Visual localization and ablation method and system for cardiovascular balloon

Abstract

The invention discloses a visual localization and ablation method and system for a cardiovascular balloon. Acquiring a real-time image sequence through a synchronous multi-view image acquisition technology; through a multi-view image reconstruction algorithm, a three-dimensional blood vessel model is obtained through conversion; in the three-dimensional blood vessel model, identifying an opening position and morphological characteristics of a target blood vessel, and extracting to obtain a blood vessel lesion area for positioning an ablation target; the optimal positioning point coordinates of the cardiovascular balloon are obtained through calculation in combination with hemodynamic simulation, and the optimal path is planned; adjusting the position of the cardiovascular balloon using a navigation algorithm; real-time parameters in the ablation process are collected, and the ablation parameters are monitored and adjusted and controlled in real time. The system comprises an image acquisition module, a two-dimensional extraction module, a three-dimensional modeling module, a vasculopathy area extraction module, an optimal path planning module, a navigation adjustment module and a parameter adjustment module. Modern imaging, hemodynamic simulation and electromagnetic positioning are combined, and dynamic control over the ablation process is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of electroablation, and particularly to a visual positioning and ablation method and system for a cardiovascular balloon. Background Art

[0002] Cardiovascular diseases (CVDs) are one of the main causes of death globally, especially diseases such as atherosclerosis, coronary heart disease, and myocardial infarction. With the aging of the global population and the change of lifestyle, the prevalence and mortality of cardiovascular diseases continue to rise. The treatment methods for these diseases mainly include drug treatment, interventional surgery, and surgical surgery, etc. Among them, balloon interventional treatment, for example, balloon dilation, balloon ablation, etc. are important means for treating coronary artery diseases. Such surgeries usually require precise positioning and real-time monitoring to ensure the treatment effect and reduce the surgical risk.

[0003] Currently, traditional balloon positioning techniques mainly rely on imaging guidance, such as angiography, CT, MRI, etc. to determine the position of the balloon. However, these methods have some deficiencies: (1) Poor real-time performance. Traditional imaging methods usually cannot provide real-time and dynamic feedback on the position of the balloon; (2) Limited positioning accuracy. The resolution of traditional imaging techniques is limited, especially in cases where the blood vessels are complex, narrow, or curved, the positioning accuracy is affected; (3) High invasiveness. Some imaging methods such as angiography require the injection of contrast agents, which increases the pain and complication risk of patients. In ablation procedures, especially during intravascular balloon ablation, precise balloon positioning, real-time monitoring, and dynamic control during the ablation process are crucial. Traditional ablation techniques usually rely on manual operation and it is difficult to achieve precise positioning and real-time adjustment of blood vessels.

[0004] With the rapid development of computer vision, artificial intelligence AI, deep learning, and imaging techniques, non-invasive, real-time, and high-precision balloon positioning and ablation process control have gradually become a research hotspot in the medical field. These emerging technologies can greatly improve the accuracy and safety of balloon treatment. However, the existing technologies still have the following problems: (1) Insufficient positioning accuracy and real-time performance. Although multi-view image and electromagnetic positioning techniques have improved the positioning accuracy to a certain extent, there are still certain errors and delays in complex vascular environments; (2) Insufficient real-time feedback control. The real-time monitoring and automatic adjustment mechanisms during the ablation process are not yet perfect. The changes in ablation parameters such as temperature and pressure need to be monitored in real time during the entire treatment process to ensure the ablation effect while avoiding damage to healthy tissues; (3) Lack of intelligent path planning and adjustment. In the existing technologies, the path planning of the balloon from the current position to the optimal positioning point is still manually adjusted, lacking intelligent and automated optimization algorithms. Summary of the Invention

[0005] The object of the present invention is to provide a visual positioning and ablation method and system for a cardiovascular balloon, which is used to solve at least one of the above technical problems and can combine modern imaging technology, artificial intelligence, hemodynamic simulation and electromagnetic positioning technology to realize dynamic control of the ablation process.

[0006] The embodiments of the present invention are implemented as follows:

[0007] A visual positioning and ablation method for a cardiovascular balloon, which includes:

[0008] Adopt a multi-channel imaging system to obtain a real-time image sequence through synchronous multi-view image acquisition technology.

[0009] Use an edge detection algorithm to extract the main contour of the blood vessel from the real-time image sequence to obtain a two-dimensional blood vessel image sequence.

[0010] Convert the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm.

[0011] In the three-dimensional blood vessel model, based on deep learning classification network and image generation adversarial network technology, identify the opening position and morphological features of the target blood vessel, and extract the blood vessel lesion area for positioning the ablation target.

[0012] Combined with hemodynamic simulation, based on the three-dimensional blood vessel model and the blood vessel lesion area, calculate the optimal positioning point coordinates of the cardiovascular balloon, and plan the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates.

[0013] Obtain the real-time position information of the cardiovascular balloon through electromagnetic positioning, calculate the relative spatial relationship between the cardiovascular balloon and the best path, and use a navigation algorithm to adjust the position of the cardiovascular balloon.

[0014] Collect real-time parameters during the ablation process, monitor and adjust the ablation parameters in real time.

[0015] In a preferred embodiment of the present invention, in the above visual positioning and ablation method for a cardiovascular balloon, the step of adopting a multi-channel imaging system to obtain a real-time image sequence through synchronous multi-view image acquisition technology includes:

[0016] Obtain cross-sectional images of blood vessels and surrounding tissues through a CT imaging channel.

[0017] Obtain images of blood vessel wall thickness and blood flow characteristics through an MRI imaging channel.

[0018] Obtain direct-view images of the inside of the cardiovascular system in real time through a flexible endoscope imaging channel.

[0019] Set the acquisition frequency and image resolution for each imaging channel, and synchronously record the images acquired by each imaging channel according to the timestamp through the image sensor, align the multi-channel data, and obtain a real-time image sequence.

[0020] Its technical effect is as follows: Through the multi-channel imaging system, different anatomical features and lesion areas of blood vessels are fully captured. Especially in complex situations such as blood vessel bending, stenosis, or branching, more comprehensive blood vessel anatomical information is provided, giving play to the complementarity of different imaging techniques, making subsequent three-dimensional reconstruction and lesion area recognition more accurate. In the case of complex blood vessel morphology or blood flow fluctuations, the real-time acquired image sequence can provide immediate feedback information, avoiding positioning errors caused by delays.

[0021] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the obtaining of the two-dimensional blood vessel image sequence by extracting the main contour of the blood vessel from the real-time image sequence using an edge detection algorithm includes:

[0022] Based on the Sobel edge detection algorithm, perform two-dimensional directional gradient calculation on the images in the real-time image sequence,

[0023] The horizontal gradient of each pixel point in the X direction

[0024] The vertical gradient of each pixel point in the Y direction

[0025] Where K x (a, b) is the Sobel convolution kernel in the X direction, K y (a, b) is the Sobel convolution kernel in the Y direction, and the Sobel convolution kernel is designed according to the thickness of the blood vessel, and I(x + a, y + b) is the pixel value at the current position in the image.

[0026] Calculate the gradient magnitude of each pixel point For representing the change intensity of the pixel point.

