A method and device for displaying the cephal vein in arteriovenous fistula formation surgery.

Abstract

This invention provides a method and apparatus for displaying the cephalic vein during arteriovenous fistula (AVF) reconstruction, relating to the field of endovascular disease diagnosis technology. The method includes: denoising the preprocessed raw radio frequency signal using a phase unwrapping algorithm and wavelet transform to generate a first signal; extracting vascular contour information from the first signal; generating a three-dimensional vascular model based on the vascular contour information and the first signal; determining the intravascular pressure gradient distribution using a hemodynamic simulation algorithm; matching vascular bifurcation points based on pre-AVF imaging data; generating a vascular topology map by combining the intravascular pressure gradient distribution and vascular bifurcation points; generating a cephalic vein probability distribution heatmap using a multi-scale convolutional neural network; spatially registering the cephalic vein probability distribution heatmap with real-time ultrasound images; superimposing the spatially registered image onto a terminal interface; and marking the AVF location coordinates on the spatially registered image. This achieves precise localization of the cephalic vein.

Description

Technical Field

[0001] This invention relates to the field of endovascular disease diagnosis technology, and in particular to a method and apparatus for displaying the cephal vein during arteriovenous fistula formation. Background Technology

[0002] Arteriovenous fistula creation is a crucial surgical procedure for establishing vascular access in hemodialysis patients, and its outcome directly impacts the patient's dialysis quality and quality of life. Accurately locating and managing the cephalic vein is paramount during the procedure; however, numerous challenges currently exist in clinical practice regarding cephalic vein localization.

[0003] Traditional surgical methods rely primarily on the surgeon's experience and simple surface landmarks to locate the cephalic vein. However, individual differences exist in the distribution of blood vessels in the human body. In some patients, the cephalic vein is located deep, has a small diameter, or exhibits vascular variations, making accurate location difficult to achieve based solely on experience. This not only increases the difficulty and risk of the surgery and prolongs the operation time but may also lead to surgical failure, requiring a second surgery and causing additional pain and financial burden for the patient.

[0004] Existing auxiliary localization techniques, such as ultrasound examination alone, can show the general shape of blood vessels, but cannot provide doctors with accurate information about the location of the cephalic vein. The ability to identify blood vessels in ultrasound images is limited, especially when blood vessels overlap or are interfered with by surrounding tissues. It is difficult for doctors to quickly and accurately identify the cephalic vein from complex images, affecting the timeliness and accuracy of surgical decisions.

[0005] Furthermore, during surgery, the lack of real-time and precise guidance for cephalic vein localization makes it easy for surgeons to accidentally injure surrounding nerves, arteries, and other vital tissues during puncture and anastomosis, leading to a series of complications such as local hematoma, infection, and limb numbness, severely impacting postoperative recovery and quality of life. Inaccurate procedures can also cause insufficient blood flow, stenosis, or occlusion of the arteriovenous fistula, reducing its lifespan and affecting dialysis outcomes. Therefore, accurately locating the cephalic vein has become a pressing clinical challenge. Summary of the Invention

[0006] To address the technical problem of accurately locating the cephalic vein in existing technologies, this invention provides a method and apparatus for displaying the cephalic vein during arteriovenous fistula formation surgery. The technical solution is as follows:

[0007] On one hand, a method for visualizing the cephal vein during arteriovenous fistula formation is provided, the method comprising:

[0008] It receives the raw radio frequency signal sent by the ultrasound probe in real time and preprocesses the raw radio frequency signal.

[0009] The preprocessed original radio frequency signal is denoised using phase unwrapping algorithm and wavelet transform to generate a first signal. The blood vessel contour information in the first signal is extracted by combining frequency domain features. A three-dimensional blood vessel model is generated based on the blood vessel contour information and the first signal.

[0010] Preoperative imaging data of arteriovenous fistula is acquired, the intravascular pressure gradient distribution is determined by hemodynamic simulation algorithm, the vascular bifurcation point is matched according to the preoperative imaging data of arteriovenous fistula, and the vascular topology map is generated by combining the intravascular pressure gradient distribution and the vascular bifurcation point.

[0011] A multi-scale convolutional neural network was used to fuse features from a 3D vascular model and a vascular topology map to generate a heatmap of the probability distribution of cephalic veins.

[0012] Spatially register the cephalic vein probability distribution heatmap with real-time B-ultrasound images, overlay the spatially registered images onto the terminal interface, and mark the coordinates of the arteriovenous fistula location on the spatially registered images.

[0013] Optionally, the preprocessing of the original radio frequency signal includes:

[0014] Bandpass filtering is applied to the original radio frequency signal to separate the blood vessel wall reflection wave and the blood flow Doppler frequency shift signal;

[0015] Hilbert transform is used to extract the envelope signal of the reflected wave from the blood vessel wall to generate an intensity distribution map of the blood vessel cross section;

[0016] Short-time Fourier transform is performed on the Doppler frequency shift signal to extract the blood flow velocity spectrum and calculate the average flow velocity;

[0017] If a sudden drop in signal amplitude occurs in the vascular cross-sectional intensity distribution map, or if the harmonic components of the blood flow velocity spectrum exceed a preset threshold, a signal reacquisition command is triggered.

[0018] Optionally, the step of generating a three-dimensional vascular model based on vascular contour information combined with a first signal includes:

[0019] The blood vessel contour information is converted into point cloud data, and an initial blood vessel surface mesh is generated based on the point cloud data using a voxel fusion algorithm.

[0020] The Laplacian smoothing algorithm was used to optimize the grid curvature continuity of the initial blood vessel surface mesh, and the blood vessel lumen was filled on the initial blood vessel surface mesh based on surface reconstruction technology to generate an initial three-dimensional blood vessel model.

[0021] Anisotropic diffusion filtering is applied to the initial 3D vascular model to enhance the edge clarity of the vascular branch structure and generate a 3D vascular model.

[0022] If the deviation between the length of the vascular branches in the 3D vascular model and the preoperative image data exceeds the preset tolerance, a local mesh resampling process is triggered to update the 3D vascular model.

