Cardiovascular image three-dimensional reconstruction method and system

Abstract

The application provides a cardiovascular image three-dimensional reconstruction method and system, a cross-sectional image sequence of a blood vessel is continuously collected, boundary detection is performed on the cross-sectional image sequence, a closed contour of a blood vessel lumen in each image frame is obtained, and a three-dimensional point cloud is constructed through the closed contour; spatial curve fitting is performed on the three-dimensional point cloud, a center line path of the blood vessel of a patient is obtained; an inherent anatomical feature of a blood vessel wall is determined, and an axial rotation offset amount of an ultrasound catheter when being pulled back in different image frames is quantified according to a relative displacement of the inherent anatomical feature; the closed contour of the corresponding frame is rotated and transformed around the center of the catheter according to the axial rotation offset amount corresponding to each image frame, the geometric center of the correction result is spatially aligned with the center line path, and a three-dimensional cardiovascular model of the patient is reconstructed. By adopting the scheme, geometric distortion caused by actual movement of the catheter in a complex blood vessel environment can be reduced.

Description

Technical Field

[0001] This application relates to the field of three-dimensional reconstruction technology, and more specifically, to a method and system for three-dimensional reconstruction of cardiovascular images. Background Technology

[0002] 3D reconstruction refers to the process of transforming discrete visual observations into three-dimensional digital entities with precise geometric properties and complete topological structures based on two-dimensional images or sequence data collected from real-world scenes through key technical steps such as feature perception, spatial calculation, and model reconstruction. It realizes a systematic digital twin of physical objects from surface texture to three-dimensional form, and ultimately constructs a high-fidelity computable model that can be used for measurement analysis, virtual simulation, and immersive interaction.

[0003] Cardiovascular 3D reconstruction refers to the process of transforming the original 2D tomographic sequences acquired by medical imaging equipment into a 3D luminal network model with anatomical accuracy and physiological characteristics through computational processes such as intelligent segmentation of vascular cavities, spatial registration of angiographic data, and fusion of hemodynamic features. This process overcomes the limitations of traditional 2D images in representing the spatial conformation of tortuous blood vessels, achieving a systematic digital mapping of cardiac structure and vascular trees from morphology to function. Traditional 3D reconstruction methods for cardiovascular imaging often rely on the assumption of uniform catheter retraction and fixed orientation, which makes it difficult to effectively compensate for the geometric distortion caused by the actual movement of the catheter in complex vascular environments. This results in spatial adaptation deviations between the reconstructed model and the real anatomical structure. For example, traditional methods may directly stack the rotational misalignment contour caused by catheter twisting along the path coordinates. The twisted luminal structure formed by this rotational offset cannot be corrected in the correspondence of anatomical features in adjacent frames, ultimately forming a spiral twist or branch vessel misalignment in the 3D model that does not conform to the actual blood vessel orientation, thus leading to geometric distortion problems in the reconstructed model. Therefore, how to reduce the geometric distortion caused by the actual movement of the catheter in complex vascular environments has become a challenge for the industry. Summary of the Invention

[0004] This application provides a method and system for three-dimensional reconstruction of cardiovascular images, which can reduce geometric distortion caused by the actual movement of catheters in complex vascular environments.

[0005] In a first aspect, this application provides a method for three-dimensional reconstruction of cardiovascular images, comprising the following steps: During the retraction of the ultrasound catheter within the patient's blood vessel, a series of cross-sectional images of the blood vessel are continuously acquired, and the initial spatial path coordinates of the catheter are recorded. Boundary detection is performed on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, and then the three-dimensional point cloud of the patient's blood vessel is constructed by the closed contour and the initial spatial path coordinates. The three-dimensional point cloud is fitted with spatial curves using the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. The inherent anatomical features of the vessel wall between adjacent frames in the cross-sectional image sequence are determined, and then the axial rotational offset of the ultrasound catheter during pull-back in different image frames is quantified based on the relative displacement of the inherent anatomical features. Based on the axial rotation offset of each image frame, the closed contour of the corresponding frame is rotated and corrected around the center of the catheter. The geometric center of the correction result is spatially aligned with the centerline path to reconstruct the three-dimensional cardiovascular model of the patient.

[0006] In some embodiments, performing boundary detection on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame specifically includes: The cross-sectional image sequence is preprocessed to enhance the edge features of the blood vessel wall; The edge detection operator is used to extract the boundary pixels of the blood vessel lumen based on the preprocessed image data; The boundary pixels are connected in an orderly manner to form a closed outline of the blood vessel lumen.

[0007] In some embodiments, constructing a three-dimensional point cloud of the patient's blood vessels using the closed contour and the initial spatial path coordinates specifically includes: Determine the three-dimensional reference position of each closed contour from the initial spatial path coordinates; The two-dimensional pixel coordinates of each closed contour are transformed to a three-dimensional spatial coordinate system using the three-dimensional reference position, thereby constructing a three-dimensional point cloud of the patient's blood vessels.

[0008] In some embodiments, the three-dimensional point cloud is fitted with spatial curves using the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. Specifically, this includes: The three-dimensional point cloud is subjected to spatial sampling processing to obtain a downsampled set of blood vessel spatial points; A reference blood vessel centerline is generated based on the aforementioned blood vessel spatial point set; The centerline of the reference blood vessel is optimized using the morphological characteristics of the blood vessel as a geometric constraint to obtain the centerline path of the patient's blood vessel.

[0009] In some embodiments, quantifying the axial rotational offset of the ultrasound catheter during retraction in different image frames based on the relative displacement of the inherent anatomical features specifically includes: Extract the inherent anatomical feature points of each image frame from the cross-sectional image sequence; Determine the relative displacement between the inherent anatomical feature points between adjacent frame images; The axial rotation offset during ultrasound catheter pullback in different image frames is determined based on all relative displacements.

