A method, apparatus, device and medium for registering three-dimensional maps of tubular structures

Abstract

This invention discloses a method for 3D atlas registration of tubular structures, comprising: contour detection of the original sample data of the tubular structure and repair of contour defects; extraction of the sample centerline of the original sample data and the template centerline of the standard atlas; obtaining the morphological transformation relationship based on the sample centerline and the template centerline, and performing morphological correction on the original sample data; calculating the forward deformation field from the corrected sample data to the standard atlas based on rigid and non-rigid registration transformations, performing an inverse transformation on the forward deformation field to obtain the inverse deformation field, and mapping the standard atlas to the original sample data based on the inverse deformation field and the morphological transformation relationship; and performing contour-guided upsampling processing on the standard atlas mapped to the original sample data to obtain a high-resolution final registration atlas. This invention enables high-precision integrated registration of 3D atlases of tubular structures, improves registration efficiency, and enhances the robustness of the registration algorithm to damaged samples.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method, apparatus, device, and medium for three-dimensional image registration of a tubular structure. Background Technology

[0002] Light-sheet fluorescence microscopy, with its high-resolution three-dimensional imaging capabilities, is one of the core tools in biomedical research. High-precision three-dimensional atlas registration of biological samples (such as tissues and organs) acquired through it is a key technical step in this field. For samples such as the spinal cord, which are approximately tubular structures with physiological curvature, unpredictable morphological changes such as twisting, shrinkage, and local damage during sample preparation make it difficult for existing technologies to directly achieve high-precision integrated three-dimensional atlas registration. At the same time, standard atlases of biological samples are mostly low-resolution general templates, which have significant resolution differences compared to the high-resolution raw data captured by light-sheet fluorescence microscopy. Mainstream pixel-level grayscale interpolation upsampling methods only improve resolution at the pixel value level, ignoring geometric features, which easily leads to "jagged" atlas outlines and seriously affects registration accuracy.

[0003] Overall, existing methods generally suffer from problems such as inability to achieve high-precision three-dimensional integrated registration, poor robustness to damaged samples, and low efficiency in processing massive amounts of data, and urgently require targeted solutions. Summary of the Invention

[0004] This invention provides a method for registering three-dimensional images of tubular structures, which can achieve high-precision integrated registration of three-dimensional images of tubular structures, improve registration efficiency and the robustness of the registration algorithm to damaged samples.

[0005] In a first aspect, embodiments of the present invention provide a three-dimensional map registration method for tubular structures, comprising: Contour detection was performed on the original sample data of the tubular structure, and the missing parts of the contour were repaired to obtain a continuous and complete sample contour. Extract the sample centerline from the original sample data and the template centerline of the standard spectrum corresponding to the original sample; Based on the sample centerline and the template centerline, the morphological transformation relationship is obtained, and the original sample data is morphologically corrected according to the morphological transformation relationship to obtain the corrected sample data. Based on rigid registration transformation and non-rigid registration transformation, the positive deformation field of the corrected sample data to the standard spectrum is calculated, and the inverse deformation field is obtained by inverse transformation of the positive deformation field. Based on the inverse deformation field and the morphological transformation relationship, the standard spectrum is mapped to the original sample data. The standard spectrum mapped to the original sample data is subjected to contour-guided upsampling to obtain a high-resolution final registration spectrum.

[0006] Furthermore, the process of performing contour detection on the original sample data of the tubular structure and repairing any missing contour portions to obtain a continuous and complete sample contour includes: Background segmentation is performed on the original sample data to obtain the foreground region and background noise region; A cylindrical three-dimensional structural element matching the scale of the tubular structure is used to perform a small-scale core erosion operation on the foreground region to remove contour impurities, and then a large-scale core expansion operation is performed to initially connect the contour fracture area. The original sample data is traversed along the long axis of the tubular structure using a three-dimensional square window of a preset size to extract the original sample contour. By fitting the missing points of the original sample data contour using B-spline basis functions, a continuous and complete sample contour is obtained.

[0007] Furthermore, obtaining the morphological transformation relationship based on the sample centerline and the template centerline includes: At each pair of corresponding points of the sample centerline and the template centerline, a local right-handed coordinate system is constructed with the corresponding point as the origin and the tangent direction of the corresponding point as the Z-axis. For each pair of corresponding points, calculate the homogeneous transformation matrix from the local right-handed coordinate system of the sample centerline to the local right-handed coordinate system of the template centerline; Based on the homogeneous transformation matrix of all corresponding points, the morphological transformation relationship from the original sample data to the standard spectrum is obtained.