[0023] Optionally, determining the intravascular pressure gradient distribution using a hemodynamic simulation algorithm includes:

[0024] The intravascular blood flow velocity field is constructed based on the Navier-Stokes equation, and the intravascular pressure gradient distribution is solved by the finite element method based on the intravascular blood flow velocity field.

[0025] Optionally, the step of using a multi-scale convolutional neural network to fuse features from the three-dimensional vascular model and the vascular topology map to generate a heatmap of the probability distribution of the cephalic vein includes:

[0026] The input layer of a multi-scale convolutional neural network receives a slice sequence of a three-dimensional blood vessel model and an adjacency matrix of a blood vessel topology graph.

[0027] The hidden layers of the multi-scale convolutional neural network use parallel convolutional kernels to extract multi-scale feature maps of blood vessel diameter, curvature, and bifurcation density;

[0028] Multi-scale convolutional neural networks use an attention mechanism to weighted fuse feature maps of different scales and output the probability of the presence of cephalic veins.

[0029] The transfer learning algorithm was used to load the pre-trained blood vessel segmentation model parameters into the multi-scale neural network, and the network weights of the multi-scale neural network were fine-tuned based on intraoperative data.

[0030] Optionally, the multi-scale convolutional neural network weighted and fused feature maps of different scales through an attention mechanism, and outputs the probability of the presence of cephal veins, including:

[0031] Channel attention is calculated on the multi-scale feature map to generate the weight coefficients for each channel;

[0032] A spatial weight mask is generated by focusing the pixel positions of the blood vessel bifurcation region through a spatial attention mechanism.

[0033] The channel weights and spatial weights are multiplied by a Hadamard product to obtain the enhanced fused feature map.

[0034] If the maximum probability value in the fused feature map is lower than the preset confidence level, a manual review prompt signal will be triggered.

[0035] Optionally, the spatial registration of the cephalic vein probability distribution heatmap with the real-time ultrasound image includes:

[0036] Extracting vascular skeleton lines from real-time ultrasound images;

[0037] The ICP algorithm is used to match the projection contour of the 3D vascular model, and the projection contour and vascular skeleton line are matched to achieve the initial alignment between the B-ultrasound image and the 3D vascular model.

[0038] An affine transformation matrix was used to align the coordinate systems of the cephalic vein probability distribution heatmap and the real-time ultrasound image;

[0039] The overlapping area between the cephalic vein probability distribution heatmap and the real-time ultrasound image is rendered using a transparency overlay algorithm, and color coding is used to distinguish the probability levels.

[0040] If the spatial registration error between the cephalic vein probability distribution heatmap and the real-time ultrasound image exceeds a preset pixel threshold, then the local elastic registration process based on feature points will be initiated.

[0041] On the other hand, an apparatus for displaying cephal veins in arteriovenous fistula (AVF) reconstruction is provided. This apparatus is used to implement the method for displaying cephal veins in AVF reconstruction provided in the embodiments of the present invention. The apparatus includes:

[0042] The preprocessing module is used to receive the raw radio frequency signal sent by the ultrasound probe in real time and preprocess the raw radio frequency signal.

[0043] The model generation module is used to perform noise reduction on the preprocessed original radio frequency signal based on phase unwrapping algorithm and wavelet transform to generate a first signal, extract blood vessel contour information from the first signal by combining frequency domain features, and generate a three-dimensional blood vessel model based on the blood vessel contour information and the first signal.

[0044] The vascular topology mapping module is used to acquire preoperative image data of arteriovenous fistula, determine the intravascular pressure gradient distribution through hemodynamic simulation algorithm, match vascular bifurcation points based on preoperative image data of arteriovenous fistula, and generate a vascular topology mapping by combining intravascular pressure gradient distribution and vascular bifurcation points.

[0045] The cephalic vein probability distribution heatmap generation module is used to fuse features of a 3D blood vessel model and a blood vessel topology map using a multi-scale convolutional neural network to generate a cephalic vein probability distribution heatmap.

[0046] The display module is used to spatially register the cephalic vein probability distribution heat map with the real-time B-ultrasound image, overlay the spatially registered image on the terminal interface, and mark the coordinates of the arteriovenous fistula location on the spatially registered image.

[0047] On the other hand, a device for displaying cephal veins in arteriovenous fistula formation is provided, the device for displaying cephal veins in arteriovenous fistula formation includes:

[0048] processor;

[0049] A memory storing computer-readable instructions, which, when executed by the processor, implement the method provided in the embodiments of the present invention.

[0050] On the other hand, a computer-readable storage medium is provided, wherein program code is stored in the computer-readable storage medium, and the program code can be invoked by a processor to execute the method provided in the embodiments of the present invention.

[0051] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0052] This invention receives raw radio frequency (RF) signals from an ultrasound probe in real time, preprocesses these signals, and uses a phase unwrapping algorithm and wavelet transform to denoise the preprocessed RF signals, generating a first signal. The vascular contour information in the first signal is extracted using frequency domain features. A three-dimensional vascular model is generated based on the vascular contour information and the first signal. Preoperative image data for arteriovenous fistula (AVF) surgery is acquired. The intravascular pressure gradient distribution is determined using a hemodynamic simulation algorithm. Vascular bifurcation points are matched based on the preoperative AVF image data. A vascular topology map is generated by combining the intravascular pressure gradient distribution and the vascular bifurcation points. A multi-scale convolutional neural network is used to fuse the features of the three-dimensional vascular model and the vascular topology map, generating a cephalic vein probability distribution heatmap. The cephalic vein probability distribution heatmap is spatially registered with the real-time ultrasound image. The spatially registered image is then overlaid and displayed on the terminal interface, and the AVF location coordinates are marked on the spatially registered image. This achieves precise localization of the cephalic vein. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This is a flowchart of a method for displaying the cephal vein in arteriovenous fistula formation according to an embodiment of the present invention;

[0055] Figure 2 This is a block diagram of a device for displaying the cephal vein during arteriovenous fistula formation, provided by an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the structure of a device for displaying the cephal vein in arteriovenous fistula formation surgery, provided by an embodiment of the present invention. Detailed Implementation

[0057] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0058] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0059] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0060] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0061] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0062] To address the technical problem of accurately locating the cephalic vein in existing technologies, this invention provides a method and apparatus for displaying the cephalic vein during arteriovenous fistula formation surgery. The technical solution is as follows:

[0063] like Figure 1 As shown, on one hand, a method for visualizing the cephal vein during arteriovenous fistula formation is provided, the method comprising:

[0064] S1. Receive the raw radio frequency signal sent by the ultrasound probe in real time and preprocess the raw radio frequency signal.