[0027] Calculate the gradient direction of each pixel point θ(x, y) = atan2(G y (x, y), G x (x, y)), for representing the direction of the edge.

[0028] According to the edge intensity and morphological characteristics of the blood vessel, set a high threshold T H and a low threshold T L .

[0029] Pixel points with a gradient magnitude greater than the high threshold are valid edges, and retain the pixel points.

[0030] For pixel points with gradient magnitude less than the high threshold but greater than the low threshold, determine whether the pixel point is connected to a strong edge. If it is connected, the pixel point is considered a valid edge and retained; if not, the pixel point is not a valid edge and is removed.

[0031] Pixel points with gradient magnitude less than the low threshold do not belong to the edge, and such pixel points are removed.

[0032] Extract the retained pixel points to obtain a binary image.

[0033] Use a contour extraction algorithm to extract the contour of the blood vessel from the binary image, obtaining a two-dimensional blood vessel image sequence containing the main contour of the blood vessel.

[0034] Its technical effect is as follows: The Sobel operator can better identify the edges of blood vessels by calculating the horizontal and vertical gradients of each pixel point in the image. For blood vessels of different thicknesses, by adjusting the size and weight of the Sobel convolution kernel, the adaptability of the edge detection algorithm under different blood vessel morphologies is improved, and the detection ability for thin blood vessels or tiny lesion areas is enhanced. Through effective threshold processing by combining the gradient direction and edge intensity, broken edges are connected and the extracted blood vessel contour is made more coherent, and the overall contour is not broken due to local weak edges. Through the intensity and direction information of the edges, the morphological features of the blood vessels can be more accurately described, including the bending, branching and other structures of the blood vessels, realizing the accurate identification of the blood vessels.

[0035] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the conversion of the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm includes:

[0036] Extract the feature points of the blood vessels from multiple perspectives in the two-dimensional blood vessel image sequence to represent the key feature regions of the blood vessels.

[0037] According to the feature points extracted from each perspective, apply the beam normal vector optimization algorithm to project the two-dimensional feature points into three-dimensional space and optimize to obtain the three-dimensional point cloud coordinates The three-dimensional point cloud coordinates conform to the projections in all perspectives, and the optimization objective is to minimize the error function where n is the total number of feature points, m is the total number of perspectives, and e i =||p i -p′ i || 2 =(x i -x′ i ) 2 +(y i -y′ i ) 2 is the projection error of each feature point, and p i =(xi , y i ) is the two-dimensional coordinate of the feature point projected by the blood vessel in the j-th perspective, P j is the camera projection matrix representing the mapping relationship from the three-dimensional space to the two-dimensional image plane, P i = (X i , Y i , Z i ) is the three-dimensional coordinate point corresponding to each two-dimensional feature point, p′ i = (x′ i , y′ i ) is the projection position of the three-dimensional point P i in the j-th perspective.

[0038] Generate a three-dimensional point cloud model representing the spatial distribution of blood vessels from the three-dimensional point cloud coordinates of each perspective.

[0039] Merge the three-dimensional point cloud models of each perspective and perform surface reconstruction to generate a three-dimensional surface model of the blood vessel.

[0040] By extracting the feature points of the blood vessel from multiple perspectives and applying the beam normal vector optimization algorithm, the two-dimensional feature points are accurately projected into the three-dimensional space, and finally a three-dimensional point cloud that conforms to the projections of all perspectives is generated. By merging the three-dimensional point clouds of multiple perspectives and using a surface reconstruction algorithm such as Delaunay triangulation, a three-dimensional surface model of the blood vessel is generated, providing accurate three-dimensional space information for subsequent cardiovascular balloon positioning, path planning, and ablation operations.

[0041] Its technical effect lies in: by comprehensively integrating the two-dimensional image feature point information of multiple perspectives, it avoids the blood vessel reconstruction errors that may be caused by a single perspective. Combining the beam normal vector optimization algorithm, according to the principle of minimizing the projection error, the two-dimensional feature points are accurately projected into the three-dimensional space. By optimizing and adjusting the position of the three-dimensional point cloud, the projection error in all perspectives is minimized, thus ensuring the accuracy of the three-dimensional blood vessel model.

[0042] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, in the three-dimensional blood vessel model, based on deep learning classification network and image generation adversarial network technologies, identifying the opening position and morphological characteristics of the target blood vessel, and extracting the blood vessel lesion area for positioning ablation targets includes:

[0043] Voxelize the three-dimensional blood vessel model to obtain three-dimensional image data.

[0044] Design a deep learning classification network model, and use the three-dimensional image data as the input of the deep learning classification network model to represent the gray scale or intensity information of the blood vessel in the three-dimensional space.

[0045] Train the deep learning classification network model using the labeled dataset, and identify the vascular opening and morphological features by optimizing the objective function.

[0046] The output of the deep learning classification network model is the classification probability of each voxel or vascular region, indicating whether the point belongs to the target vascular opening region.

[0047] Extract the vascular morphological features from the intermediate feature map and label map of the deep learning classification network model. The vascular morphological features include at least one of the vascular branch angle, vascular diameter, and curvature.

[0048] Design the generator and discriminator of the image generation adversarial network model. The generator receives the three-dimensional image data and generates the vascular lesion region. The discriminator distinguishes the region generated by the generator from the real vascular lesion region and outputs the spatial distribution of the vascular morphology and lesion region.

[0049] According to the extracted lesion region and the spatial distribution, calculate the position of the lesion region in the three-dimensional space, and obtain the vascular lesion region for positioning the ablation target in the three-dimensional vascular model.

[0050] The technical effect is that the GAN generator can generate possible lesion regions according to the three-dimensional image data. Through the adversarial training with the discriminator, the generator continuously optimizes the generation result. The discriminator is responsible for distinguishing the generated lesion region from the real lesion region, enhancing the accuracy of the spatial distribution of the vascular lesion region. Combining the vascular morphological features and the lesion region improves the accuracy of treatment target positioning. It can not only identify the location of the lesion region but also provide more characteristic information about the morphology, size, degree, etc. of the lesion region. Determine the optimal positioning point of the balloon through the morphological features of the blood vessel and optimize the movement path of the balloon in the blood vessel. Based on the generated vascular lesion region and the corresponding spatial distribution data, the coordinates of the lesion region in the three-dimensional space can be accurately calculated. By accurately calculating the spatial distribution and position of the lesion region, high-precision three-dimensional space data can be provided for balloon positioning and path planning.

[0051] In a preferred embodiment of the present invention, in the above visual positioning and ablation method for a cardiovascular balloon, the combination of hemodynamic simulation, based on the three-dimensional vascular model and the vascular lesion region, calculating the optimal positioning point coordinates of the cardiovascular balloon, and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates includes:

[0052] Establish a hemodynamic simulation equation where ρ is the density of blood, v is the velocity field, e is the pressure, μ is the viscosity of blood, F is the external force, and ▽ is the gradient operator.

[0053] From the three-dimensional blood vessel model and the blood vessel lesion area, select a candidate positioning point R, where the candidate positioning point R is located at the front end of the blood vessel lesion area or near the blood vessel lesion area.