[0008] Furthermore, the step of performing morphological correction on the original sample data according to the morphological transformation relationship to obtain corrected sample data includes: Define a dense sampling grid in the three-dimensional space of the original sample data; Determine the local right-handed coordinate system to which each sampling point belongs, and use the homogeneous transformation matrix corresponding to the coordinate system to transform the coordinates of the sampling point to the local coordinate system space of the standard atlas; The intensity values ​​of the original sample are resampled at the transformed spatial location to obtain the corrected sample data.

[0009] Furthermore, the step of calculating the positive deformation field from the corrected sample data to the standard spectrum based on rigid and non-rigid registration transformations, and then performing an inverse transformation on the positive deformation field to obtain the inverse deformation field, includes: Rigid registration is performed on the corrected sample data to obtain rigid transformation parameters; Based on the rigid registration results, non-rigid registration is performed to obtain non-rigid transformation parameters; Based on the rigid transformation parameters and the non-rigid transformation parameters, the positive deformation field of the corrected sample data to the standard spectrum is calculated; The positive deformation field is interpolated into a dense positive dense deformation field based on the spline interpolation method; The inverse operation of the forward dense deformation field is performed using the Newton-Raphson iteration method to obtain the dense inverse dense deformation field; The inverse dense deformation field is reparameterized into a spline control point network, and the inverse deformation field is obtained by fitting it using the least squares method.

[0010] Furthermore, mapping the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship includes: Based on the inverse deformation field, the standard spectrum is mapped to the corrected sample data space through resampling transformation; Based on the inverse transformation relationship of the morphological transformation relationship, the standard spectrum mapped to the corrected sample data space is restored to the three-dimensional coordinate space of the original sample data, so as to realize the mapping from the standard spectrum to the original sample data.

[0011] Furthermore, the step of performing contour-guided upsampling processing on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum includes: A boundary search algorithm is used to extract the three-dimensional contour surface of the mapped standard atlas; A high-precision interpolation algorithm is used to smoothly fit the three-dimensional contour surface, and the contour points are densified according to the resolution of the original sample to generate a high-resolution contour surface. The pixels enclosed by the high-resolution contour surface are filled and rasterized according to the values ​​of the corresponding regions in the standard atlas to obtain the final high-resolution registration map.

[0012] Secondly, embodiments of the present invention provide a three-dimensional map registration device for a tubular structure, comprising: The contour repair module is used to detect the contour of the original sample data of the tubular structure and repair the missing parts of the contour to obtain a continuous and complete sample contour. The centerline extraction module is used to extract the sample centerline of the original sample data and the template centerline of the standard spectrum corresponding to the original sample. The morphology correction module is used to obtain the morphology transformation relationship based on the sample centerline and the template centerline, and to perform morphology correction on the original sample data according to the morphology transformation relationship to obtain the corrected sample data. The inverse mapping module is used to calculate the forward deformation field of the corrected sample data to the standard spectrum based on rigid registration transformation and non-rigid registration transformation, perform inverse transformation on the forward deformation field to obtain the inverse deformation field, and map the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship. The upsampling module is used to perform contour-guided upsampling processing on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum.

[0013] Thirdly, embodiments of the present invention provide an electronic device, comprising: Memory, used to store computer programs; Processor, used to execute the computer program; Wherein, when the processor executes the computer program, it implements the three-dimensional map registration method for the tubular structure described in any of the first aspects above.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed, implements the three-dimensional map registration method for tubular structures as described in any of the first aspects above.

[0015] Compared with existing technologies, the three-dimensional image registration method for tubular structures provided in this invention has the following advantages: Contour detection is performed on the original sample data of the tubular structure, and missing contour parts are repaired to obtain a continuous and complete sample contour; the sample centerline of the original sample data and the template centerline of the standard image corresponding to the original sample are extracted; a morphological transformation relationship is obtained based on the sample centerline and the template centerline, and the original sample data is morphologically corrected according to the morphological transformation relationship to obtain corrected sample data; based on rigid and non-rigid registration transformations, the forward deformation field of the corrected sample data to the standard image is calculated, and the inverse deformation field is obtained by inverse transformation; based on the inverse deformation field and the morphological transformation relationship, the standard image is mapped to the original sample data; contour-guided upsampling processing is performed on the standard image mapped to the original sample data to obtain a high-resolution final registration image. This invention can achieve high-precision integrated registration of three-dimensional images of tubular structures, improve registration efficiency, and enhance the robustness of the registration algorithm to damaged samples. Attached Figure Description

[0016] To more clearly illustrate the technical features of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a three-dimensional map registration method for a tubular structure provided in an embodiment of the present invention; Figure 2 This is a visualization of the centerline of one embodiment of a three-dimensional map registration method for tubular structures provided by this invention. Figure 3 This is a contour repair effect diagram of an embodiment of a three-dimensional map registration method for tubular structures provided by the present invention; Figure 4 This is a morphological correction effect diagram of an embodiment of a three-dimensional map registration method for tubular structures provided by the present invention; Figure 5 This is a reverse mapping effect diagram of an embodiment of a three-dimensional map registration method for tubular structures provided by the present invention; Figure 6 This is an upsampling effect diagram of an embodiment of a three-dimensional map registration method for tubular structures provided by the present invention; Figure 7 This is a schematic diagram of a tubular three-dimensional map registration device provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0021] In a first aspect, embodiments of the present invention provide a method for three-dimensional map registration of tubular structures, see [link to previous section]. Figure 1 This is a flowchart illustrating an embodiment of a three-dimensional map registration method for a tubular structure provided by the present invention.