[0010] In some embodiments, the rotational transformation correction of the closed contour of the corresponding image frame around the center of the conduit based on the axial rotation offset of each image frame specifically includes: Establish a rotation transformation matrix with the center of the catheter as the rotation reference; The closed contour of the corresponding frame is rotated using the rotation transformation matrix and the axial rotation offset corresponding to each image frame.

[0011] In some embodiments, spatially aligning the geometric center of the correction result with the centerline path to reconstruct a three-dimensional cardiovascular model of the patient specifically includes: Determine the geometric center of the correction result; Register all geometric centers to the corresponding positions on the centerline path to obtain the registered contour sequence; A three-dimensional cardiovascular model of the patient was reconstructed based on the registered contour sequence.

[0012] Secondly, this application provides a three-dimensional reconstruction system for cardiovascular imaging, comprising: The acquisition module is used to continuously acquire a sequence of cross-sectional images of the blood vessel during the retraction of the ultrasound catheter within the patient's blood vessel, and to record the initial spatial path coordinates of the catheter. The processing module is used to perform boundary detection on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, and then construct a three-dimensional point cloud of the patient's blood vessel through the closed contour and the initial spatial path coordinates. The processing module is also used to perform spatial curve fitting on the three-dimensional point cloud with the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. The processing module is also used to determine the inherent anatomical features of the vessel wall between adjacent frames in the cross-sectional image sequence, and then quantify the axial rotation offset of the ultrasound catheter during pull-back in different image frames based on the relative displacement of the inherent anatomical features. The execution module is used to perform rotational transformation correction on the closed contour of the corresponding frame around the catheter center according to the axial rotation offset of each image frame, and spatially align the geometric center of the correction result with the centerline path to reconstruct the three-dimensional cardiovascular model of the patient.

[0013] Thirdly, this application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described cardiovascular image three-dimensional reconstruction method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described cardiovascular image three-dimensional reconstruction method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The cardiovascular imaging three-dimensional reconstruction method and system provided in this application obtains the centerline path of the patient's blood vessels by fitting a spatial curve to a three-dimensional point cloud with vascular morphology features as geometric constraints; a centerline model conforming to the physiological curvature characteristics of blood vessels is established, and the path coordinate offset caused by catheter wall-attachment is corrected by geometric constraints, providing an accurate spatial reference for subsequent contour alignment; thus, it provides an anatomical rationale for centerline positioning; subsequently, by determining the inherent anatomical features of the blood vessel wall between adjacent frames in the cross-sectional image sequence, the axial rotation offset of the ultrasound catheter during retraction in different image frames is quantified based on the relative displacement of the inherent anatomical features; this process can accurately identify the actual axial rotation of the catheter in complex vascular environments, and dynamically track the real-time torsional state of the catheter during retraction by the relative positional changes of inherent features such as calcified plaques and branch openings between adjacent frames; furthermore, in the three-dimensional model reconstruction process, the axial rotation offset corresponding to each image frame is used to reconstruct the model. The closed contour of the catheter frame is rotated and corrected around the catheter center, and the precisely quantified rotation compensation parameters are applied to each cross-sectional contour to achieve contour-level spatial attitude correction. This process dynamically corrects the orientation deviation of each cross-section caused by catheter torsion through rotation transformation, providing contour data with rotational consistency verification for the final model reconstruction. Finally, the geometric center of the correction result is spatially aligned with the centerline path to ensure that the rotated and corrected vessel cross-sections are accurately arranged along the centerline path that conforms to anatomical reality, achieving continuous topological reconstruction of the vessel lumen in three-dimensional space. In other words, through spatial curve fitting of the centerline path and dynamic quantification correction of axial rotation offset, a dual compensation mechanism that can adapt to the complex motion state of the catheter is constructed, thereby effectively suppressing geometric distortions such as vessel lumen distortion and branch misalignment caused by the combined movement of the catheter against the wall and axial torsion. In summary, this scheme can reduce the geometric distortion caused by the actual movement of the catheter in complex vascular environments. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a three-dimensional reconstruction method for cardiovascular images according to some embodiments of this application; Figure 2 This is a schematic flowchart illustrating the determination of a closed profile according to some embodiments of this application; Figure 3 This is a schematic diagram of the process for generating a reference blood vessel centerline according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a three-dimensional reconstruction system for cardiovascular imaging according to some embodiments of this application; Figure 5 This is an internal structural diagram of a computer device for implementing a three-dimensional reconstruction method for cardiovascular images, according to some embodiments of this application. Detailed Implementation

[0017] To better understand the technical solutions in this embodiment, the technical solutions in this embodiment will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0018] refer to Figure 1 The figure is a flowchart illustrating a three-dimensional reconstruction method for cardiovascular images according to some embodiments of this application. The three-dimensional reconstruction method for cardiovascular images mainly includes the following steps: In step 101, during the retraction of the ultrasound catheter within the patient's blood vessel, a sequence of cross-sectional images of the blood vessel is continuously acquired, and the initial spatial path coordinates of the catheter are recorded.

[0019] In practice, a mechanically rotating intravascular ultrasound catheter system can be used, with an ultrasound transducer integrated at its tip. The transducer is driven by a drive mechanism inside the catheter to perform a circular scan. During the process of the catheter being pulled back at a preset speed along the blood vessel lumen by an automatic retraction system, the ultrasound transducer continuously emits ultrasound pulses at a preset sampling frequency and receives echo signals reflected from the blood vessel wall. After signal preprocessing, the signals are transmitted to an external image processing system, where a series of cross-sectional grayscale images are formed through digital scanning conversion and reconstruction, constituting the cross-sectional image sequence. Simultaneously, a two-dimensional perspective image sequence containing the catheter marker loop is acquired using the dual-plane imaging device of the angiography system. Based on the principle of stereo vision, the position of the catheter marker loop in three-dimensional space is reconstructed. Combined with the time synchronization information during the retraction process, the initial spatial path coordinates consisting of a series of three-dimensional coordinate points are generated.