[0054] Calculate the optimal positioning point coordinates of the cardiovascular balloon Among them, the optimal positioning point is the point with the least impact on blood flow when the balloon passes through the blood vessel, Q(R) is the flow rate of the candidate positioning point, and ▽e is the pressure gradient of the candidate positioning point.

[0055] Establish a blood flow perturbation model during the movement of the balloon. For any path point r u , calculate the blood flow perturbation Among them, v u is the blood flow velocity at the path point r u , is the velocity of the balloon, and ▽r u is the pressure gradient at the path point r u .

[0056] Minimize the blood flow perturbation δQ u and optimize the path length Distance(r u-1 , r u ) as the goal to establish an optimization function Among them, L path is the total path cost, representing the total cost of the balloon from the current position to the optimal positioning point, and w 1 is the distance weight coefficient.

[0057] Solve the optimization function, calculate the total cost through the path selection algorithm, and the cost update formula is C(r v ) = min(C(r v ), C(r u ) + δQ u + w 2 ·Distance(r u , r v ))), where r v is the adjacent path node of the path point r u , C(r v ) is the cost of the r v node, Distance(r u , r v ) is the geometric distance between the node r v and the path point r u , representing the physical distance of the balloon from one node to another node, and w 2 is the distance weight coefficient. Select the path with the minimum cost as the final path to obtain the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates.

[0058] Its technical effects are as follows: Based on the physical characteristics of blood flow, it can accurately simulate the influence of blood flow on the movement of the balloon, and then optimize the movement path of the balloon. By calculating the relationship between the vascular lesion area and blood flow characteristics, the optimal positioning point with the least influence of the balloon on blood flow is determined, maximizing the treatment effect and minimizing the disturbance to blood flow. By calculating and minimizing blood flow disturbance in real time during the path planning process, that is, the changes in blood flow velocity and pressure gradient, excessive disturbance to blood flow caused by the rapid movement or unstable expansion of the balloon is avoided. By optimizing the weight coefficients of the path length and blood flow disturbance in the optimization function, it is ensured that the influence of the balloon on blood flow during movement is as small as possible, while ensuring an appropriate path length and optimizing the speed and accuracy of treatment. By calculating the total cost of the path in real time and selecting the path with the minimum cost, the movement trajectory of the balloon can be adjusted in real time during the treatment process to adapt to the requirements of vascular morphology and lesion changes. By optimizing the cost function of the path, including geometric distance and blood flow disturbance, the most suitable path can be efficiently selected, thereby reducing the treatment time and unnecessary risks during the treatment while ensuring the position accuracy of the balloon.

[0059] In a preferred embodiment of the present invention, in the above visual positioning and ablation method for a cardiovascular balloon, after obtaining the optimal path, the optimal path is smoothed to obtain a smoothed optimal path. Among them, is the B-spline basis function, r u is the path point r u is the control point of.

[0060] Its technical effects are as follows: There may be bends or small local irregularities in the blood vessel. The path without smoothing may contain sharp turns or unnatural corners, which are likely to cause difficulties in balloon operation or unnecessary compression on the blood vessel wall. Through B-spline smoothing, the path will be more natural and avoid unstable path movement. The optimized smoothed path helps to reduce the friction between the balloon and the blood vessel wall, reduce the oscillation and instability during the movement of the balloon, and enhance the path operability and control accuracy.

[0061] In a preferred embodiment of the present invention, in the above visual positioning and ablation method for a cardiovascular balloon, the method of obtaining the real-time position information of the cardiovascular balloon through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the optimal path, and using a navigation algorithm to adjust the position of the cardiovascular balloon includes:

[0062] Embed a micro sensor in the balloon catheter to detect the electromagnetic field signal and position information in real time and send them outwards.

[0063] Receive the electromagnetic field signal and the position information, and calculate the spatial coordinates P of the cardiovascular balloon.sensor =(X sensor , Y sensor , Z sensor ) and the direction angle θ sensor .

[0064] Calculate the position error ΔP(t) and the direction error Δθ between the current position of the cardiovascular balloon and the target path point.

[0065] Calculate the adjustment amounts of the orientation and movement speed of the cardiovascular balloon, and obtain an adjustment control signal using a proportional-integral-derivative control algorithm for controlling the movement of the balloon. Among them, the formula of the proportional-integral-derivative control algorithm is u(t) is the control signal, K p is the proportional coefficient, K d is the differential coefficient, K i is the integral gain coefficient, ΔP(t) is the position error, is the rate of change of the position error.

[0068] Its technical effect is that: by embedding a micro sensor in the balloon catheter, the electromagnetic field signal is detected in real time and the position information of the balloon is transmitted, which is used to calculate the specific position of the balloon in the blood vessel, helping the automation system to adjust the movement of the balloon at any time to ensure that the balloon is always in the target area or the optimal path. By calculating the position error and the direction error, the proportional-integral-derivative PID control algorithm is used to dynamically adjust the balloon, smoothly guiding the balloon to move along the target path, avoiding over-adjustment or over-compensation, and ensuring that the balloon can quickly follow the path during movement without generating drastic movement changes. Through the electromagnetic positioning and PID control automation control system, the movement of the balloon can be adjusted and feedback in real time. Even in a complex blood vessel environment, the balloon can dynamically adjust the path according to the real-time calculated data, accurately position the balloon in the target area, enabling the balloon to accurately avoid the obstacles in the blood vessel, improving the treatment efficiency and accuracy.

[0069] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the acquisition of real-time parameters during ablation and the monitoring and real-time adjustment control of ablation parameters include:

[0070] Acquire the temperature, pressure and energy transmission conditions during ablation.

[0071] Real-time monitor whether the temperature reaches the preset ablation range, whether the balloon pressure exceeds the normal range, and whether the energy transmission rate exceeds the normal range.

[0072] Based on the parameters monitored in real time, adjust the control signal based on a preset threshold to control the ablation strategy.

[0073] Its technical effect lies in: by collecting real-time parameters during the ablation process and adjusting the control signal based on the real-time monitoring data and preset thresholds to optimize the ablation strategy, ensuring precise control of the ablation process and avoiding over-treatment or under-treatment.

[0074] A visual positioning and ablation system for a cardiovascular balloon, comprising:

[0075] An image acquisition module for acquiring a real-time image sequence by using a multi-channel imaging system and synchronous multi-view image acquisition technology.

[0076] A two-dimensional extraction module for extracting the main contour of blood vessels from the real-time image sequence by using an edge detection algorithm to obtain a two-dimensional blood vessel image sequence.

[0077] A three-dimensional modeling module for converting the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm.

[0078] A blood vessel lesion area extraction module for identifying the opening position and morphological features of the target blood vessel in the three-dimensional blood vessel model based on deep learning classification network and image generation adversarial network technologies, and extracting the blood vessel lesion area for positioning the ablation target.

[0079] An optimal path planning module for calculating the optimal positioning point coordinates of the cardiovascular balloon and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates by combining hemodynamic simulation based on the three-dimensional blood vessel model and the blood vessel lesion area.

[0080] A navigation adjustment module for obtaining the real-time position information of the balloon catheter through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the best path, and adjusting the position of the cardiovascular balloon using a navigation algorithm.