[0022] like Figure 1 As shown, the method includes the following steps: S1: Perform contour detection on the original sample data of the tubular structure and repair the missing parts of the contour to obtain a continuous and complete sample contour. S2: Extract the sample centerline of the original sample data and the template centerline of the standard spectrum corresponding to the original sample; S3: Obtain the morphological transformation relationship based on the sample centerline and the template centerline, and perform morphological correction on the original sample data according to the morphological transformation relationship to obtain the corrected sample data; S4: Based on rigid registration transformation and non-rigid registration transformation, calculate the positive deformation field of the corrected sample data to the standard spectrum, perform inverse transformation on the positive deformation field to obtain the inverse deformation field, and map the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship. S5: Perform contour-guided upsampling on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum.

[0023] In practice, the first step is to acquire sample data of tubular biological structures with physiological curvature, such as the spinal cord, to perform contour detection, and to repair any missing parts of the contour, in order to solve the problem of contour extraction failure caused by damage or breakage that easily occurs during sample preparation.

[0024] It should be noted that the sample data and standard spectra involved in this invention are preferably three-dimensional volume data, and the related registration, deformation and reconstruction are all completed in three-dimensional space.

[0025] Furthermore, the sample centerline of the original sample data and the template centerline of the standard spectrum corresponding to the original sample are extracted separately. For example, the Zhang-Suen 3D skeletonization extraction algorithm can be used to process the sample contour and the standard spectrum contour separately. By performing 26-neighborhood topology refinement, while maintaining the overall connectivity of the tubular structure and the integrity of its head and tail endpoints, the 3D tubular contour is gradually shrunk inwards layer by layer, ultimately obtaining a single-pixel-width centerline that is connected head and tail and maintains topological connectivity. These are denoted as the sample centerline and the template centerline, respectively. For an example, see [link to example]. Figure 2 The image shown is a visualization of the centerline extracted in this embodiment.

[0026] It should be noted that this step can also employ methods based on model segmentation and clustering (including hierarchical clustering segmentation, Reeb plot and geodesic distance methods, cylindrical decomposition CSD method, normal vector shrinkage and hierarchical clustering methods, etc.), methods based on geometric analysis and optimization (including Laplace shrinkage method, L1 median skeleton method, generalized rotational symmetry axis method, local correlation point balancing method, etc.), and methods based on deep learning (end-to-end skeleton prediction model, generative adversarial networks, graph neural networks, etc.) to solve for the centerline. This application is not limited to the methods listed above for solving the centerline. Those skilled in the art can use other existing methods or reasonable modifications to achieve the same result without departing from the concept of this application, and all such methods should be included within the scope of protection of this application.

[0027] Furthermore, based on the sample centerline and the template centerline, the global morphological transformation relationship from the original sample data to the standard spectrum is obtained. The original sample data is then morphologically corrected according to the morphological transformation relationship to obtain morphologically corrected sample data. The morphology of this sample data is completely aligned with the standard spectrum, eliminating morphological distortions such as physiological curvature and sample preparation distortion of the original sample data.

[0028] Furthermore, based on the cascaded strategy of rigid and non-rigid registration transformation, the forward deformation field from the corrected sample data to the standard spectrum is calculated. Then, the inverse deformation field is obtained by performing an inverse transformation on the forward deformation field. Based on the inverse deformation field and the morphological transformation relationship, the standard spectrum is mapped to the original sample data. This solves the technical problem of the traditional registration method, which directly deforms the sample data to the spectrum, causing distortion of the original sample data structure. It achieves accurate and lossless mapping of the standard spectrum to the original sample space.

[0029] Finally, contour-guided upsampling is performed on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum.