[0020] It should be noted that, in this application, the ultrasound catheter retraction refers to the data acquisition operation in which, after the intravascular ultrasound catheter is delivered to the target vascular segment during interventional treatment, the catheter is moved along the vascular axis by an electric retraction device; the cross-sectional image sequence refers to a series of two-dimensional ultrasound cross-sectional images acquired along the vertical direction of the long axis of the vessel, and its pixel grayscale value characterizes the distribution of ultrasound echo intensity of the vascular tissue; the initial spatial path coordinates refer to the approximate value of the catheter's motion trajectory in three-dimensional space obtained based on the reconstruction of dual-plane perspective images, and this coordinate sequence contains three-dimensional position information with time stamps; the mechanically rotating intravascular ultrasound catheter system includes an ultrasound transducer, a drive mechanism, a retraction device, and an image processing system; the dual-plane imaging device refers to a combination of devices that simultaneously image the same object using two X-ray sources and detectors at different angles.

[0021] In step 102, boundary detection is performed on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, and then the three-dimensional point cloud of the patient's blood vessel is constructed by the closed contour and the initial spatial path coordinates.

[0022] In some embodiments, reference Figure 2 As shown in the figure, this is a schematic flowchart illustrating the process of determining a closed contour in some embodiments of this application. Boundary detection of the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame can be achieved through the following steps: First, in step 1021, the cross-sectional image sequence is preprocessed to enhance the edge features of the blood vessel wall; Then, in step 1022, the boundary pixels of the blood vessel lumen are extracted based on the preprocessed image data using an edge detection operator; Finally, in step 1023, the boundary pixels are connected in an orderly manner to form a closed contour of the blood vessel lumen.

[0023] In specific implementation, image preprocessing of the cross-sectional image sequence to enhance the edge features of the blood vessel wall can be achieved in the following ways: First, median filtering is performed on the cross-sectional image sequence. The median of the gray values ​​of all pixels within a predetermined neighborhood window of each pixel is used to replace the value of the center pixel, thereby suppressing the inherent speckle noise of the ultrasound image. Next, a contrast-limited adaptive histogram equalization method is used. Each frame of the image is divided into multiple rectangular sub-regions of the same size. A gray-level histogram is independently calculated and histogram equalization is performed in each sub-region. At the same time, the height of the histogram is constrained by a preset contrast limit value to prevent local noise from being excessively enhanced, thereby improving the local contrast at the junction of the blood vessel wall and the lumen. Finally, anisotropic diffusion filtering is performed. A diffusion coefficient function is constructed based on the image gradient information. A greater degree of diffusion smoothing is implemented in flat areas with small image gradients, while a smaller degree of diffusion is maintained in blood vessel wall edge areas with large gradients. Finally, a preprocessed image sequence with effectively suppressed noise and enhanced blood vessel wall edge features is obtained as the preprocessed image data. The contrast limit value can be set according to the image noise level, and this application does not limit it.

[0024] In specific implementation, extracting the boundary pixels of blood vessel lumens using an edge detection operator based on preprocessed image data can be achieved in the following way: For example, in each frame of the preprocessed image, the Canny edge detection operator is applied. First, a Gaussian filter with a predetermined standard deviation is used to convolve the image to achieve smoothing. Then, the Sobel operator is used to calculate the gradient components of each pixel in the horizontal and vertical directions, thereby obtaining the gradient magnitude and direction. Then, non-maximum suppression is performed, comparing the gradient magnitude of each pixel with the gradient magnitudes of its two adjacent pixels along the gradient direction. The process involves comparing the gradient values ​​and retaining only local maxima. A dual-threshold hysteresis connection mechanism is then employed, setting two predetermined thresholds: a high threshold and a low threshold. Pixels with gradient magnitudes higher than the high threshold are marked as strong edge points, while those lower than the low threshold are directly excluded. Pixels in between are only retained as weak edge points when adjacent to strong edge points. Finally, the set of all pixels identified as strong edge points and those meeting the criteria as weak edge points is used as the boundary pixels of the blood vessel lumen. The ratio of the high threshold to the low threshold can be set according to the continuity requirements of the blood vessel wall edge, and this application does not impose any limitations on this.

[0025] In specific implementation, the orderly connection of the boundary pixels to form the closed contour of the blood vessel lumen can be achieved in the following way: First, select an unvisited boundary point from the set of boundary pixels as the starting point; then, according to the 8-connected neighborhood principle, check the positions of the eight adjacent pixels of the current boundary point in turn. If there is an unvisited boundary point, add it to the current contour chain and update the current position; repeat this search and connection process until the tracking path returns to the starting point, forming a closed contour; for cases with multiple unvisited adjacent boundary points encountered during the tracking process, according to the leftmost rule in the prior art, the first boundary point encountered in the clockwise direction from the current pixel position is selected as the next tracking direction. This rule ensures that each independent closed contour can be formed at the branch of the blood vessel lumen; after completing the current contour tracking, continue to select a new starting point from the remaining unvisited boundary points and repeat the above tracking process until all boundary points are visited; finally, the closed contour of the blood vessel lumen is generated; each closed contour is composed of an orderly connected sequence of boundary pixels, ensuring that each closed contour is a simply connected region in the topological structure.

[0026] It should be noted that the closed contour of the blood vessel lumen described in this application refers to a continuous closed curve used to characterize the geometric shape of the boundary of the inner wall of the blood vessel in a single cross-sectional image.