[0081] A parameter adjustment module for collecting real-time parameters during the ablation process and monitoring and real-time adjusting and controlling the ablation parameters.

[0082] The beneficial effects of the embodiments of the present invention are:

[0083] By using multi-channel imaging systems such as CT, MRI, and flexible endoscopes, combined with synchronous multi-view image acquisition technology, the vascular structure can be captured in real-time and comprehensively, especially in complex blood vessels or lesion areas, providing multi-level information with different angles and resolutions. CT imaging provides high-resolution cross-sectional images of blood vessels and surrounding tissues. MRI imaging can obtain images of the thickness of the blood vessel wall and blood flow characteristics. The endoscope provides real-time direct-view images, thus comprehensively analyzing information such as the morphology, lesions, and flow patterns of blood vessels, and helping with subsequent precise model reconstruction. By synchronously recording and aligning multi-channel data with timestamps, it ensures that the image information at each moment is accurately registered, providing high-quality input data for subsequent image processing and 3D modeling, and reducing the risk of data deviation or misalignment.

[0084] Through the Sobel edge detection algorithm for two-dimensional gradient calculation to extract the main contours of blood vessels, it can accurately identify the blood vessel boundaries, especially in complex and small blood vessels, effectively reducing image noise and misidentification, and improving the accuracy of contour extraction. By reasonably setting the high threshold and low threshold, edge detection can not only highlight the main edges of blood vessels but also remove irrelevant noise, ensuring that the extracted blood vessel contours are more precise and complete. The contour extraction algorithm further enhances the ability to extract blood vessel edges.

[0085] Through the multi-view image reconstruction algorithm, the two-dimensional blood vessel image sequence can be transformed into an accurate three-dimensional blood vessel model. The optimized three-dimensional model can more realistically reflect the geometric morphology of blood vessels, especially in the performance of curved, branched, or lesion areas, enhancing the realism and operability of the blood vessel structure. The three-dimensional point cloud coordinates optimized by this algorithm conform to the projections of all views, minimizing the reconstruction error, significantly improving the accuracy of the three-dimensional blood vessel model, ensuring the accurate positioning of the lesion area, and providing accurate data support for the subsequent treatment process.

[0086] Based on deep learning classification networks and the image generation adversarial network GAN technology, it can automatically identify and extract the blood vessel lesion areas, avoiding the errors of manual operations, and being able to efficiently and accurately identify the target areas, especially in the case of complex and tiny lesions. Through the classification probability output by the deep learning classification network and the lesion areas generated by GAN, it can provide treatment goals matching the specific conditions of the patient, ensuring that the treatment for each patient can be personalized and optimized according to the real morphology of the lesion area, calculating the precise position of the lesion area in three-dimensional space, and providing high-quality spatial data for the precise positioning of the balloon.

[0087] Based on hemodynamic simulation equations, accurately calculate the optimal positioning point of the cardiovascular balloon in the blood vessel. By calculating the flow rate, pressure gradient, and blood flow disturbance along the path, the system can select the path point with the least impact on blood flow, reduce the negative impact on the blood vessel, ensure the treatment effect while maintaining blood flow stability. Through the blood flow disturbance model and path planning algorithm, it is possible to select the most suitable path according to optimization objectives such as minimizing blood flow disturbance and optimizing path length, avoiding excessive physical impact of the balloon on the blood vessel or excessive disturbance to the blood flow, thereby reducing the risk of complications.

[0088] Through electromagnetic positioning technology, obtain the position information of the balloon in real time, enabling the balloon to move precisely in complex blood vessels. Combining the calculation of position error and direction error, the navigation algorithm can dynamically adjust the position of the balloon to ensure that the balloon accurately reaches the treatment target. Through the proportional-integral-derivative (PID) control algorithm, the system can finely adjust the movement trajectory of the balloon to ensure that the balloon moves precisely along the optimal path, avoiding the accumulation of path errors and position deviations. Brief Description of the Drawings

[0089] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0090] Figure 1 It is a flowchart of the visual positioning and ablation method for the cardiovascular balloon of the present invention. Detailed Embodiments

[0091] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0092] Please refer to Figure 1, the first embodiment of the present invention provides a visual positioning and ablation method for a cardiovascular balloon, including: using a multi-channel imaging system to obtain a real-time image sequence through synchronous multi-view image acquisition technology; using an edge detection algorithm to extract the main contour of blood vessels from the real-time image sequence to obtain a two-dimensional blood vessel image sequence; converting the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm; in the three-dimensional blood vessel model, based on deep learning classification network and image generation adversarial network technologies, identifying the opening position and morphological characteristics of the target blood vessel, and extracting the blood vessel lesion area for positioning the ablation target; combining hemodynamic simulation, based on the three-dimensional blood vessel model and the blood vessel lesion area, calculating the optimal positioning point coordinates of the cardiovascular balloon, and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates; obtaining the real-time position information of the cardiovascular balloon through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the best path, and using a navigation algorithm to adjust the position of the cardiovascular balloon; collecting real-time parameters during the ablation process, and monitoring and real-time adjusting the ablation parameters.

[0093] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the step of using a multi-channel imaging system to obtain a real-time image sequence through synchronous multi-view image acquisition technology includes: obtaining cross-sectional images of blood vessels and surrounding tissues through a CT imaging channel; obtaining images of the thickness of the blood vessel wall and blood flow characteristics through an MRI imaging channel; obtaining direct-view images inside the cardiovascular system in real time through a flexible endoscope imaging channel; setting the acquisition frequency and image resolution of each imaging channel, and synchronously recording the images collected by each imaging channel according to the time stamp through an image sensor, aligning the multi-channel data, and obtaining a real-time image sequence.

[0094] The technical effect is that: through the multi-channel imaging system, different anatomical features and lesion areas of blood vessels can be fully captured. Especially in complex situations such as blood vessel bending, stenosis or branching, more comprehensive blood vessel anatomical information is provided, and the complementary nature of different imaging technologies is utilized, making subsequent three-dimensional reconstruction and lesion area identification more accurate. In cases where the blood vessel morphology is complex or there are blood flow fluctuations, the real-time acquired image sequence can provide immediate feedback information, avoiding positioning errors caused by delays.

[0095] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the step of using an edge detection algorithm to extract the main contour of blood vessels from the real-time image sequence to obtain a two-dimensional blood vessel image sequence includes: based on the Sobel edge detection algorithm, performing two-dimensional directional gradient calculation on the images in the real-time image sequence, and the horizontal gradient of each pixel point in the X direction The vertical gradient of each pixel point in the Y direction Among them, K x (a, b) is the Sobel convolution kernel in the X direction, and K y (a, b) is the Sobel convolution kernel in the Y direction. The Sobel convolution kernel is designed according to the thickness of the blood vessels. I(x + a, y + b) is the pixel value at the current position in the image;

[0096] Specifically, the thickness of the blood vessels will directly affect the change of edges in the image. For thicker blood vessels, the edge change is more obvious, and a stronger gradient signal can help detection; while for thinner blood vessels, the edge change is smaller, and a more refined convolution kernel design is required to detect the tiny gradient change. Therefore, when designing the Sobel convolution kernels in the X and Y directions, the size and weight of the convolution kernel can be adjusted according to the thickness of the blood vessels. For thicker blood vessels, a larger convolution kernel is used, such as 3x3, 5x5 or even 7x7. In this way, the convolution kernel can capture the edge information in a larger area and help extract obvious edges; for thinner blood vessels, a smaller convolution kernel is used, such as 3x3 or a custom smaller size. The small convolution kernel can focus on the local features of the image, effectively extract the subtle gradient changes, and avoid over-smoothing the details.