[0030] In summary, this invention addresses the issue of contour extraction failure caused by damage and breakage during sample preparation, particularly for tubular biological structures like the spinal cord with physiological curvature, by repairing contour defects. This improves the robustness of the registration algorithm to damaged samples. Centerline-based morphological correction accurately corrects complex morphological distortions such as physiological curvature and sample preparation distortion, eliminating the need for segmented registration and manual intervention, thus overcoming the limitation of traditional methods in achieving high-precision, integrated 3D registration. Rigid and non-rigid cascade registration and inverse deformation field solving enable lossless and accurate mapping of the standard atlas to the original sample space, preserving the original sample's morphological and structural information. Contour-guided upsampling completely resolves the jagged contours and geometric distortion issues caused by traditional pixel interpolation, ultimately outputting a high-resolution, smooth-edge registered atlas. The overall solution achieves high-precision integrated registration of 3D atlases of tubular structures, significantly improving registration efficiency and making it widely applicable to various 3D atlas registration scenarios for tubular biological tissues.

[0031] In one optional implementation, the process of performing contour detection on the original sample data of the tubular structure and repairing any missing contour portions to obtain a continuous and complete sample contour includes: Background segmentation is performed on the original sample data to obtain the foreground region and background noise region; A cylindrical three-dimensional structural element matching the scale of the tubular structure is used to perform a small-scale core erosion operation on the foreground region to remove contour impurities, and then a large-scale core expansion operation is performed to initially connect the contour fracture area. The original sample data is traversed along the long axis of the tubular structure using a three-dimensional square window of a preset size to extract the original sample contour. By fitting the missing points of the original sample data contour using B-spline basis functions, a continuous and complete sample contour is obtained.

[0032] Specifically, background segmentation is performed on the original sample data to separate the foreground region containing the tubular structure from the background noise region. The background noise region is directly removed, retaining only the valid foreground data. This avoids direct processing of TB-level raw data, significantly reducing the amount of data processing. For example, a threshold can be preset based on the prior light intensity parameters of the image data acquisition system, and a global threshold segmentation method can be used to separate the foreground and background. Furthermore, a cylindrical three-dimensional structural element is selected, with its axis aligned with the contour of the tubular structure. The following operations are performed sequentially on the foreground region: Smaller-scale nuclear corrosion operation: Corrosion is performed using smaller-scale cylindrical structural elements to remove isolated noise points, burrs and artifacts attached to the contour without damaging the fine contour structure of the pipe wall, thus ensuring the initial integrity of the contour. Large-scale kernel expansion operation: Large-scale cylindrical structuring elements are used to perform expansion, automatically connecting small fractures and defects in the contour, filling small gaps, and laying the foundation for subsequent complete contour extraction.

[0033] It should be noted that the structural element scale can be dynamically adjusted according to the imaging resolution of the sample data. For high-resolution imaging data, a smaller structural element scale can be used to avoid damaging the fine structure, while for low-resolution imaging data, a larger structural element scale can be used to improve the efficiency of fracture connection.

[0034] Furthermore, along the long axis of the tubular structure, the original sample's three-dimensional data is traversed using a three-dimensional square window. The specific parameters and operations are as follows: Window parameter settings: The window size matches the diameter of the tubular structure to ensure that the window can completely cover a single segment of the tubular structure. The window sliding step size is set to 1 / 2 to 2 / 3 of the window size to ensure that there is an overlapping area between adjacent traversed image sub-regions and to avoid omissions in contour extraction. Adaptive window pose adjustment: The window angle fits the local curvature of the contour in real time. By calculating the tangent direction of the local contour, the long axis direction of the window is dynamically adjusted to ensure that the long axis of the window is always consistent with the contour direction, accurately adapting to the physiological curvature of the tubular structure. Contour point extraction: Within each window sub-region, the inner and outer contour points of the tubular structure are extracted using an edge detection algorithm. The contour points of all windows are then integrated to obtain the original sample contour.

[0035] Furthermore, using the effective contour points on both sides of the broken area as feature control points, and through B-spline basis functions with the feature control points as constraints, spline curve parameters that fit the broken area are calculated to determine the fitting trajectory that matches the natural shape of the tubular structure. The fitting trajectory is then interpolated at equal intervals to generate missing contour points with the same density as the original contour points. The missing contour points are then merged with the original contour points to complete the compensation and repair of the broken area, ultimately resulting in a continuous, smooth, and complete sample contour.

[0036] For example, see Figure 3 As shown, Figure 3 The diagram below shows the contour repair effect of this embodiment. The upper diagram shows the direct contour detection result, with the yellow arrow pointing to the broken and damaged area of ​​the real sample data. It can be seen that there are obvious breaks and defects in the contour of this area. The lower diagram shows the result after processing by the broken contour fitting compensation method of this embodiment. The broken and damaged area pointed to by the yellow arrow has been effectively filled, and the continuity and smoothness of the contour have been significantly improved. Finally, a continuous, complete sample contour that fits the shape of the real sample is formed.