[0027] In some embodiments, the construction of a three-dimensional point cloud of the patient's blood vessels using the closed contour and the initial spatial path coordinates can be achieved through the following steps: Determine the three-dimensional reference position of each closed contour from the initial spatial path coordinates; The two-dimensional pixel coordinates of each closed contour are transformed to a three-dimensional spatial coordinate system using the three-dimensional reference position, thereby constructing a three-dimensional point cloud of the patient's blood vessels.

[0028] In specific implementation, the determination of the three-dimensional reference position of each closed contour from the initial spatial path coordinates can be achieved in the following ways: First, extract the acquisition timestamp of the image frame corresponding to each closed contour and accurately match it with the timestamp recorded in the initial spatial path coordinate sequence; when the acquisition timestamp of the closed contour is completely consistent with a certain timestamp in the initial spatial path coordinate sequence, the corresponding three-dimensional coordinate is directly used as the three-dimensional reference position of the closed contour; when the acquisition timestamp of the closed contour is between two adjacent timestamps in the initial spatial path coordinate sequence, the three-dimensional reference position is calculated using a linear interpolation method, that is, the X, Y, and Z components of two adjacent three-dimensional coordinate points are interpolated according to the proportional relationship of the timestamps, thereby obtaining the three-dimensional reference position of each closed contour; as a preferred embodiment, for closed contours with missing path coordinates in the pull-back start or end stage, the endpoint extrapolation method can be used to supplement its three-dimensional reference position based on the movement trend of existing path points, ensuring that each closed contour obtains a corresponding three-dimensional spatial positioning reference, which is not limited in this application.

[0029] It should be noted that the three-dimensional reference position of the closed contour mentioned in this application refers to the spatial reference point used to locate the blood vessel contour in the two-dimensional ultrasound image to the three-dimensional spatial coordinate system, and to establish the coordinate mapping relationship between the two-dimensional image data and the three-dimensional anatomical space.

[0030] In specific implementation, the transformation of the two-dimensional pixel coordinates of each closed contour to a three-dimensional spatial coordinate system through the three-dimensional reference position to construct the three-dimensional point cloud of the patient's blood vessels can be achieved in the following way: taking the three-dimensional reference position of each closed contour as the reference point, and combining the pixel spacing calibration parameters of the ultrasound image, the two-dimensional image coordinates of each boundary pixel on the contour are transformed into three-dimensional spatial coordinates; specifically, a two-dimensional coordinate system is established with the image center as the origin, and the pixel coordinates are converted into physical size offsets according to the pixel spacing, and then the offsets are superimposed on the three-dimensional reference position through coordinate transformation; the boundary pixels of all closed contours are traversed, and the above coordinate transformation process is repeated to map all two-dimensional contour points to a unified world coordinate system; finally, the generated three-dimensional spatial point set is deduplicated to remove duplicate points caused by overlapping of adjacent contours or interpolation errors, thereby generating the constructed three-dimensional point cloud of the patient's blood vessels, in which each three-dimensional point retains its original image grayscale information and spatial topological relationship.

[0031] It should be noted that the three-dimensional point cloud mentioned in this application refers to a set of discrete points used to characterize the geometric shape of the inner wall surface of blood vessels in three-dimensional space, and is used to provide the original spatial data basis for three-dimensional reconstruction of the blood vessel lumen.

[0032] In step 103, the three-dimensional point cloud is fitted with spatial curves using the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels.

[0033] In some embodiments, the process of fitting the three-dimensional point cloud with the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels can be achieved through the following steps: The three-dimensional point cloud is subjected to spatial sampling processing to obtain a downsampled set of blood vessel spatial points; A reference blood vessel centerline is generated based on the aforementioned blood vessel spatial point set; The centerline of the reference blood vessel is optimized using the morphological characteristics of the blood vessel as a geometric constraint to obtain the centerline path of the patient's blood vessel.

[0034] In specific implementation, spatial sampling processing is performed on the three-dimensional point cloud to obtain a downsampled vascular spatial point set. This can be achieved in the following way: First, the three-dimensional point cloud is voxelized, dividing the three-dimensional space into a uniform cubic mesh. The voxel size can be set according to the point cloud density. Smaller voxel sizes are used in high-density areas to preserve details, while larger voxel sizes are used in low-density areas to improve processing efficiency. Within each voxel, the distance from all points to the voxel center is calculated, and the closest point is retained as a representative point, while other points within the voxel are removed. Then, curvature-based sampling is performed on the retained point set. By calculating the curvature features of local regions of the point cloud, more points are retained in the curved areas of the blood vessels with greater curvature, and fewer points are retained in the straight segments of the blood vessels with less curvature. Finally, a downsampled vascular spatial point set is obtained while maintaining the geometric features of the blood vessels.

[0035] It should be noted that the vascular spatial point set mentioned in this application refers to a three-dimensional spatial point set used to characterize the key features of vascular morphology after downsampling, which is used to improve the efficiency and stability of subsequent centerline calculation while maintaining the geometric features of the blood vessels.

[0036] For specific implementation, refer to Figure 3As shown in the figure, this is a schematic diagram of the process for generating a reference vascular centerline according to some embodiments of this application. The generation of the reference vascular centerline based on the vascular spatial point set can be achieved in the following way: First, the Euclidean distance between any two points in the vascular spatial point set is calculated, and the two points with the greatest distance are selected as the start and end points of the vascular system; then, an adjacency graph is constructed using the vascular spatial point set, where each point is a graph node, and the connection relationship between points is determined by a preset connection distance threshold; the Dijkstra algorithm is used to calculate the shortest path from the start point to the end point in the adjacency graph, and the shortest path is used as the reference vascular centerline; for vascular structures with branches, the shortest path to the end of each branch is calculated at the branch point, and the complete reference vascular centerline containing the branch structure is formed by connecting the paths.