[0097] Calculate the gradient magnitude of each pixel point to represent the change intensity of the pixel point; calculate the gradient direction of each pixel point θ(x, y) = atan2(G y (x, y), G x (x, y)), to represent the direction of the edge; according to the edge intensity and morphological features of the blood vessels, set the high threshold T H and the low threshold T L ;

[0098] Specifically, the edge intensity of the blood vessels is usually represented by calculating the gradient magnitude of each pixel point in the image. The gradient magnitude reflects the degree of gray change in the image. For the blood vessel area, its edge usually shows a sharp change in gray level, especially at the boundary between the blood vessel wall and the surrounding tissue. The area with a larger edge intensity indicates the main contour of the blood vessel in this area. The high threshold T H is used to screen out the areas with larger edge intensity, that is, the obvious blood vessel edges. The strong edges correspond to the clear boundaries of the blood vessels, usually the transition area between the background and the blood vessels; the low threshold T L is used to screen out the areas with weaker edges. These weak edges may be the detailed parts of the blood vessels or the areas with relatively blurred edge transitions. The weak edges usually need to be connected to the strong edges to be considered as valid blood vessel edges.

[0099] The morphology of blood vessels, such as straight lines, curves, or branches, also affects the setting of the threshold. Usually, the edges of blood vessels can appear curved or irregular, and the edge strength in these areas may be relatively weak. For blood vessels with complex morphologies, a high threshold T H helps to remove weak noise and irregular morphologies, thus retaining the most prominent edges; at blood vessel branch points or bends, weak edges may still represent valid blood vessel edges, so a low threshold T L needs to be set lower to ensure that these important details are retained.

[0100] Pixels with a gradient magnitude greater than the high threshold are valid edges, and the pixel is retained; for pixels with a gradient magnitude less than the high threshold but greater than the low threshold, it is determined whether the pixel is connected to a strong edge. If it is connected, the pixel is considered a valid edge and retained; if not, the pixel is not a valid edge and is removed; pixels with a gradient magnitude less than the low threshold do not belong to the edge and are removed; the retained pixels are extracted to obtain a binary image; using a contour extraction algorithm, such as the findContours function in OpenCV, the contours of the blood vessels are extracted from the binary image to obtain a two-dimensional blood vessel image sequence containing the main contours of the blood vessels.

[0101] The technical effect is that the Sobel operator can better identify the edges of blood vessels by calculating the horizontal and vertical gradients of each pixel in the image. For blood vessels of different thicknesses, by adjusting the size and weight of the Sobel convolution kernel, the adaptability of the edge detection algorithm under different blood vessel morphologies is improved, and the detection ability for thin blood vessels or small lesion areas is enhanced. By performing effective threshold processing by combining the gradient direction and edge strength, broken edges are connected and the extracted blood vessel contours are made more coherent, and the overall contour is not broken due to local weak edges. Through the edge strength and direction information, the morphological characteristics of blood vessels, including the curvature, branches, etc. of the blood vessels, can be more accurately described, realizing the accurate identification of blood vessels.

[0102] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the conversion of the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model by a multi-view image reconstruction algorithm includes: extracting feature points of blood vessels from multiple views in the two-dimensional blood vessel image sequence to represent key feature regions of the blood vessels; according to the feature points extracted in each view, applying a beam method vector optimization algorithm to project the two-dimensional feature points into three-dimensional space to optimize and obtain three-dimensional point cloud coordinates The three-dimensional point cloud coordinates conform to the projections in all views, and the optimization objective is to minimize the error function where n is the total number of feature points, m is the total number of views, and e i =||p i -p′ i ||2 =(x i -x' i ) 2 +(y i -y' i ) 2 is the projection error of each feature point, p i =(x i , y i ) is the two-dimensional coordinate of the feature point projected by the blood vessel in the j-th view, P j is the camera projection matrix representing the mapping relationship from the three-dimensional space to the two-dimensional image plane, P i =(X i , Y i , Z i ) is the three-dimensional coordinate point corresponding to each two-dimensional feature point, p' i =(x' i , y' i ) is the projection position of the three-dimensional point P i in the j-th view; generate a three-dimensional point cloud model representing the spatial distribution of the blood vessel from the three-dimensional point cloud coordinates of each view; merge the three-dimensional point cloud models of each view and perform surface reconstruction to generate a three-dimensional surface model of the blood vessel.

[0103] By extracting the feature points of the blood vessel from multiple views and applying the bundle adjustment algorithm, the two-dimensional feature points are accurately projected into the three-dimensional space, and finally a three-dimensional point cloud that conforms to the projections of all views is generated. By merging the three-dimensional point clouds of multiple views and using a surface reconstruction algorithm such as Delaunay triangulation, a three-dimensional surface model of the blood vessel is generated, providing accurate three-dimensional spatial information for subsequent cardiovascular balloon positioning, path planning, and ablation operations.

[0104] Its technical effect is that: by integrating the two-dimensional image feature point information of multiple views, the blood vessel reconstruction error caused by a single view is avoided. Combining the bundle adjustment algorithm, according to the principle of minimizing the projection error, the two-dimensional feature points are accurately projected into the three-dimensional space. By optimizing and adjusting the position of the three-dimensional point cloud, the projection error in all views is minimized, thus ensuring the accuracy of the three-dimensional blood vessel model.

[0105] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, in the three-dimensional vascular model, based on deep learning classification network and image generation adversarial network technologies, identifying the opening position and morphological features of the target blood vessel, and extracting the blood vessel lesion area for positioning the ablation target includes: performing voxelization processing on the three-dimensional vascular model to obtain three-dimensional image data; designing a deep learning classification network model, using the three-dimensional image data as the input of the deep learning classification network model, which is used to represent the gray scale or intensity information of the blood vessel in three-dimensional space; training the deep learning classification network model with a labeled data set, and identifying the blood vessel opening and morphological features by optimizing the objective function; the output of the deep learning classification network model is the classification probability of each voxel or blood vessel area, indicating whether the point belongs to the target blood vessel opening area; extracting the blood vessel morphological features from the intermediate feature map and label map of the deep learning classification network model, and the blood vessel morphological features include at least one of the blood vessel branch angle, blood vessel diameter, and curvature; designing the generator and discriminator of the image generation adversarial network model, the generator receives the three-dimensional image data and generates the blood vessel lesion area, the discriminator distinguishes the area generated by the generator from the real blood vessel lesion area, and outputs the spatial distribution of the blood vessel morphology and lesion area; according to the extracted lesion area and the spatial distribution, calculating the position of the lesion area in three-dimensional space, and obtaining the blood vessel lesion area for positioning the ablation target in the three-dimensional vascular model.