[0065] These raw radio frequency signals contain rich vascular information, but also contain noise and interference, thus requiring preprocessing. By performing bandpass filtering on the raw radio frequency signals, the reflected waves from the vessel wall and the Doppler frequency shift signal of the blood flow are separated, providing a clear signal basis for subsequent analysis.

[0066] S2. Based on the phase unwrapping algorithm and wavelet transform, the preprocessed original radio frequency signal is denoised to generate a first signal. The blood vessel contour information in the first signal is extracted by combining the frequency domain features. Based on the blood vessel contour information and the first signal, a three-dimensional blood vessel model is generated.

[0067] After noise reduction, vascular contour information is extracted by combining frequency domain features. This contour information is further used to generate a three-dimensional vascular model, which intuitively displays the morphology and distribution of blood vessels in a three-dimensional form, providing a basic framework for subsequent analysis and processing.

[0068] S3. Obtain preoperative imaging data of arteriovenous fistula, determine the intravascular pressure gradient distribution through hemodynamic simulation algorithm, match the vascular bifurcation points based on the preoperative imaging data of arteriovenous fistula, and generate a vascular topology map by combining the intravascular pressure gradient distribution and vascular bifurcation points.

[0069] The vascular topology diagram shows in detail the connections and branches between blood vessels, helping doctors understand the overall structure of blood vessels and providing an important basis for the localization of arteriovenous fistulas.

[0070] S4. A multi-scale convolutional neural network is used to fuse features of the three-dimensional blood vessel model and the blood vessel topology map to generate a heatmap of the probability distribution of the cephalic vein.

[0071] A multi-scale convolutional neural network is employed to fuse features from a 3D vascular model and a vascular topology map. This fusion method enables a more comprehensive analysis of vascular features, outputting a cephalic vein probability distribution heatmap. The cephalic vein probability distribution heatmap visually displays the probability of cephalic veins appearing at different locations using different colors and brightness levels, providing doctors with intuitive clues about the possible locations of cephalic veins.

[0072] S5. Spatially register the cephalic vein probability distribution heatmap with the real-time B-ultrasound image, overlay the spatially registered image onto the terminal interface, and mark the coordinates of the arteriovenous fistula location on the spatially registered image.

[0073] Doctors can use these coordinates to quickly locate the fistula, improving the accuracy and efficiency of the surgery.

[0074] Optionally, the preprocessing of the original radio frequency signal includes:

[0075] The original radio frequency signal is bandpass filtered to separate the blood vessel wall reflection wave and the blood flow Doppler frequency shift signal.

[0076] Hilbert transform is used to extract the envelope signal of the reflected wave from the blood vessel wall, and an intensity distribution map of the blood vessel cross section is generated.

[0077] Short-time Fourier transform is performed on the Doppler frequency shift signal to extract the blood flow velocity spectrum and calculate the average flow velocity.

[0078] If a sudden drop in signal amplitude occurs in the vascular cross-sectional intensity distribution map, or if the harmonic components of the blood flow velocity spectrum exceed a preset threshold, a signal reacquisition command is triggered.

[0079] Optionally, the step of generating a three-dimensional vascular model based on vascular contour information combined with a first signal includes:

[0080] The blood vessel contour information is converted into point cloud data, and an initial blood vessel surface mesh is generated based on the point cloud data using a voxel fusion algorithm.

[0081] The Laplacian smoothing algorithm was used to optimize the grid curvature continuity of the initial blood vessel surface mesh, and the blood vessel lumen was filled on the initial blood vessel surface mesh based on surface reconstruction technology to generate an initial three-dimensional blood vessel model.

[0082] Anisotropic diffusion filtering is applied to the initial 3D vascular model to enhance the edge clarity of the vascular branch structure and generate a 3D vascular model.

[0083] If the deviation between the length of the vascular branches in the 3D vascular model and the preoperative image data exceeds the preset tolerance, a local mesh resampling process is triggered to update the 3D vascular model.

[0084] Optionally, determining the intravascular pressure gradient distribution using a hemodynamic simulation algorithm includes:

[0085] The intravascular blood flow velocity field is constructed based on the Navier-Stokes equation, and the intravascular pressure gradient distribution is solved by the finite element method based on the intravascular blood flow velocity field.

[0086] Optionally, the step of using a multi-scale convolutional neural network to fuse features from the three-dimensional vascular model and the vascular topology map to generate a heatmap of the probability distribution of the cephalic vein includes:

[0087] The input layer of a multi-scale convolutional neural network receives slice sequences of a three-dimensional blood vessel model and an adjacency matrix of the blood vessel topology graph.

[0088] The hidden layers of the multi-scale convolutional neural network use parallel convolutional kernels to extract multi-scale feature maps of blood vessel diameter, curvature, and bifurcation density.

[0089] Multi-scale convolutional neural networks use an attention mechanism to weighted fuse feature maps of different scales and output the probability of the presence of cephalic veins.

[0090] The transfer learning algorithm was used to load the pre-trained blood vessel segmentation model parameters into the multi-scale neural network, and the network weights of the multi-scale neural network were fine-tuned based on intraoperative data.

[0091] Optionally, the multi-scale convolutional neural network weighted and fused feature maps of different scales through an attention mechanism, and outputs the probability of the presence of cephal veins, including:

[0092] Channel attention is calculated on the multi-scale feature map to generate the weight coefficients for each channel.