[0037] It should be noted that the reference vascular centerline mentioned in this application refers to a three-dimensional spatial curve used to preliminarily describe the direction of the center of the vascular anatomy.

[0038] In specific implementation, the centerline of the reference blood vessel is optimized using the morphological characteristics of the blood vessel as geometric constraints. The resulting centerline path of the patient's blood vessel can be achieved as follows: the reference blood vessel centerline is represented as a parameterized B-spline curve, with the continuity and smoothness of the blood vessel as geometric constraints. The continuity constraint is achieved by ensuring the continuity of the first derivative of adjacent B-spline curve segments at the connection points, and the smoothness constraint is achieved by minimizing the curvature change of the entire B-spline curve. Key center points are determined based on the geometric distribution characteristics of the blood vessel spatial point set, including blood vessel bifurcation points identified by analyzing the point set density distribution, and blood vessel bending feature points determined by calculating local curvature extrema. A system containing numerous... The objective function is optimized by combining the data fitting term and the geometric constraint term. The data fitting term constrains the distance deviation between the optimized B-spline curve and the key center point, while the geometric constraint term ensures smoothness by applying a regularization penalty to the second derivative norm of the B-spline curve. The coordinates of the control points of the B-spline curve are iteratively adjusted using a gradient descent algorithm. In each iteration, the gradient vector of the objective function with respect to the coordinates of each control point is calculated, and the position of the control points is updated along the negative gradient direction until the change in the objective function value is less than a preset convergence threshold. The finally optimized B-spline curve is used as the centerline path of the patient's blood vessels. This path maintains geometric consistency with the original blood vessel spatial point set while satisfying the requirements of continuity and smoothness of the vascular anatomy.

[0039] It should be noted that the centerline path of the patient's blood vessels mentioned in this application refers to an optimized three-dimensional spatial curve used to accurately describe the direction of the center of the vascular anatomy.

[0040] In step 104, the inherent anatomical features of the vessel wall between adjacent frames in the cross-sectional image sequence are determined, and then the axial rotation offset of the ultrasound catheter during pullback in different image frames is quantified based on the relative displacement of the inherent anatomical features.

[0041] It should be noted that the inherent anatomical features of the blood vessel wall mentioned in this application refer to anatomical structures in the blood vessel wall tissue that have a stable morphology and can be continuously observed in consecutive image frames, including calcified plaques, fibrous plaque boundaries, vascular branch openings, and the intima-vessel interface. These inherent anatomical features can be identified from the cross-sectional image sequence using image feature extraction algorithms, specifically including plaque feature detection based on local image texture analysis, vascular branch opening localization based on morphological processing, and vascular wall interface identification based on grayscale gradient changes. In specific implementations, scale-invariant feature transformation algorithms or accelerated robust feature algorithms can be used to extract the coordinate positions and features of the inherent anatomical features. In another embodiment, the location coordinates and feature vectors of the inherent anatomical features of the blood vessel wall can be directly output through a feature point detection network in deep learning. The extraction of the inherent anatomical features can be performed after image preprocessing to ensure the accuracy of feature detection. In some embodiments, the extraction density of the inherent anatomical features can be adjusted according to the degree of vascular lesions. For example, more feature points can be extracted in plaque-rich areas to improve subsequent matching accuracy, while the number of feature points can be appropriately reduced in relatively smooth areas of the blood vessel wall to improve computational efficiency. In other embodiments, the feature extraction strategy can also be dynamically adjusted according to the clinical importance of the blood vessel segment, which is not limited in this application.

[0042] In some embodiments, quantifying the axial rotational offset of the ultrasound catheter during retraction in different image frames based on the relative displacement of the inherent anatomical features can be achieved using the following steps: Extract the inherent anatomical feature points of each image frame from the cross-sectional image sequence; Determine the relative displacement between the inherent anatomical feature points between adjacent frame images; The axial rotation offset during ultrasound catheter pullback in different image frames is determined based on all relative displacements.

[0043] In specific implementation, the extraction of inherent anatomical feature points from each image frame in the cross-sectional image sequence can be achieved in the following way: First, a Gaussian pyramid is constructed for each preprocessed cross-sectional image frame. By detecting the extreme points of the Gaussian difference function at different scales, the potential locations and scales of inherent anatomical features such as calcified plaques, fibrous plaque boundaries, and vascular branch openings in the vessel wall region are determined. Then, a gradient direction histogram is calculated in the scale-space neighborhood of each feature point, and the peak direction of the histogram is taken as the principal direction of the feature point to ensure the rotational invariance of the feature description. Finally, within the neighborhood aligned with the principal direction of the feature point... The image is divided into sub-regions, and the gradient direction statistical histogram of sampling points in each sub-region is calculated to generate feature descriptor vectors with scale and rotation invariance, thereby completing the extraction of inherent anatomical feature points of the blood vessel wall in each image frame. In a preferred embodiment, a preset contrast threshold can be applied to the response value of the difference of Gaussian function to exclude unstable feature points in low-contrast regions, thereby improving the repeatability and robustness of the extracted feature points. In other embodiments, a feature point detection network based on deep learning can also be used to directly regress the coordinate positions and feature descriptor vectors of inherent anatomical feature points in the image, which is not limited in this application.