[0106] The technical effect is that the GAN generator can generate possible lesion areas according to the three-dimensional image data. Through the adversarial training with the discriminator, the generator continuously optimizes the generation result, and the discriminator is responsible for distinguishing the generated lesion area from the real lesion area, enhancing the accuracy of the spatial distribution of the blood vessel lesion area. Combining the blood vessel morphological features and the lesion area improves the accuracy of treatment target positioning. It can not only identify the position of the lesion area, but also provide more feature information about the morphology, size, degree, etc. of the lesion area, determine the optimal positioning point of the balloon through the morphological features of the blood vessel, and optimize the movement path of the balloon in the blood vessel. Based on the generated blood vessel lesion area and the corresponding spatial distribution data, the coordinates of the lesion area in three-dimensional space can be accurately calculated. By accurately calculating the spatial distribution and position of the lesion area, high-precision three-dimensional space data can be provided for balloon positioning and path planning.

[0107] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the step of combining hemodynamic simulation, based on the three-dimensional vascular model and the blood vessel lesion area, calculating the optimal positioning point coordinates of the cardiovascular balloon, and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates includes: establishing a hemodynamic simulation equation Among them, ρ is the density of blood, v is the velocity field, e is the pressure, μ is the viscosity of blood, F is the external force, such as the frictional force of the blood vessel wall or gravity, and ▽ is the gradient operator. Specifically, when performing the gradient operation on the pressure field e, its physical meaning is to represent the direction and rate of pressure change at a certain point. In the hemodynamic simulation equation, the gradient ▽e represents the pressure gradient, that is, the pressure change rate of the fluid; when performing the gradient operation on the velocity field v, the result is a tensor ▽v matrix, which represents the change of the velocity field in space, representing the change rate and direction of the velocity field; ▽ 2 ▽²v is the Laplace operator of the velocity field, representing the diffusion effect of the velocity field, that is, the viscosity of the fluid; according to the three-dimensional blood vessel model and the blood vessel lesion area, a candidate positioning point R is selected, and the candidate positioning point R is located at the front end of the blood vessel lesion area or near the blood vessel lesion area; calculate the optimal positioning point coordinates of the cardiovascular balloon Among them, the optimal positioning point is the point with the least impact on blood flow when the balloon passes through the blood vessel, Q(R) is the blood flow at the candidate positioning point, and ▽e is the pressure gradient at the candidate positioning point; establish a blood flow disturbance model during the movement of the balloon. For any path point r u , calculate the blood flow disturbance Among them, v u is the blood flow velocity at the path point r u , is the velocity of the balloon, and ▽r u is the pressure gradient at the path point r u ; to minimize the blood flow disturbance δQ u and optimize the path length Distance(r u-1 , r u ) as the goal to establish an optimization function Among them, L path is the total path cost, representing the total cost of the balloon from the current position to the optimal positioning point, and w 1 is the distance weight coefficient, which is used to control the relative importance between the blood flow disturbance and the path length. A higher w 1 value will make the path length more important, and a lower w 1 value pays more attention to the blood flow disturbance; solve the optimization function, calculate the total cost through the path selection algorithm, and the cost update formula is C(r v ) = min(C(r v ), C(r u ) + δQ u + w 2 ·Distance(r u , r v ))), where r v is the adjacent path node of the path point r u , and C(r v) is r v The cost of the node, Distance(r u , r v ) is the node r v and the path point r u The geometric distance between them represents the physical distance of the balloon from one node to another node. w 2 is the distance weight coefficient. The path with the minimum cost is selected as the final path to obtain the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates.

[0108] Its technical effect is as follows: Based on the physical characteristics of blood flow, it can accurately simulate the influence of blood flow on the movement of the balloon, and then optimize the movement path of the balloon. By calculating the relationship between the vascular lesion area and blood flow characteristics, the optimal positioning point with the least influence of the balloon on blood flow is determined, maximizing the treatment effect and minimizing the disturbance to blood flow. By calculating and minimizing the blood flow disturbance in real time during the path planning process, that is, the changes in blood flow velocity and pressure gradient, excessive disturbance to blood flow caused by the rapid movement or unstable expansion of the balloon is avoided. By optimizing the weight coefficients of the path length and blood flow disturbance in the optimization function, it is ensured that the influence of the balloon on blood flow during movement is as small as possible, and at the same time, the path length is appropriate, optimizing the speed and accuracy of treatment. By calculating the total cost of the path in real time and selecting the path with the minimum cost, the movement trajectory of the balloon can be adjusted in real time during the treatment process to adapt to the needs of vascular morphology and lesion changes. By optimizing the cost function of the path, including geometric distance and blood flow disturbance, the most suitable path can be efficiently selected, thus reducing the treatment time and unnecessary risks during the treatment while ensuring the position accuracy of the balloon.

[0109] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, after obtaining the best path, the best path is smoothed to obtain a smoothed best path. Among them, is the B-spline basis function, r u is the path point r u The control point of.

[0110] Its technical effect is as follows: There may be bends or small local irregularities in the blood vessel. The path without smoothing may contain sharp turns or unnatural corners, which are likely to cause difficulties in balloon operation or unnecessary compression on the blood vessel wall. Through B-spline smoothing, the path will be more natural, avoiding unsteady path movement. The optimized smoothed path helps to reduce the friction between the balloon and the blood vessel wall, reduce the oscillation and instability during the movement of the balloon, and enhance the path operability and control accuracy.

[0111] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, obtaining the real-time position information of the cardiovascular balloon through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the optimal path, and using a navigation algorithm to adjust the position of the cardiovascular balloon includes: embedding a micro-sensor in the balloon catheter to detect electromagnetic field signals and position information in real time and transmit them outward; receiving the electromagnetic field signals and the position information, and calculating the spatial coordinates P sensor =(X sensor , Y sensor , Z sensor ) and the direction angle θ sensor ; calculating the position error ΔP(t) and the direction error Δθ between the current position of the cardiovascular balloon and the target path point; calculating the adjustment amounts of the orientation and movement speed of the cardiovascular balloon, and using a proportional-integral-derivative control algorithm to obtain an adjustment control signal for controlling the movement of the balloon, where the formula of the proportional-integral-derivative control algorithm is u(t) is the control signal, K p is the proportional coefficient, K d is the differential coefficient, K i is the integral gain coefficient, ΔP(t) is the position error, is the change rate of the position error.

[0114] Its technical effect is as follows: By embedding a micro-sensor in the balloon catheter, electromagnetic field signals are detected in real time and the position information of the balloon is transmitted, which is used to calculate the specific position of the balloon in the blood vessel, helping the automation system to adjust the movement of the balloon at any time to ensure that the balloon is always in the target area or the optimal path. By calculating the position error and the direction error, the proportional-integral-derivative PID control algorithm is used to dynamically adjust the balloon, smoothly guiding the balloon to move along the target path, avoiding over-adjustment or over-compensation, and ensuring that the balloon can quickly follow the path during movement without generating violent movement changes. Through the electromagnetic positioning and PID control automation control system, the movement of the balloon can be adjusted and feedback in real time. Even in a complex blood vessel environment, the balloon can dynamically adjust the path according to the real-time calculated data, accurately position the balloon in the target area, enabling the balloon to accurately avoid obstacles in the blood vessel, and improving the treatment efficiency and accuracy.