[0093] A spatial weight mask is generated by focusing the pixel positions of the blood vessel bifurcation region through a spatial attention mechanism.

[0094] The channel weights and spatial weights are subjected to the Hadamard product to obtain the enhanced fused feature map.

[0095] If the maximum probability value in the fused feature map is lower than the preset confidence level, a manual review prompt signal will be triggered.

[0096] Optionally, the spatial registration of the cephalic vein probability distribution heatmap with the real-time ultrasound image includes:

[0097] Extract the vascular skeleton line from real-time ultrasound images.

[0098] The ICP algorithm is used to match the projection contour of the 3D vascular model, and the projection contour and vascular skeleton line are matched to achieve the initial alignment between the ultrasound image and the 3D vascular model.

[0099] An affine transformation matrix was used to align the coordinate system of the cephalic vein probability distribution heatmap with that of the real-time ultrasound image.

[0100] The overlapping area between the cephalic vein probability distribution heatmap and the real-time ultrasound image is rendered using a transparency overlay algorithm, and color coding is used to distinguish the probability levels.

[0101] If the spatial registration error between the cephalic vein probability distribution heatmap and the real-time ultrasound image exceeds a preset pixel threshold, then the local elastic registration process based on feature points will be initiated.

[0102] The present invention will be further described below with reference to Examples 1 to 6:

[0103] Example 1:

[0104] Once the ultrasound probe acquires the raw radiofrequency signal, the bandpass filter is activated first. The purpose of the bandpass filter is to set an appropriate filtering frequency range based on the frequency characteristics of the blood vessel wall reflection wave and the blood flow Doppler frequency shift signal, thereby effectively separating the two. For example, if experiments and experience determine that the frequency range of the blood vessel wall reflection wave is mainly concentrated in [X1,X2] Hz, while the frequency range of the blood flow Doppler frequency shift signal is in [Y1,Y2] Hz, then the bandpass filter can be set to allow signals within the [X1,X2] Hz and [Y1,Y2] Hz ranges to pass through, while blocking noise signals of other frequencies.

[0105] After separating the reflected wave from the blood vessel wall, the envelope signal is extracted using Hilbert transform. Hilbert transform is a commonly used signal processing method that converts the amplitude and phase information of the original signal to extract the envelope signal. The resulting cross-sectional intensity distribution map of the blood vessel clearly shows the signal intensity distribution at different locations along the vessel cross-section, allowing doctors to visually observe changes in the morphology and structure of the blood vessel wall.

[0106] For Doppler shift signals, a short-time Fourier transform (SFT) is used for processing. The SFT performs localized analysis of the signal in the time and frequency domains, enabling the extraction of blood flow velocity spectra at different times. The average flow velocity is then calculated by analyzing the blood flow velocity spectrum. The formula for calculating the average flow velocity is: Where V avg V represents the average flow velocity, n is the number of sampling points, and V i Let be the flow rate at the i-th sampling point.

[0107] To ensure the quality of the acquired signals, a sudden drop in signal amplitude in the vascular cross-sectional intensity distribution map may indicate an abnormality in the vessel wall or signal interference. Similarly, when the harmonic components of the blood flow velocity spectrum exceed a preset threshold, it also indicates a signal abnormality. Upon the occurrence of these abnormalities, a signal re-acquisition command will be automatically triggered to reacquire the original radio frequency signal, ensuring the accuracy of subsequent analysis and processing.

[0108] Example 2:

[0109] After acquiring the blood vessel contour information, it is first converted into point cloud data. Point cloud data is a dataset composed of a large number of discrete points, each representing a location on the blood vessel contour. Using a voxel fusion algorithm, this point cloud data is converted into an initial blood vessel surface mesh. The voxel fusion algorithm combines adjacent points into small voxel units based on the spatial relationships of the point clouds; these voxel units are then connected to form the initial blood vessel surface mesh.

[0110] To make the generated mesh smoother and more natural, the Laplacian smoothing algorithm was used to optimize the mesh curvature continuity. After optimizing the mesh curvature, the vascular lumen was filled using surface reconstruction technology. Surface reconstruction technology utilizes the surface information of the mesh to construct the surface of the vascular lumen, thus fully presenting the three-dimensional structure of the blood vessel.

[0111] To enhance the edge sharpness of vascular branch structures, anisotropic diffusion filtering was applied to the reconstructed 3D model. Anisotropic diffusion filtering is an image gradient-based filtering method that smooths the image while preserving its edge information. In the vascular model, it makes the edges of vascular branches clearer, facilitating observation and analysis by doctors.

[0112] To ensure the accuracy of the 3D model, the system compares the length of the blood vessel branches with preoperative image data. If the deviation exceeds a preset tolerance, it indicates that the model may have errors, triggering a local mesh resampling process. Local mesh resampling resamples and regenerates the mesh in areas with large deviations to improve the model's accuracy and make it more consistent with the actual blood vessel conditions.

[0113] Example 3:

[0114] Hemodynamic simulation algorithms are crucial for accurately determining the location of arteriovenous fistulas (AFs). First, the intravascular blood flow velocity field is constructed based on the Navier-Stokes equations. The Navier-Stokes equations describe the conservation of momentum in viscous incompressible fluids. Solving these equations using the finite element method (FEM) yields the intravascular pressure gradient distribution. The FEM divides the vascular region into multiple small elements, discretizing the Navier-Stokes equations within each element, and then solving the equation set yields the pressure gradient distribution.

[0115] During the simulation, the bifurcation angle and diameter change rate of the vessel in the preoperative images of the arteriovenous fistula were extracted and used as boundary conditions input into the simulation model. The bifurcation angle and diameter change rate affect the blood flow state, and taking these factors into account can make the simulation results more accurate. For example, when the bifurcation angle is large, the blood flow at the bifurcation will be more complex, and the pressure distribution will also be different.