[0044] In specific implementation, determining the relative displacement between the inherent anatomical feature points between adjacent frame images can be achieved in the following ways: for example, by calculating the Euclidean distance between feature descriptor vectors to evaluate the similarity of feature points, and screening initial matching pairs based on a preset matching distance threshold; then, optimizing the initial matching pairs by applying a random sampling consensus algorithm, estimating the fundamental matrix transformation model by iteratively sampling the minimum point set, and calculating the distance from all matching point pairs to their corresponding epipolar lines, marking point pairs with distances less than the preset consensus threshold as inliers, and removing mismatched point pairs that do not conform to the epipolar geometric constraints; finally, based on the retained correct matching point pairs, calculating the pixel coordinate displacement vector of each matching feature point between adjacent frame images to form a set of displacement vectors describing the relative displacement of feature points between adjacent frames; wherein, the preset consensus threshold can be adjusted according to the image resolution and feature point positioning accuracy, and this application does not limit it in this regard.

[0045] It should be noted that the relative displacement mentioned in this application refers to a two-dimensional vector quantity used to characterize the positional changes of inherent anatomical features between adjacent image frames, and is used to reflect the relative motion state of the ultrasound catheter during the pull-back process.

[0046] In specific implementation, determining the axial rotation offset of the ultrasound catheter during pullback in different image frames based on all relative displacements can be achieved in the following way: For example, a two-dimensional point set correspondence can be constructed based on the coordinate displacement vectors of all correctly matched feature point pairs between adjacent image frames. The least squares estimation algorithm is used to solve for the optimal two-dimensional rigid body transformation matrix that minimizes the sum of squared coordinate transformation errors of all matched point pairs. The rigid body transformation matrix contains both rotation and translation components. Then, the rotation angle parameter and the relative rotation offset between adjacent image frames are extracted from the rigid body transformation matrix. Next, the axial rotation offset of each image frame in the cross-sectional image sequence relative to the initial reference frame is obtained by accumulating the relative rotation offset frame by frame starting from the initial reference frame. Finally, a sliding window averaging filter is used to smooth the absolute rotation offset sequence to eliminate random fluctuations caused by feature point matching errors. The size of the sliding window can be set according to the pullback speed and image sampling frequency. If the size is too small, the smoothing effect may be insufficient; if the size is too large, it may mask the true trend of catheter rotation. This application does not limit this.

[0047] It should be noted that the axial rotation offset mentioned in this application refers to a parameter used to quantify the angle of rotation of the ultrasound catheter around its axis, and is used to compensate for image orientation deviation caused by catheter torsion.

[0048] In step 105, the closed contour of the corresponding frame is rotated and corrected around the center of the catheter according to the axial rotation offset of each image frame. The geometric center of the correction result is spatially aligned with the centerline path to reconstruct the three-dimensional cardiovascular model of the patient.

[0049] In some embodiments, the rotational transformation correction of the closed contour of the corresponding frame around the center of the conduit based on the axial rotation offset corresponding to each image frame can be achieved by the following steps: Establish a rotation transformation matrix with the center of the catheter as the rotation reference; The closed contour of the corresponding frame is rotated using the rotation transformation matrix and the axial rotation offset corresponding to each image frame.

[0050] In specific implementation, the rotation transformation matrix established with the catheter center as the rotation reference can be implemented in the following way, for example: First, a two-dimensional coordinate system is established with the center point of each image frame as the catheter center position, and this center point corresponds to the actual physical position of the ultrasonic transducer; then, according to the axial rotation offset corresponding to each image frame, a two-dimensional rigid body transformation matrix containing rotation and translation components is constructed, wherein the rotation component is determined by the rotation offset, and the translation component ensures that the catheter center position remains unchanged before and after rotation; the construction of the rotation transformation matrix is ​​realized by the standard two-dimensional rigid body transformation formula, converting the rotation angle into rotation matrix elements, and combining the translation components to form a complete transformation matrix; wherein, the sign convention of the rotation angle follows the right-hand rule to ensure that the rotation direction is consistent with the actual rotation direction of the catheter; as a preferred embodiment, coordinate normalization processing can be added during the construction of the transformation matrix to eliminate the influence of image resolution differences on transformation accuracy; in other embodiments, homogeneous coordinate representation can also be used to simplify matrix operations, and this application does not limit this.

[0051] It should be noted that the rotation transformation matrix mentioned in this application refers to a mathematical tool used to realize the rotation transformation of geometric coordinates in a two-dimensional image plane.

[0052] In specific implementation, the rotation transformation of the closed contour of the corresponding frame using the rotation transformation matrix and the axial rotation offset corresponding to each image frame can be achieved in the following way: First, convert the coordinates of all boundary pixels of each closed contour into homogeneous coordinates so that they can be multiplied by the rotation transformation matrix; then, multiply the homogeneous coordinates of each boundary point by the corresponding rotation transformation matrix to achieve the rotation transformation of the point around the center of the conduit, generating new coordinates after rotation correction; next, repeat the above matrix multiplication operation for all boundary points to complete the rotation transformation of the entire closed contour; finally, convert the transformed homogeneous coordinates back to standard two-dimensional coordinates to obtain the orientation-corrected closed contour as the correction result; wherein, as a preferred embodiment, a bilinear interpolation method can be used to process non-integer coordinate positions during the rotation transformation process to maintain the smoothness of the contour; in other embodiments, the parallel computing capability of the graphics processor can also be used to accelerate the transformation calculation of a large number of coordinate points, which is not limited in this application.

[0053] In some embodiments, spatially aligning the geometric center of the correction result with the centerline path to reconstruct a three-dimensional cardiovascular model of the patient can be achieved through the following steps: Determine the geometric center of the correction result; Register all geometric centers to the corresponding positions on the centerline path to obtain the registered contour sequence; A three-dimensional cardiovascular model of the patient was reconstructed based on the registered contour sequence.