[0037] It should be noted that other 3D contour smoothing algorithms can also be used to achieve contour fitting compensation, such as those based on parametric curve / surface fitting (Bezier curve fitting, piecewise polynomial fitting, etc.), implicit function and energy optimization (radial basis function RBF interpolation, regularization / variable models, etc.), statistics and machine learning (principal component analysis and statistical shape models, contour prediction using deep learning PointNet, etc.), geometric processing and filtering (Laplacian smoothing, anisotropic diffusion, etc.), or reconstruction and meshing (Poisson surface reconstruction, moving cubes, etc.). It is understood that those skilled in the art can also use other existing methods or reasonable modifications to achieve the same result without departing from the concept of this application, and all such modifications should be included within the scope of protection of this application.

[0038] This embodiment effectively solves the problem of contour extraction failure caused by distortion, shrinkage, local damage, and breakage during sample preparation by coordinating background segmentation, three-dimensional morphological preprocessing, sliding window contour extraction, and spline fitting compensation, and greatly improves the registration algorithm's adaptability to damaged samples.

[0039] In one optional implementation, obtaining the morphological transformation relationship based on the sample centerline and the template centerline includes: At each pair of corresponding points of the sample centerline and the template centerline, a local right-handed coordinate system is constructed with the corresponding point as the origin and the tangent direction of the corresponding point as the Z-axis. For each pair of corresponding points, calculate the homogeneous transformation matrix from the local right-handed coordinate system of the sample centerline to the local right-handed coordinate system of the template centerline; Based on the homogeneous transformation matrix of all corresponding points, the morphological transformation relationship from the original sample data to the standard spectrum is obtained.

[0040] Specifically, a one-to-one point matching relationship is established between the sample centerline and the template centerline according to their axial positions. For each pair of corresponding points, the tangent direction vector at that point is calculated. Taking that point as the origin and the tangent direction as the Z-axis of the local coordinate system, orthogonalization is performed through vector cross product to construct X and Y axes perpendicular to the Z-axis, thus forming a uniquely determined local right-handed coordinate system.

[0041] For each set of corresponding points, obtain the three-axis directions of the sample's local coordinate system and the template's local coordinate system. Calculate the rotation matrix R from the sample coordinate system to the template coordinate system to align the sample's three-axis directions with the template's three-axis directions. Calculate the scaling matrix S based on the pipe diameter and scale ratio of the sample and template. Calculate the translation vector t between corresponding points. Combine the rotation R, scaling S, and translation t into a homogeneous transformation matrix to uniformly represent the local spatial transformation relationship at the corresponding point. The homogeneous transformation matrix is ​​expressed as follows: ; Next, all local transformations are integrated to form a global morphological transformation relationship covering the entire three-dimensional space.

[0042] This embodiment can stably handle various distortions such as bending, twisting, and shrinkage, and achieve accurate, smooth, and globally consistent three-dimensional integrated morphological adaptation between the sample and the standard spectrum. It solves the technical problem that traditional methods cannot handle complex morphological distortions and provides a reliable spatial transformation basis for subsequent reverse mapping registration.

[0043] In one optional implementation, the step of performing morphological correction on the original sample data according to the morphological transformation relationship to obtain corrected sample data includes: Define a dense sampling grid in the three-dimensional space of the original sample data; Determine the local right-handed coordinate system to which each sampling point belongs, and use the homogeneous transformation matrix corresponding to the coordinate system to transform the coordinates of the sampling point to the local coordinate system space of the standard atlas; The intensity values ​​of the original sample are resampled at the transformed spatial location to obtain the corrected sample data.

[0044] Specifically, a set of sampling points covering the entire sample domain is constructed in the three-dimensional space of the original sample data. For each sampling point, its corresponding local transformation relationship is determined, and the sampling point is transformed from the original sample space to the standard spectral space using the corresponding homogeneous transformation matrix.

[0045] Using the transformed coordinates as a reference, the intensity value at the corresponding position is obtained by interpolation from the original sample data to generate the morphologically corrected sample data.

[0046] For example, see Figure 4 As shown, Figure 4 The images shown are the morphological correction results of this embodiment. The top image is the standard atlas template, the middle image is the actual sample data, which has obvious physiological curvature and sample preparation distortion, and its morphology is significantly different from the standard atlas, making direct integrated registration impossible. The bottom image is the sample data after morphological correction using the method of this embodiment. It can be seen that the complex curvature and distortion of the original sample data have been effectively corrected, and the overall morphology is highly consistent with the standard atlas. Moreover, the internal structure and grayscale characteristics of the sample are completely preserved, without distortion or breakage.

[0047] This embodiment realizes fully automatic and high-precision global morphology correction of tubular structure samples from their original form to their standard spectral form. It effectively solves the registration problem caused by physiological bending of the sample and sample preparation distortion / shrinkage. It eliminates the need for segmented registration and manual intervention, greatly improving registration efficiency and result repeatability.