[0116] The vorticity intensity at the bifurcation point is calculated using the following formula: Where ω represents the vorticity intensity. This is a velocity vector. If the vorticity value exceeds a preset threshold, it is marked as a potential arteriovenous fistula anastomosis area. This is because in areas with larger vorticity, the blood flow pattern is more complex, which may make arteriovenous fistula anastomosis more suitable.

[0117] Finally, the simulation results were non-rigidly registered with the intraoperative ultrasound images. Non-rigid registration can better adapt to the deformation of blood vessels during the operation, making the simulation results more consistent with the actual images, thereby outputting the corrected arteriovenous fistula location coordinates and providing doctors with more accurate fistula localization information.

[0118] Example 4:

[0119] In practice, the training process of a multi-scale convolutional neural network includes:

[0120] At the start of training, the input layer receives a sequence of slices from a 3D vascular model and an adjacency matrix of the vascular topology graph. The slice sequence of the 3D vascular model is a series of 2D images obtained by cutting the complete 3D model along a specific direction. These images preserve the morphological and structural information of the blood vessels at different levels. The adjacency matrix of the vascular topology graph describes the connections between blood vessels in the form of a mathematical matrix. The elements in the matrix correspond to whether there is a connection between different vascular nodes and the nature of the connection. In this way, the neural network can "understand" the overall architecture of the blood vessels.

[0121] After receiving the input data, the hidden layer uses parallel convolutional kernels for feature extraction. Different sized convolutional kernels serve different purposes: large kernels capture large-scale features of blood vessels, such as their general direction and overall curvature; small kernels focus on extracting minute details, such as subtle protrusions on the vessel wall and the precise shape of vessel bifurcation. By using these convolutional kernels in parallel, the network can comprehensively acquire multi-scale features such as vessel diameter, curvature, and bifurcation density, providing rich information for accurately determining the location of the cephalic vein.

[0122] Attention mechanisms play a crucial role in feature fusion. During channel attention calculation, they analyze the importance of each channel's feature map and generate corresponding weight coefficients for each channel. For example, channels closely related to cephalic vein features, such as those reflecting vessel wall thickness and blood flow signal characteristics, are assigned higher weights; while less relevant channels have lower weights.

[0123] Spatial attention focuses on the pixel locations of blood vessel bifurcation regions, generating a spatial weight mask. Blood vessel bifurcation regions are often key areas for determining the location of the cephalic vein because the cephalic vein exhibits unique characteristics at bifurcation points. Spatial attention can highlight pixels in these key regions, making the neural network pay more attention to these potentially cephalic vein locations.

[0124] The channel weights and spatial weights are multiplied by a Hadamard product to obtain the enhanced fused feature map. This operation preserves important feature information from different channels while enhancing features at key spatial locations, making the fused feature map more accurately reflect the characteristics of the cephalic vein and thus outputting a more reliable probability of the presence of the cephalic vein.

[0125] If the maximum probability value in the fused feature map is lower than the preset confidence level, it means that the current prediction result has low credibility and significant uncertainty. To ensure the accuracy and safety of the surgery, a manual review prompt signal will be triggered at this time. For example, a prominent prompt box may pop up on the surgical operation interface, or a specific sound alarm may be issued to remind the doctor to manually check and confirm the prediction result, so as to avoid the surgical operation being affected by incorrect predictions.

[0126] Example 5:

[0127] Spatial registration algorithms are the core component for achieving accurate overlay display of cephalic vein probability distribution heatmaps and real-time ultrasound images. During surgery, real-time ultrasound images continuously acquire real-time information about the patient's blood vessels. To accurately match the cephalic vein probability distribution heatmap with this information, the vascular skeleton lines in the real-time ultrasound images must first be extracted. These vascular skeleton lines, obtained through specific image processing algorithms, represent the central axis of the blood vessels and reflect their approximate course.

[0128] After extracting the vascular skeleton lines, the Iterative Closest Point (ICP) algorithm is used to match the projected contour of the 3D vascular model. The ICP algorithm works by iteratively finding the optimal matching relationship between two sets of points. In this process, the 3D vascular model is projected onto the same plane as the ultrasound image to obtain the projected contour, which is then matched with the vascular skeleton lines in the ultrasound image. During each iteration, the algorithm calculates the distance between the two sets of points and adjusts the position and orientation of the 3D vascular model to gradually reduce the distance until an optimal matching state is reached, achieving initial alignment.

[0129] After initial alignment, an affine transformation matrix is ​​used to align the coordinate systems of the cephalic vein probability distribution heatmap and the ultrasound image. An affine transformation matrix is ​​a mathematical matrix that incorporates transformations such as translation, rotation, and scaling. By adjusting the parameters of the affine transformation matrix, the cephalic vein probability distribution heatmap can be transformed accordingly to perfectly match the ultrasound image in its coordinate system. For example, if the ultrasound image undergoes some rotation or translation during acquisition, the affine transformation matrix can perform the reverse rotation and translation operation on the cephalic vein probability distribution heatmap, achieving coordinate system unification between the two.

[0130] To make the overlay display clearer and more intuitive, a transparency overlay algorithm is used to render the overlapping area of ​​the heatmap and ultrasound image, and color coding is employed to distinguish probability levels. The transparency overlay algorithm allows the overlapping areas of the heatmap and ultrasound image to display their respective information without obscuring each other. For example, setting the transparency of the heatmap to 50% allows doctors to see both the actual morphology of blood vessels in the ultrasound image and the probability distribution of the cephalic vein. Simultaneously, color coding is used to distinguish probability levels; for instance, red represents high-probability areas, meaning these areas are highly likely to be the location of the cephalic vein, while blue represents low-probability areas, facilitating doctors' quick identification of possible cephalic vein locations.