[0115] In a preferred embodiment of the present invention, in the above-mentioned visual positioning and ablation method for a cardiovascular balloon, the acquisition of real-time parameters during ablation and the monitoring and real-time adjustment control of ablation parameters include: acquiring the temperature, pressure, and energy transmission conditions during ablation; real-time monitoring whether the temperature reaches the preset ablation range, whether the balloon pressure exceeds the normal range, and whether the energy transmission rate exceeds the normal range; based on the parameters monitored in real time, adjusting the control signal based on a preset threshold to control the ablation strategy.

[0116] The technical effect is that by acquiring real-time parameters during ablation and adjusting the control signal based on real-time monitoring data and preset thresholds to optimize the ablation strategy, precise control of the ablation process is ensured, avoiding over-treatment or under-treatment.

[0117] The second embodiment of the present invention provides a visual positioning and ablation system for a cardiovascular balloon, including: an image acquisition module for acquiring a real-time image sequence by using a multi-channel imaging system through synchronous multi-view image acquisition technology; a two-dimensional extraction module for extracting the main contour of blood vessels from the real-time image sequence by using an edge detection algorithm to obtain a two-dimensional blood vessel image sequence; a three-dimensional modeling module for converting the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm; a blood vessel lesion area extraction module for identifying the opening position and morphological characteristics of the target blood vessel in the three-dimensional blood vessel model based on deep learning classification network and image generation adversarial network technologies, and extracting the blood vessel lesion area for positioning the ablation target; an optimal path planning module for calculating the optimal positioning point coordinates of the cardiovascular balloon based on the three-dimensional blood vessel model and the blood vessel lesion area by combining hemodynamic simulation, and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates; a navigation adjustment module for obtaining the real-time position information of the balloon catheter through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the best path, and adjusting the position of the cardiovascular balloon using a navigation algorithm; a parameter adjustment module for acquiring real-time parameters during ablation and monitoring and real-time adjustment control of ablation parameters.

[0118] The computer program product of the visual positioning and ablation method and device for a cardiovascular balloon provided by the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods in the foregoing method embodiments. For specific implementation, reference can be made to the method embodiments and will not be elaborated here.

[0119] Specifically, the storage medium can be a general storage medium, such as a removable disk, a hard disk, etc. When the computer program on the storage medium is run, it can execute the above-mentioned visual positioning and ablation method for a cardiovascular balloon, so as to combine modern imaging technology, artificial intelligence, hemodynamic simulation and electromagnetic positioning technology to achieve dynamic control of the ablation process.

[0120] If the above-mentioned functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0121] Finally, it should be noted that the above-mentioned embodiments are only specific implementation manners of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions recorded in the foregoing embodiments or easily conceive of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A visual positioning and ablation method for cardiovascular balloons, characterized in that: include: Adopt multi-channel imaging system to obtain real-time image sequences through synchronous multi-view image acquisition technology; Extracting the main contour of the blood vessel from the real-time image sequence using an edge detection algorithm to obtain a two-dimensional blood vessel image sequence; The two-dimensional vascular image sequence is converted into a three-dimensional vascular model through a multi-view image reconstruction algorithm; In the three-dimensional blood vessel model, based on the deep learning classification network and the image generation adversarial network technology, the opening position and morphological characteristics of the target blood vessel are identified, and the vascular lesion area used to locate the ablation target is extracted; In combination with hemodynamic simulation, based on the three-dimensional vascular model and the vascular lesion area, the optimal positioning point coordinates of the cardiovascular balloon are calculated, and the optimal path of the cardiovascular balloon from the current position to the optimal positioning point coordinates is planned; Acquiring real-time position information of the cardiovascular balloon through electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the optimal path, and adjusting the position of the cardiovascular balloon using a navigation algorithm; Collect real-time parameters during the ablation process, monitor and adjust and control the ablation parameters in real time.

2. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The multi-channel imaging system is used to obtain a real-time image sequence through a synchronous multi-view image acquisition technology, including: Obtain cross-sectional images of blood vessels and surrounding tissues through CT imaging channels; The thickness of the blood vessel wall and the blood flow characteristics images are obtained through the MRI imaging channel; Obtain direct images of the cardiovascular system in real time through the flexible endoscope imaging channel; The acquisition frequency and image resolution of each imaging channel are set, and the acquired images of each imaging channel are synchronously recorded according to the timestamp through the image sensor, and the multi-channel data are aligned to obtain a real-time image sequence.

3. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The extracting the main contour of the blood vessel from the real-time image sequence using an edge detection algorithm to obtain a two-dimensional blood vessel image sequence comprises: Based on the Sobel edge detection algorithm, two-dimensional directional gradient calculation is performed on the images in the real-time image sequence. The horizontal gradient of each pixel in the X direction The vertical gradient of each pixel in the Y direction Among them, K x (a, b) is the Sobel convolution kernel in the X direction, K y (a, b) is the Sobel convolution kernel in the Y direction, which is designed according to the thickness of the blood vessel, and I (x+a, y+b) is the pixel value at the current position in the image; Calculate the gradient magnitude for each pixel Used to indicate the change intensity of the pixel point; Calculate the gradient direction of each pixel θ(x,y)=atan2(G y (x,y),G x (x,y)), used to indicate the direction of the edge; According to the edge strength and morphological characteristics of the blood vessels, a high threshold T is set. H and low threshold T L ; A pixel point whose gradient amplitude is greater than the high threshold is a valid edge, and the pixel point is retained; For pixels whose gradient amplitude is less than the high threshold but greater than the low threshold, determine whether the pixel is connected to a strong edge. If so, the pixel is considered to be a valid edge and is retained. If not, the pixel is not a valid edge and is removed. Pixels whose gradient amplitude is less than the low threshold do not belong to the edge and are removed; Extract the retained pixels to obtain a binary image; The contour extraction algorithm is used to extract the contour of the blood vessel from the binary image to obtain a two-dimensional blood vessel image sequence containing the main contour of the blood vessel.

4. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The converting the two-dimensional blood vessel image sequence into a three-dimensional blood vessel model through a multi-view image reconstruction algorithm comprises: Extracting feature points of blood vessels at multiple viewing angles from the two-dimensional blood vessel image sequence to represent key feature areas of the blood vessels; According to the feature points extracted from each perspective, the beam normal vector optimization algorithm is applied to project the two-dimensional feature points into three-dimensional space to optimize the three-dimensional point cloud coordinates. The three-dimensional point cloud coordinates conform to the projections in all viewing angles, and the optimization goal is to minimize the error function Among them, n is the total number of feature points, m is the total number of viewing angles, and e i =||p i -p′ i || 2 =(x i -x′ i ) 2 +(y i -y′ i ) 2 is the projection error for each feature point, p i =(x i ,y i ) is the two-dimensional coordinate of the feature point projected by the blood vessel in the jth viewing angle, P j is the camera projection matrix that represents the mapping relationship from the three-dimensional space to the two-dimensional image plane, P i =(X i ,Y i ,Z i ) is the three-dimensional coordinate point corresponding to each two-dimensional feature point, p′ i =(x′ i ,y′ i ) is a three-dimensional point P i The projection position at the jth viewing angle; The three-dimensional point cloud coordinates of each viewing angle are used to generate a three-dimensional point cloud model representing the spatial distribution of blood vessels; The three-dimensional point cloud models of each viewing angle are merged, and surface reconstruction is performed to generate a three-dimensional surface model of the blood vessel.

5. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: In the three-dimensional blood vessel model, based on the deep learning classification network and the image generation adversarial network technology, the opening position and morphological characteristics of the target blood vessel are identified, and the vascular lesion area for locating the ablation target is extracted, including: voxelize the three-dimensional blood vessel model to obtain three-dimensional image data; Designing a deep learning classification network model, wherein the three-dimensional image data is used as an input of the deep learning classification network model to represent the grayscale or intensity information of the blood vessel in the three-dimensional space; Using the labeled data set to train the deep learning classification network model, identifying blood vessel openings and morphological features by optimizing the objective function; The output of the deep learning classification network model is the classification probability of each voxel or blood vessel region, indicating whether the point belongs to the target blood vessel opening region; Extracting blood vessel morphological features from the intermediate feature map and label map of the deep learning classification network model, wherein the blood vessel morphological features include at least one of a blood vessel branching angle, a blood vessel diameter, and a tortuosity; Design a generator and a discriminator of an image generation adversarial network model, wherein the generator receives the three-dimensional image data and generates a vascular lesion area, and the discriminator distinguishes the area generated by the generator from the real vascular lesion area, and outputs the vascular morphology and the spatial distribution of the lesion area; The position of the lesion area in the three-dimensional space is calculated according to the extracted lesion area and the spatial distribution, and the vascular lesion area used for locating the ablation target is obtained in the three-dimensional blood vessel model.

6. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The combining hemodynamic simulation, based on the three-dimensional vascular model and the vascular lesion area, calculating the optimal positioning point coordinates of the cardiovascular balloon, and planning the best path of the cardiovascular balloon from the current position to the optimal positioning point coordinates includes: Establishing hemodynamic simulation equations Among them, ρ is the density of blood, v is the velocity field, e is the pressure, μ is the viscosity of blood, F is the external force, is the gradient operator; According to the three-dimensional blood vessel model and the blood vessel lesion area, a candidate positioning point R is selected, wherein the candidate positioning point R is located at the front end of the blood vessel lesion area or close to the blood vessel lesion area; Calculate the optimal positioning point coordinates of cardiovascular balloons The optimal positioning point is the point where the balloon has the least effect on blood flow when passing through the blood vessel, Q(R) is the flow rate of the candidate positioning point, is the pressure gradient of the candidate positioning point; Establish a blood flow disturbance model during balloon movement. For any path point r u , calculate blood flow disturbance Among them, v u is the path point r u The blood flow velocity at is the velocity of the balloon, is the path point r u The pressure gradient at To minimize the blood flow disturbance δQ u and the optimal path length Distance(r u-1 ,r u ) Establish an optimization function for the target Among them, L path is the total path cost, which indicates the total cost of the balloon from the current position to the optimal positioning point, and w1 is the distance weight coefficient; Solve the optimization function and calculate the total cost through the path selection algorithm. The cost update formula is C(r v )=min(C(r v ),C(r u )+δQ u +w2 Distance(r u ,r v )), where r v is the path point r u The adjacent path nodes, C(r v ) is r v The cost of the node, Distance(r u ,r v ) is the node r v and the path point r u The geometric distance between them represents the physical distance of the balloon from one node to another node, w2 is the distance weight coefficient, and the path with the minimum cost is selected as the final path to obtain the optimal path of the cardiovascular balloon from the current position to the coordinates of the optimal positioning point.

7. The visual positioning and ablation method for cardiovascular balloon according to claim 6, characterized in that: After obtaining the optimal path, the optimal path is smoothed to obtain the smoothed optimal path. Among them, is the B-spline basis function, r u is the path point r u control point.

8. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The steps of acquiring the real-time position information of the cardiovascular balloon by electromagnetic positioning, calculating the relative spatial relationship between the cardiovascular balloon and the optimal path, and adjusting the position of the cardiovascular balloon by using a navigation algorithm include: Embed a micro sensor in the balloon catheter to detect electromagnetic field signals and position information in real time and send them outward; Receive the electromagnetic field signal and the position information, and calculate the spatial coordinates P of the cardiovascular balloon sensor =(X sensor ,Y sensor ,Z sensor ) and direction angle θ sensor ; Calculating a position error ΔP(t) and a direction error Δθ between the current position of the cardiovascular balloon and a target path point; Calculate the adjustment amount of the direction and movement speed of the cardiovascular balloon, and use the proportional integral differential control algorithm to obtain an adjustment control signal for controlling the movement of the balloon, wherein the formula of the proportional integral differential control algorithm is: u(t) is the control signal, K p is the proportionality coefficient, K d is the differential coefficient, K i is the integral gain coefficient, ΔP(t) is the position error, is the rate of change of position error.

9. The visual positioning and ablation method for cardiovascular balloon according to claim 1, characterized in that: The collecting of real-time parameters during the ablation process, monitoring and real-time adjustment and control of the ablation parameters include: Collect temperature, pressure and energy transmission during ablation; Real-time monitoring of whether the temperature reaches the preset ablation range, whether the balloon pressure exceeds the normal range, and whether the energy transfer rate exceeds the normal range; Through real-time monitoring of various parameters, the control signal is adjusted based on the preset threshold to control the ablation strategy.

10. A visual positioning and ablation system for cardiovascular balloons, characterized in that: include: An image acquisition module, used to acquire real-time image sequences by using a multi-channel imaging system and a synchronous multi-view image acquisition technology; A two-dimensional extraction module, used for extracting the main contour of the blood vessel from the real-time image sequence using an edge detection algorithm to obtain a two-dimensional blood vessel image sequence; A three-dimensional modeling module, which converts the two-dimensional vascular image sequence into a three-dimensional vascular model through a multi-view image reconstruction algorithm; A vascular lesion region extraction module is used to identify the opening position and morphological characteristics of the target blood vessel in the three-dimensional blood vessel model based on a deep learning classification network and an image generation adversarial network technology, and extract the vascular lesion region for locating the ablation target; An optimal path planning module, for calculating the optimal positioning point coordinates of the cardiovascular balloon based on the three-dimensional vascular model and the vascular lesion area in combination with hemodynamic simulation, and planning the optimal path of the cardiovascular balloon from the current position to the optimal positioning point coordinates; A navigation adjustment module, used to obtain real-time position information of the balloon catheter through electromagnetic positioning, calculate the relative spatial relationship between the cardiovascular balloon and the optimal path, and adjust the position of the cardiovascular balloon using a navigation algorithm; The parameter adjustment module is used to collect real-time parameters during the ablation process, monitor the ablation parameters and adjust and control them in real time.

Images