[0048] In one optional implementation, the step of calculating the positive deformation field from the corrected sample data to the standard spectrum based on rigid and non-rigid registration transformations, and then performing an inverse transformation on the positive deformation field to obtain the inverse deformation field, includes: Rigid registration is performed on the corrected sample data to obtain rigid transformation parameters; Based on the rigid registration results, non-rigid registration is performed to obtain non-rigid transformation parameters; Based on the rigid transformation parameters and the non-rigid transformation parameters, the positive deformation field of the corrected sample data to the standard spectrum is calculated; The positive deformation field is interpolated into a dense positive dense deformation field based on the spline interpolation method; The inverse operation of the forward dense deformation field is performed using the Newton-Raphson iteration method to obtain the dense inverse dense deformation field; The inverse dense deformation field is reparameterized into a spline control point network, and the inverse deformation field is obtained by fitting it using the least squares method.

[0049] Specifically, for the morphologically corrected sample data and the standard spectrum image, rigid registration is first performed to obtain rigid transformation parameters. Based on rigid registration, the local nonlinear differences between the sample data and the standard spectrum are captured to obtain fine non-rigid transformation parameters. For example, the rigid registration algorithm can be a block matching algorithm, and the non-rigid registration algorithm can be an algorithm based on the B-spline free deformation model (FFD).

[0050] By combining rigid and non-rigid transformation parameters, a complete positive deformation field C describing the mapping relationship from the calibrated sample data to the standard spectrum is generated. Using B-spline interpolation, the sparse positive control point displacement field is expanded into a continuous dense deformation field D at image resolution. Newton's iteration method is then used to numerically inverse the positive dense deformation field D, yielding the inverse dense deformation field describing the mapping relationship from the standard spectrum space to the calibrated sample space. .

[0051] Finally, the dense inverse deformation field The network is reparameterized into a B-spline control point network, and the inverse displacement of each pixel position is fitted using the least squares method to obtain the final inverse transformed control point displacement field. This is used for subsequent image mapping operations.

[0052] It should be noted that rigid registration algorithms can also employ methods based on mutual information registration transformation, normalized cross-correlation transformation, feature points (SIFT, SURF, ORB, etc.), moments and principal axes, and iterative nearest point (3D point cloud registration). Non-rigid registration algorithms can also employ methods based on Demons registration, physical models (viscous fluids, elastic bodies), optical flow methods, and deep learning models. It is understood that those skilled in the art can implement other existing methods or reasonable modifications without departing from the concept of this application, and all such modifications should be included within the scope of protection of this application.

[0053] This embodiment establishes a forward mapping relationship through a coarse-to-fine registration strategy and uses a numerical iteration method to achieve a high-precision inverse transformation solution, ultimately providing a reliable basis for accurately mapping the standard spectrum back to the original sample space.

[0054] In one optional implementation, mapping the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship includes: Based on the inverse deformation field, the standard spectrum is mapped to the corrected sample data space through resampling transformation; Based on the inverse transformation relationship of the morphological transformation relationship, the standard spectrum mapped to the corrected sample data space is restored to the three-dimensional coordinate space of the original sample data, so as to realize the mapping from the standard spectrum to the original sample data.

[0055] Specifically, using the solved inverse deformation field, the standard spectrum is transformed from the template space to the morphologically corrected sample space, completing the precise registration of the spectrum and the corrected sample data. Then, using the morphological transformation relationship established in the morphological correction stage, its inverse transformation is solved, restoring the registered spectrum in the corrected sample space to the three-dimensional coordinate space of the original sample data. The registered spectrum is precisely restored from the straight shape after correction to the physiological curvature of the original sample data, realizing the structural correspondence between the standard spectrum and the original sample data. This ensures that the registration result completely preserves the original data structure and morphological characteristics of the sample, meeting the core needs of analyzing the original information of the sample in biomedical research, and completely solving the technical problem of losing the original characteristics of the sample in traditional registration methods.

[0056] For example, see Figure 5 As shown, Figure 5 This is a reverse mapping effect diagram of this embodiment. The upper diagram shows the data registration result after sample data correction. The standard spectrum has been mapped to the morphologically corrected sample space, presenting a straight tubular structure consistent with the standard spectrum. The lower diagram shows the result after mapping to the original sample data. The standard spectrum is accurately restored to the physiological curvature of the original sample data, and it perfectly matches the contour and structure of the original sample data.

[0057] This embodiment, while completing the atlas registration, preserves the morphological, structural, and anatomical information of the original sample data, achieving a high-precision mapping from the standard atlas to the original sample space.