[0054] In specific implementation, the geometric center of the correction result can be determined in the following way, for example: First, for each closed contour (i.e., the correction result) after rotational transformation correction, extract the two-dimensional coordinate set of all boundary pixels of each closed contour; then, by calculating the arithmetic mean of the boundary point coordinates in the X and Y axes, obtain the geometric center coordinates of each closed contour; specifically, traverse the contour boundary point set, accumulate the horizontal and vertical coordinate values ​​of all points and divide by the total number of boundary points to obtain the geometric center representing the spatial position of the contour; wherein, as a preferred embodiment, a weighted average calculation can be performed based on the curvature distribution characteristics of the boundary points, and higher weights can be assigned to areas with higher blood vessel curvature to more accurately reflect the actual geometric center position of the blood vessel lumen; in other embodiments, a calculation method based on shape moments can also be used to obtain the geometric center, which is not limited in this application.

[0055] In specific implementation, all geometric centers are registered to their corresponding positions on the centerline path. The resulting registered contour sequence can be achieved in the following ways: First, based on the temporal information of each closed contour in the acquisition sequence, the corresponding spatial reference point is located on the litigation centerline path. Then, the translation vector from each geometric center to its corresponding centerline reference point is calculated, and the target normal of the contour plane is determined by combining the tangent direction of the centerline at that point. Next, a rigid body transformation is performed on each closed contour to make its geometric center coincide with the centerline reference point, while adjusting the contour direction so that its plane normal is perpendicular to the tangent of the centerline. Finally, spatial registration is completed for all contours in sequence to form a registered contour sequence that is precisely arranged in the three-dimensional coordinate system. In a preferred embodiment, an iterative nearest-point algorithm can be used to optimize the registration process, gradually improving the spatial consistency between the contour and the centerline through multiple iterations. In other embodiments, the registration efficiency can also be improved based on the feature point matching method, which is not limited in this application.

[0056] It should be noted that the registered contour sequence mentioned in this application refers to a set of aligned contours used to accurately characterize the continuous morphology of the vascular lumen in three-dimensional space.

[0057] In specific implementation, the reconstruction of a patient's three-dimensional cardiovascular model based on the registered contour sequence can be achieved in the following ways: First, the registered contour sequence is used as a three-dimensional spatial constraint, and an initial vascular model is constructed using the traveling cube algorithm. A triangular mesh surface is generated by calculating isosurfaces in a three-dimensional voxel mesh. Then, Laplacian smoothing is performed on the initial mesh model, and the positions of the mesh vertices are iteratively adjusted to move them toward the centroid of the neighborhood, eliminating irregular noise on the model surface. Finally, the smoothed mesh is simplified using an edge-folding algorithm, reducing the model complexity while maintaining key geometric features such as vascular branches and bends, and generating the final three-dimensional cardiovascular model. In a preferred embodiment, feature preservation weights can be set during the mesh simplification process to retain more details in vascular regions with large curvature. In other embodiments, the Poisson reconstruction method can also be used to directly generate a continuous surface from the contour point cloud, which is not limited in this application.

[0058] It should be noted that the three-dimensional cardiovascular model described in this application refers to a three-dimensional geometric entity used to completely reproduce the anatomical structure of a patient's blood vessels, and to provide a basis for the visualization and quantitative analysis of blood vessel morphology.

[0059] In another aspect, in some embodiments, this application provides a cardiovascular imaging three-dimensional reconstruction system, with reference to... Figure 4 The figure is a schematic diagram of the structure of a cardiovascular imaging three-dimensional reconstruction system according to some embodiments of this application. The cardiovascular imaging three-dimensional reconstruction system 200 includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to continuously acquire cross-sectional image sequences of blood vessels and record the initial spatial path coordinates of the catheter when pulling back the ultrasound catheter in the patient's blood vessels. Processing module 202, in this application, is mainly used to perform boundary detection on the cross-sectional image sequence, obtain the closed contour of the blood vessel lumen in each image frame, and then construct the three-dimensional point cloud of the patient's blood vessel through the closed contour and the initial spatial path coordinates; In addition, the processing module 202 in this application is also used to perform spatial curve fitting on the three-dimensional point cloud with the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. In addition, the processing module 202 in this application is also used to determine the inherent anatomical features of the blood vessel wall between adjacent frames in the cross-sectional image sequence, and then quantify the axial rotation offset of the ultrasound catheter during pull-back in different image frames based on the relative displacement of the inherent anatomical features. The execution module 203 in this application is mainly used to perform rotational transformation correction on the closed contour of the corresponding frame around the center of the catheter according to the axial rotation offset of each image frame, and to spatially align the geometric center of the correction result with the centerline path, thereby reconstructing the three-dimensional cardiovascular model of the patient.

[0060] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described cardiovascular image three-dimensional reconstruction method.

[0061] In some embodiments, reference Figure 5 This figure is an internal structural diagram of a computer device for implementing a three-dimensional reconstruction method for cardiovascular images, according to some embodiments of this application. The three-dimensional reconstruction method for cardiovascular images in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 300 includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0062] The processor 301 may be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the cardiovascular imaging three-dimensional reconstruction method in this application.

[0063] The communication bus 302 is used to transmit information between the aforementioned components.

[0064] Memory 303 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 (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks 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. Memory 303 may exist independently and be connected to processor 301 via communication bus 302. Memory 303 may also be integrated with processor 301.

[0065] The memory 303 stores program code for executing the scheme of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the cardiovascular imaging three-dimensional reconstruction method can be implemented by the processor 301 and one or more software modules in the program code in the memory 303.

[0066] Communication interface 304 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0067] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core processor or a multi-core processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0068] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device may be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0069] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for three-dimensional reconstruction of cardiovascular images.