[0131] If the registration error exceeds a preset pixel threshold, it indicates that the current registration effect is unsatisfactory and may affect the doctor's judgment. In this case, a feature-point-based local elastic registration process can be initiated. Feature-point-based local elastic registration selects certain feature points in the image; these feature points are typically locations with unique characteristics, such as vascular bifurcation points or prominent protrusions on the vessel wall. Based on the positional differences of these feature points in the ultrasound image and the 3D vascular model, local adjustments are made to the image. For example, if the positional deviation of a certain feature point in the ultrasound image and the thermal image is large, elastic transformations such as stretching and twisting are applied to the area around that feature point. This allows the cephalic vein probability distribution thermal image and the ultrasound image to better match in these key areas, thereby improving the registration accuracy and ensuring precise alignment between the cephalic vein probability distribution thermal image and the real-time ultrasound image, providing the doctor with accurate display information.

[0132] Example 6:

[0133] During the surgery, real-time data is collected on the surgeon's operating path and instrument positioning. The surgeon's operating path records the movement trajectory of the instruments used during the surgery, while the instrument positioning data accurately identifies the position of the instruments within the patient's body. This data reflects the actual situation of the surgical procedure and allows for continuous model optimization.

[0134] The weights of the last layer of a multi-scale convolutional neural network are updated using stochastic gradient descent. Stochastic gradient descent is a commonly used optimization algorithm that randomly selects a sample or a small batch of samples from the collected data for gradient calculation each time, and then updates the network weights based on the gradient. The advantages of this method are its fast computation speed and rapid convergence on large-scale data. When updating the weights, the weight values ​​are adjusted in the direction that reduces the loss function, based on the gradient of the loss function with respect to the weights, allowing the model to continuously learn and adapt to the data from actual surgeries.

[0135] If the rate of decrease of the loss function is lower than the preset convergence threshold for N consecutive iterations, it means that the multi-scale convolutional neural network may have converged to a local optimum, and continuing the current training may not further improve the model's performance. To avoid the multi-scale convolutional neural network getting stuck in a local optimum, the network parameters are frozen and no longer updated, while a backup multi-scale convolutional neural network model is activated. The backup multi-scale convolutional neural network model is a pre-trained standby model that can take over the work of the original multi-scale convolutional neural network in a timely manner when the original multi-scale convolutional neural network encounters problems or performs poorly, ensuring the smooth progress of the transformation.

[0136] As the Euclidean distance between the instrument positioning data and the predicted cephalic vein location continues to decrease, it indicates that the surgical procedure is getting closer to the predicted cephalic vein location. At this point, the transparency of the heatmap display area is dynamically reduced. For example, when the Euclidean distance decreases to a certain extent, the transparency of the heatmap is reduced from 50% to 30%. This allows the surgeon to observe the actual surgical area more clearly and intuitively perceive the degree of closeness between the predicted result and the actual situation. This helps the surgeon to perform the surgical procedure more accurately and improves the success rate of the operation.

[0137] On the other hand, such as Figure 2 As shown, a device for displaying the cephal vein in arteriovenous fistula (AVF) surgery is provided. This device is used to implement the method for displaying the cephal vein in AVF surgery provided in this embodiment of the invention. The device includes:

[0138] The preprocessing module 201 is used to receive the raw radio frequency signal sent by the ultrasound probe in real time and preprocess the raw radio frequency signal.

[0139] The model generation module 202 is used to perform noise reduction processing on the preprocessed original radio frequency signal based on the phase unwrapping algorithm and wavelet transform to generate a first signal, extract blood vessel contour information from the first signal by combining frequency domain features, and generate a three-dimensional blood vessel model based on the blood vessel contour information and the first signal.

[0140] The vascular topology mapping module 203 is used to acquire preoperative image data of arteriovenous fistula, determine the intravascular pressure gradient distribution through hemodynamic simulation algorithm, match vascular bifurcation points based on preoperative image data of arteriovenous fistula, and generate a vascular topology mapping by combining intravascular pressure gradient distribution and vascular bifurcation points.

[0141] The cephalic vein probability distribution heatmap generation module 204 is used to perform feature fusion on the three-dimensional blood vessel model and blood vessel topology map using a multi-scale convolutional neural network to generate a cephalic vein probability distribution heatmap.

[0142] The display module 205 is used to spatially register the cephalic vein probability distribution heat map with the real-time B-ultrasound image, overlay the spatially registered image on the terminal interface, and mark the coordinates of the arteriovenous fistula location on the spatially registered image.

[0143] On the other hand, a device for displaying cephal veins in arteriovenous fistula formation is provided, the device for displaying cephal veins in arteriovenous fistula formation includes:

[0144] processor.

[0145] A memory storing computer-readable instructions, which, when executed by the processor, implement the method provided in the embodiments of the present invention.

[0146] On the other hand, a computer-readable storage medium is provided, wherein program code is stored in the computer-readable storage medium, and the program code can be invoked by a processor to execute the method provided in the embodiments of the present invention.

[0147] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0148] The cephalic vein visualization method for arteriovenous fistula formation according to this invention has significant beneficial effects in improving surgical precision, optimizing surgical procedures, and improving patient prognosis. In terms of improving surgical precision, it integrates multiple advanced algorithms and technologies. Through operations such as bandpass filtering, Hilbert transform, and short-time Fourier transform, it can accurately separate the vessel wall reflection wave and blood flow Doppler frequency shift signal from the original radio frequency signal, generating a vessel cross-sectional intensity distribution map and blood flow velocity spectrum, providing reliable data for subsequent analysis. If the signal is abnormal, such as a sudden drop in signal amplitude in the vessel cross-sectional intensity distribution map or harmonic components of the blood flow velocity spectrum exceeding a preset threshold, a re-acquisition command is automatically triggered to ensure data accuracy and avoid cephalic vein positioning deviations caused by signal errors.

[0149] Simultaneously, noise reduction is achieved using phase unwrapping algorithms and wavelet transforms, and vascular contour information is extracted by combining frequency domain features. The resulting 3D vascular model realistically recreates the spatial morphology of blood vessels. During the 3D modeling process, the application of voxel fusion algorithms, Laplacian smoothing algorithms, and surface reconstruction techniques optimizes the model's mesh quality and structural integrity. If the deviation between the vascular branch length and preoperative image data exceeds a preset tolerance, a local mesh resampling process is triggered to further ensure the model's accuracy, allowing doctors to more intuitively and accurately observe the distribution of blood vessels and more accurately locate the cephalic vein.