[0058] In one optional implementation, the contour-guided upsampling processing of the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum includes: A boundary search algorithm is used to extract the three-dimensional contour surface of the mapped standard atlas; A high-precision interpolation algorithm is used to smoothly fit the three-dimensional contour surface, and the contour points are densified according to the resolution of the original sample to generate a high-resolution contour surface. The pixels enclosed by the high-resolution contour surface are filled and rasterized according to the values ​​of the corresponding regions in the standard atlas to obtain the final high-resolution registration map.

[0059] Specifically, a boundary search algorithm is used to extract the edge contour surfaces of all partitions from the standard atlas that has been reverse-mapped, which serve as the geometric constraint benchmark for upsampling. For example, the SUKUZI contour search algorithm can be used. This algorithm is based on the connected components and boundary topology of the image and can efficiently and accurately extract the two-dimensional manifold contour surfaces of each anatomical partition in the three-dimensional image.

[0060] High-precision interpolation algorithms, such as cubic spline interpolation or B-spline interpolation, are used to smooth and refine the point density of the extracted low-resolution 3D contour surface, generating a high-resolution contour that matches the resolution of the original sample data.

[0061] Based on the high-resolution contour surface, the region enclosed by the surface is filled with the partition values ​​of the standard map to complete the rasterization and obtain the final high-resolution registration map.

[0062] For example, see Figure 6 As shown, Figure 6 The images shown are the upsampling results of this embodiment. The left image shows the upsampling result of the traditional interpolation method. After the standard atlas is magnified by traditional pixel-level interpolation, its contour edges show obvious jagged and blocky effects, the partition boundaries are blurred, and the smoothness is poor, which seriously affects the visual quality and structural accuracy of the high-resolution atlas. The right image shows the upsampling result of the contour-guided method (this embodiment). While maintaining the same overall shape and partition color as the left image, the atlas contour edges are smooth and continuous without jagged edges, and the internal partition structure is clear and delicate, perfectly matching the high-resolution shape of the original sample.

[0063] This embodiment solves the problems of contour jaggedness and structural distortion caused by direct upsampling of low-resolution standard atlases in the prior art. Under the premise of ensuring the integrity of the atlas anatomical information, it realizes the output of ultra-high-definition atlases that match the original high-resolution sample data.

[0064] Secondly, embodiments of the present invention provide a three-dimensional map registration device for a tubular structure, see [link to previous document]. Figure 7 This is a schematic diagram of one embodiment of a tubular three-dimensional map registration device provided by the present invention.

[0065] like Figure 7 As shown, the device includes: The contour repair module 21 is used to perform contour detection on the original sample data of the tubular structure and repair the missing parts of the contour to obtain a continuous and complete sample contour. Centerline extraction module 22 is used to extract the sample centerline of the original sample data and the template centerline of the standard spectrum corresponding to the original sample; The morphology correction module 23 is used to obtain the morphology transformation relationship based on the sample centerline and the template centerline, and to perform morphology correction on the original sample data according to the morphology transformation relationship to obtain the corrected sample data. The inverse mapping module 24 is used to calculate the positive deformation field of the corrected sample data to the standard spectrum based on rigid registration transformation and non-rigid registration transformation, perform inverse transformation on the positive deformation field to obtain the inverse deformation field, and map the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship. The upsampling module 25 is used to perform contour-guided upsampling processing on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum.

[0066] In one optional implementation, the process of performing contour detection on the original sample data of the tubular structure and repairing any missing contour portions to obtain a continuous and complete sample contour includes: Background segmentation is performed on the original sample data to obtain the foreground region and background noise region; A cylindrical three-dimensional structural element matching the scale of the tubular structure is used to perform a small-scale core erosion operation on the foreground region to remove contour impurities, and then a large-scale core expansion operation is performed to initially connect the contour fracture area. The original sample data is traversed along the long axis of the tubular structure using a three-dimensional square window of a preset size to extract the original sample contour. By fitting the missing points of the original sample data contour using B-spline basis functions, a continuous and complete sample contour is obtained.

[0067] In one optional implementation, obtaining the morphological transformation relationship based on the sample centerline and the template centerline includes: At each pair of corresponding points of the sample centerline and the template centerline, a local right-handed coordinate system is constructed with the corresponding point as the origin and the tangent direction of the corresponding point as the Z-axis. For each pair of corresponding points, calculate the homogeneous transformation matrix from the local right-handed coordinate system of the sample centerline to the local right-handed coordinate system of the template centerline; Based on the homogeneous transformation matrix of all corresponding points, the morphological transformation relationship from the original sample data to the standard spectrum is obtained.

[0068] In one optional implementation, the step of performing morphological correction on the original sample data according to the morphological transformation relationship to obtain corrected sample data includes: Define a dense sampling grid in the three-dimensional space of the original sample data; Determine the local right-handed coordinate system to which each sampling point belongs, and use the homogeneous transformation matrix corresponding to the coordinate system to transform the coordinates of the sampling point to the local coordinate system space of the standard atlas; The intensity values ​​of the original sample are resampled at the transformed spatial location to obtain the corrected sample data.

