, applying a global threshold or an adaptive threshold algorithm to convert the feature map into a binary image, wherein the binary image includes contour boundary information of the anterior and posterior lobes of the mitral valve; A two-dimensional boundary point cloud is generated according to the boundary information in the binary image, and the two-dimensional boundary point cloud is converted into an initial three-dimensional valve model through a three-dimensional reconstruction technology.

7. The system according to claim 6, characterized in that The building blocks are specifically used for: Based on the initial valve model, a deformation model algorithm and topology optimization technology are used to adjust the valve shape to generate a three-dimensional valve model that conforms to biomechanical characteristics; Based on the three-dimensional valve model, mechanical modeling is performed according to the stress distribution characteristics of the valve tissue to obtain a dynamic three-dimensional valve model that can truly reflect the movement state of the valve.

8. The system according to claim 7, characterized in that The evaluation module is specifically used to: Based on the dynamic three-dimensional valve model, a finite element analysis algorithm is applied to perform dynamic simulation to simulate the motion state of the valve under different cardiac cycles; Determining evaluation indicators based on valve motion simulation results, wherein the evaluation indicators include valve opening and closing angle, valve displacement, and stress distribution; A pre-trained machine learning model is used to quantitatively score the valve status according to the evaluation indicators. The model is trained to enable it to identify the characteristic differences between normal valves and valves with different degrees of prolapse, and output the corresponding prolapse grades. A mitral valve physical model corresponding to the dynamic three-dimensional valve model is printed using 3D printing technology to verify mitral valve prolapse and its prolapse grade.

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