[0150] The blood flow velocity field was constructed based on the Navier-Stokes equations, and the pressure gradient distribution was solved using the finite element method. The bifurcation angle and diameter change rate from preoperative arteriovenous fistula (AVF) imaging data were used as boundary conditions to accurately calculate the vorticity intensity at the bifurcation point and mark potential AVF anastomosis areas. The simulation results were then non-rigidly registered with intraoperative ultrasound images to output the corrected AVF location coordinates, significantly improving the accuracy of AVF localization and reducing the time and error associated with AVF location during surgery.

[0151] A multi-scale convolutional neural network is employed to fuse features from a 3D vascular model and a vascular topology map. The parallel convolutional kernels in its hidden layers extract multi-scale features of vessel diameter, curvature, and bifurcation density. An attention mechanism is used to weight and fuse feature maps of different scales, resulting in a more accurate cephalic vein probability distribution heatmap. If the maximum probability value in the fused feature map is lower than a preset confidence level, a manual review prompt is triggered, further ensuring the reliability of the prediction results. This allows surgeons to anticipate the potential location of the cephalic vein and plan the surgical path accordingly.

[0152] The cephalic vein probability distribution heatmap is spatially registered and overlaid with real-time ultrasound images, marking the coordinates of the arteriovenous fistula location. The spatial registration algorithm employs operations such as extracting vascular skeleton lines, ICP algorithm matching, affine transformation matrix alignment of the coordinate system, and transparency overlay algorithms to accurately fuse the cephalic vein probability distribution heatmap with the ultrasound image. If the registration error exceeds a preset pixel threshold, a feature-point-based local elastic registration process is initiated, ensuring that surgeons can clearly and in real-time see the correspondence between the cephalic vein probability distribution and the actual vascular image during surgery. This allows for intuitive acquisition of the cephalic vein's location information, leading to more accurate surgical procedures and reduced surgical risks.

[0153] From the perspective of improving patient prognosis, the technical solution of this invention reduces surgical errors caused by inaccurate cephalic vein localization. It lowers the risk of accidental injury to peripheral nerves, arteries, and other important tissues, and reduces the probability of complications such as local hematoma, infection, and limb numbness. Precise cephalic vein localization improves the quality of arteriovenous fistulas (AVFs), reduces the occurrence of problems such as insufficient blood flow, stenosis, or occlusion, and extends the lifespan of the AVF, thereby improving the patient's dialysis effectiveness and quality of life, and has a positive impact on the patient's long-term health and rehabilitation.

[0154] Figure 3 This is a schematic diagram of a device for displaying the cephal vein during arteriovenous fistula formation, provided by an embodiment of the present invention. Figure 3 As shown, optionally, the 310 for cephal vein visualization in arteriovenous fistula formation may include a first processor 2001.

[0155] Optionally, the device 310 for displaying the cephal vein in arteriovenous fistula formation may also include a memory 2002 and a transceiver 2003.

[0156] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0157] The following is combined with Figure 3 A detailed description of the various components of the device 310 used for cephal vein visualization in arteriovenous fistula formation is provided below:

[0158] The first processor 2001 is the control center of the device 310 for displaying the cephal vein in arteriovenous fistula formation surgery. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0159] Optionally, the first processor 2001 can perform various functions of the device 310 for cephal vein display in arteriovenous fistula formation by running or executing software programs stored in memory 2002 and calling data stored in memory 2002.

[0160] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 3 CPU0 and CPU1 are shown in the diagram.

[0161] In a specific implementation, as one embodiment, the device 310 for displaying the cephal vein in arteriovenous fistula formation may also include multiple processors, such as... Figure 3 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, "processor" can refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0162] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0163] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the device 310 for displaying the cephal vein in arteriovenous fistula reconstruction. Figure 3 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0164] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0165] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 3 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0166] Alternatively, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected via the interface circuit of the device 310 for displaying the cephal vein in arteriovenous fistula angioplasty. Figure 3 (Not shown in the figure) is coupled to the first processor 2001, and the embodiments of the present invention do not specifically limit this.

[0167] It should be noted that, Figure 3 The structure of the device 310 for cephal vein visualization in arteriovenous fistula formation shown in the figure does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0168] Furthermore, the technical effects of the device 310 for displaying the cephal vein in arteriovenous fistula formation can be referred to the technical effects of the method for displaying the cephal vein in arteriovenous fistula formation described in the above method embodiments, and will not be repeated here.

[0169] It should be understood that the first processor 2001 in this embodiment of the invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or 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.

[0170] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0171] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via infrared, microwave, or other means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0172] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0173] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0174] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0175] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0176] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0177] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0178] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0179] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0180] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0181] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for visualizing the cephal vein during arteriovenous fistula formation, characterized in that, The method includes: It receives the raw radio frequency signal sent by the ultrasound probe in real time and preprocesses the raw radio frequency signal. The preprocessed original radio frequency signal is denoised using phase unwrapping algorithm and wavelet transform to generate a first signal. The blood vessel contour information in the first signal is extracted by combining frequency domain features. A three-dimensional blood vessel model is generated based on the blood vessel contour information and the first signal. Preoperative imaging data of arteriovenous fistula is acquired, the intravascular pressure gradient distribution is determined by hemodynamic simulation algorithm, the vascular bifurcation point is matched according to the preoperative imaging data of arteriovenous fistula, and the vascular topology map is generated by combining the intravascular pressure gradient distribution and the vascular bifurcation point. A multi-scale convolutional neural network was used to fuse features from a 3D vascular model and a vascular topology map to generate a heatmap of the probability distribution of cephalic veins. Spatial registration is performed between the cephalic vein probability distribution heat map and real-time B-ultrasound image. The spatially registered image is then overlaid and displayed on the terminal interface, and the coordinates of the arteriovenous fistula location are marked on the spatially registered image. The step of using a multi-scale convolutional neural network to fuse features from a 3D vascular model and a vascular topology map to generate a heatmap of cephalic vein probability distribution includes: The input layer of a multi-scale convolutional neural network receives a slice sequence of a three-dimensional blood vessel model and an adjacency matrix of a blood vessel topology graph. The hidden layers of the multi-scale convolutional neural network use parallel convolutional kernels to extract multi-scale feature maps of blood vessel diameter, curvature, and bifurcation density; Multi-scale convolutional neural networks use an attention mechanism to weighted fuse feature maps of different scales and output the probability of the presence of cephalic veins. The transfer learning algorithm was used to load the pre-trained blood vessel segmentation model parameters into the multi-scale neural network, and the network weights of the multi-scale neural network were fine-tuned based on intraoperative data.