[0070] In summary, the cardiovascular imaging three-dimensional reconstruction method and system disclosed in this application continuously acquires cross-sectional image sequences of the blood vessels during ultrasound catheter retraction within the patient's blood vessels and records the initial spatial path coordinates of the catheter. Boundary detection is performed on the cross-sectional image sequences to obtain the closed contour of the blood vessel lumen in each image frame. Then, a three-dimensional point cloud of the patient's blood vessels is constructed using the closed contour and the initial spatial path coordinates. The three-dimensional point cloud is fitted with spatial curves using the morphological features of the blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. The inherent anatomical features of the blood vessel walls between adjacent frames in the cross-sectional image sequence are determined, and the axial rotation offset during ultrasound catheter retraction in different image frames is quantified based on the relative displacement of the inherent anatomical features. The closed contour of the corresponding frame is rotated and corrected around the catheter center according to the axial rotation offset corresponding to each image frame. The geometric center of the corrected result is spatially aligned with the centerline path, thereby reconstructing a three-dimensional cardiovascular model of the patient. This reduces geometric distortion caused by the actual movement of the catheter in complex vascular environments.

[0071] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0072] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for three-dimensional reconstruction of cardiovascular images, characterized in that, Includes the following steps: During the retraction of the ultrasound catheter within the patient's blood vessel, a series of cross-sectional images of the blood vessel are continuously acquired, and the initial spatial path coordinates of the catheter are recorded. Boundary detection is performed on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, and then the three-dimensional point cloud of the patient's blood vessel is constructed by the closed contour and the initial spatial path coordinates. The three-dimensional point cloud is fitted with spatial curves using the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. The inherent anatomical features of the vessel wall between adjacent frames in the cross-sectional image sequence are determined, and then the axial rotational offset of the ultrasound catheter during pull-back in different image frames is quantified based on the relative displacement of the inherent anatomical features. Based on the axial rotation offset of each image frame, the closed contour of the corresponding frame is rotated and corrected around the center of the catheter. The geometric center of the correction result is spatially aligned with the centerline path to reconstruct the three-dimensional cardiovascular model of the patient.

2. The method as described in claim 1, characterized in that, Boundary detection is performed on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, specifically including: The cross-sectional image sequence is preprocessed to enhance the edge features of the blood vessel wall; The edge detection operator is used to extract the boundary pixels of the blood vessel lumen based on the preprocessed image data; The boundary pixels are connected in an orderly manner to form a closed outline of the blood vessel lumen.

3. The method as described in claim 1, characterized in that, The construction of the three-dimensional point cloud of the patient's blood vessels using the closed contour and the initial spatial path coordinates specifically includes: Determine the three-dimensional reference position of each closed contour from the initial spatial path coordinates; The two-dimensional pixel coordinates of each closed contour are transformed to a three-dimensional spatial coordinate system using the three-dimensional reference position, thereby constructing a three-dimensional point cloud of the patient's blood vessels.

4. The method as described in claim 1, characterized in that, The three-dimensional point cloud is fitted with spatial curves using the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. Specifically, this includes: The three-dimensional point cloud is subjected to spatial sampling processing to obtain a downsampled set of blood vessel spatial points; A reference blood vessel centerline is generated based on the aforementioned blood vessel spatial point set; The centerline of the reference blood vessel is optimized using the morphological characteristics of the blood vessel as a geometric constraint to obtain the centerline path of the patient's blood vessel.

5. The method as described in claim 1, characterized in that, The axial rotational offset of the ultrasound catheter during pullback in different image frames is specifically quantified based on the relative displacement of the inherent anatomical features, including: Extract the inherent anatomical feature points of each image frame from the cross-sectional image sequence; Determine the relative displacement between the inherent anatomical feature points between adjacent frame images; The axial rotation offset during ultrasound catheter pullback in different image frames is determined based on all relative displacements.

6. The method as described in claim 1, characterized in that, Based on the axial rotation offset of each image frame, the closed contour of the corresponding frame is rotated and corrected around the center of the conduit. Specifically, this includes: Establish a rotation transformation matrix with the center of the catheter as the rotation reference; The closed contour of the corresponding frame is rotated using the rotation transformation matrix and the axial rotation offset corresponding to each image frame.

7. The method as described in claim 1, characterized in that, The geometric center of the correction result is spatially aligned with the centerline path to reconstruct the patient's three-dimensional cardiovascular model. This process specifically includes: Determine the geometric center of the correction result; Register all geometric centers to the corresponding positions on the centerline path to obtain the registered contour sequence; A three-dimensional cardiovascular model of the patient was reconstructed based on the registered contour sequence.

8. A three-dimensional reconstruction system for cardiovascular imaging, characterized in that, include: The acquisition module is used to continuously acquire a sequence of cross-sectional images of the blood vessel during the retraction of the ultrasound catheter within the patient's blood vessel, and to record the initial spatial path coordinates of the catheter. The processing module is used to perform boundary detection on the cross-sectional image sequence to obtain the closed contour of the blood vessel lumen in each image frame, and then construct a three-dimensional point cloud of the patient's blood vessel through the closed contour and the initial spatial path coordinates. The processing module is also used to perform spatial curve fitting on the three-dimensional point cloud with the morphological features of blood vessels as geometric constraints to obtain the centerline path of the patient's blood vessels. The processing module is also used to determine the inherent anatomical features of the vessel wall between adjacent frames in the cross-sectional image sequence, and then quantify the axial rotation offset of the ultrasound catheter during pull-back in different image frames based on the relative displacement of the inherent anatomical features. The execution module is used to perform rotational transformation correction on the closed contour of the corresponding frame around the catheter center according to the axial rotation offset of each image frame, and spatially align the geometric center of the correction result with the centerline path to reconstruct the three-dimensional cardiovascular model of the patient.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the cardiovascular imaging three-dimensional reconstruction method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the cardiovascular imaging three-dimensional reconstruction method as described in any one of claims 1 to 7.

Images