2. The method for visualizing the cephal vein in arteriovenous fistula formation according to claim 1, characterized in that, The preprocessing of the raw radio frequency signal includes: Bandpass filtering is applied to the original radio frequency signal to separate the blood vessel wall reflection wave and the blood flow Doppler frequency shift signal; Hilbert transform is used to extract the envelope signal of the reflected wave from the blood vessel wall to generate an intensity distribution map of the blood vessel cross section; Short-time Fourier transform is performed on the Doppler frequency shift signal to extract the blood flow velocity spectrum and calculate the average flow velocity; If a sudden drop in signal amplitude occurs in the vascular cross-sectional intensity distribution map, or if the harmonic components of the blood flow velocity spectrum exceed a preset threshold, a signal reacquisition command is triggered.

3. The method for visualizing the cephal vein during arteriovenous fistula formation according to claim 1, characterized in that, The process of generating a three-dimensional blood vessel model based on blood vessel contour information combined with a first signal includes: The blood vessel contour information is converted into point cloud data, and an initial blood vessel surface mesh is generated based on the point cloud data using a voxel fusion algorithm. The Laplacian smoothing algorithm was used to optimize the grid curvature continuity of the initial blood vessel surface mesh, and the blood vessel lumen was filled on the initial blood vessel surface mesh based on surface reconstruction technology to generate an initial three-dimensional blood vessel model. Anisotropic diffusion filtering is applied to the initial 3D vascular model to enhance the edge clarity of the vascular branch structure and generate a 3D vascular model. If the deviation between the length of the vascular branches in the 3D vascular model and the preoperative image data exceeds the preset tolerance, a local mesh resampling process is triggered to update the 3D vascular model.

4. The method for visualizing the cephal vein in arteriovenous fistula formation according to claim 1, characterized in that, The determination of intravascular pressure gradient distribution using hemodynamic simulation algorithms includes: The intravascular blood flow velocity field is constructed based on the Navier-Stokes equation, and the intravascular pressure gradient distribution is solved by the finite element method based on the intravascular blood flow velocity field.

5. The method for visualizing the cephal vein during arteriovenous fistula formation according to claim 1, characterized in that, The multi-scale convolutional neural network uses an attention mechanism to weighted fuse feature maps of different scales, and outputs the probability of cephal vein presence, including: Channel attention is calculated on the multi-scale feature map to generate the weight coefficients for each channel; A spatial weight mask is generated by focusing the pixel positions of the blood vessel bifurcation region through a spatial attention mechanism. The channel weights and spatial weights are multiplied by a Hadamard product to obtain the enhanced fused feature map. If the maximum probability value in the fused feature map is lower than the preset confidence level, a manual review prompt signal will be triggered.

6. The method for displaying the cephal vein in arteriovenous fistula formation according to claim 1, characterized in that, The spatial registration of the cephalic vein probability distribution heatmap with real-time ultrasound images includes: Extracting vascular skeleton lines from real-time ultrasound images; The ICP algorithm is used to match the projection contour of the 3D vascular model, and the projection contour and vascular skeleton line are matched to achieve the initial alignment between the B-ultrasound image and the 3D vascular model. An affine transformation matrix was used to align the coordinate systems of the cephalic vein probability distribution heatmap and the real-time ultrasound image; The overlapping area between the cephalic vein probability distribution heatmap and the real-time ultrasound image is rendered using a transparency overlay algorithm, and color coding is used to distinguish the probability levels. If the spatial registration error between the cephalic vein probability distribution heatmap and the real-time ultrasound image exceeds a preset pixel threshold, then the local elastic registration process based on feature points will be initiated.

7. A device for displaying a cephal vein in arteriovenous fistula (AVF) reconstruction, wherein the device is used to implement the method for displaying a cephal vein in AVF reconstruction as described in any one of claims 1-6, characterized in that... The device includes: The preprocessing module is used to receive the raw radio frequency signal sent by the ultrasound probe in real time and preprocess the raw radio frequency signal. The model generation module is used to perform noise reduction on the preprocessed original radio frequency signal based on phase unwrapping algorithm and wavelet transform to generate a first signal, extract blood vessel contour information from the first signal by combining frequency domain features, and generate a three-dimensional blood vessel model based on the blood vessel contour information and the first signal. The vascular topology mapping module is used to acquire preoperative image data of arteriovenous fistula, determine the intravascular pressure gradient distribution through hemodynamic simulation algorithm, match vascular bifurcation points based on preoperative image data of arteriovenous fistula, and generate vascular topology mapping by combining intravascular pressure gradient distribution and vascular bifurcation points. The cephalic vein probability distribution heatmap generation module is used to fuse features of a 3D blood vessel model and a blood vessel topology map using a multi-scale convolutional neural network to generate a cephalic vein probability distribution heatmap. The display module is used to spatially register the cephalic vein probability distribution heat map with real-time B-ultrasound images, overlay the spatially registered images on the terminal interface, and mark the coordinates of the arteriovenous fistula location on the spatially registered images.

8. A device for displaying the cephal vein during arteriovenous fistula formation surgery, characterized in that, The device for cephal vein visualization during arteriovenous fistula formation includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 6.

Images