[0069] In one optional implementation, the step of calculating the positive deformation field from the corrected sample data to the standard spectrum based on rigid and non-rigid registration transformations, and then performing an inverse transformation on the positive deformation field to obtain the inverse deformation field, includes: Rigid registration is performed on the corrected sample data to obtain rigid transformation parameters; Based on the rigid registration results, non-rigid registration is performed to obtain non-rigid transformation parameters; Based on the rigid transformation parameters and the non-rigid transformation parameters, the positive deformation field of the corrected sample data to the standard spectrum is calculated; The positive deformation field is interpolated into a dense positive dense deformation field based on the spline interpolation method; The inverse operation of the forward dense deformation field is performed using the Newton-Raphson iteration method to obtain the dense inverse dense deformation field; The inverse dense deformation field is reparameterized into a spline control point network, and the inverse deformation field is obtained by fitting it using the least squares method.

[0070] In one optional implementation, mapping the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship includes: Based on the inverse deformation field, the standard spectrum is mapped to the corrected sample data space through resampling transformation; Based on the inverse transformation relationship of the morphological transformation relationship, the standard spectrum mapped to the corrected sample data space is restored to the three-dimensional coordinate space of the original sample data, so as to realize the mapping from the standard spectrum to the original sample data.

[0071] In one optional implementation, the contour-guided upsampling processing of the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum includes: A boundary search algorithm is used to extract the three-dimensional contour surface of the mapped standard atlas; A high-precision interpolation algorithm is used to smoothly fit the three-dimensional contour surface, and the contour points are densified according to the resolution of the original sample to generate a high-resolution contour surface. The pixels enclosed by the high-resolution contour surface are filled and rasterized according to the values ​​of the corresponding regions in the standard atlas to obtain the final high-resolution registration map.

[0072] It should be noted that the tubular three-dimensional map registration device provided in this embodiment of the invention is used to execute all the process steps of the tubular three-dimensional map registration method of the above embodiment. The working principle and beneficial effect of the two are one-to-one, so they will not be described again.

[0073] Thirdly, embodiments of the present invention provide an electronic device, see [link to previous document]. Figure 8 The diagram shown is a structural schematic of an electronic device provided in an embodiment of the present invention.

[0074] like Figure 8 As shown, the device includes: Memory 31 is used to store computer programs; Processor 32 is used to execute the computer program; When the processor 32 executes the computer program, it implements the three-dimensional map registration method for tubular structures as described in any of the above embodiments.

[0075] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 32 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0076] The processor 32 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0077] The memory 31 can be used to store the computer programs and / or modules. The processor 32 implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory 31 and calling the data stored in the memory 31. The memory 31 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 31 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0078] It should be noted that the aforementioned electronic devices include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 8 The structural diagram is merely an example of the electronic device described above and does not constitute a limitation on the electronic device. It may include more components than shown in the diagram, or combine certain components, or use different components.

[0079] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed, implements the three-dimensional map registration method for tubular structures described in any of the above embodiments.

[0080] It should be understood that all or part of the processes in the above-described method for three-dimensional map registration of tubular structures can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-described method for three-dimensional map registration of tubular structures. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0081] It should be noted that the technical solution of the present invention is not only applicable to the tubular structure of the spinal cord, but can also be directly adapted to the three-dimensional image atlas registration of other tubular biological structures such as blood vessels, esophagus, and intestines.

[0082] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. It should be noted that, for those skilled in the art, several equivalent obvious modifications and / or equivalent substitutions can be made without departing from the technical principles of the present invention, and these obvious modifications and / or equivalent substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method of registering three-dimensional maps of tubular structures, characterized by, include: Contour detection was performed on the original sample data of the tubular structure, and the missing parts of the contour were repaired to obtain a continuous and complete sample contour. Extract the sample centerline from the original sample data and the template centerline of the standard spectrum corresponding to the original sample; Based on the sample centerline and the template centerline, the morphological transformation relationship is obtained, and the original sample data is morphologically corrected according to the morphological transformation relationship to obtain the corrected sample data. Based on rigid registration transformation and non-rigid registration transformation, the positive deformation field of the corrected sample data to the standard spectrum is calculated, and the inverse deformation field is obtained by inverse transformation of the positive deformation field. Based on the inverse deformation field and the morphological transformation relationship, the standard spectrum is mapped to the original sample data. The standard spectrum mapped to the original sample data is subjected to contour-guided upsampling to obtain a high-resolution final registration spectrum.

2. The method of tubular structure three-dimensional atlas registration of claim 1, wherein, The process of performing contour detection on the original sample data of the tubular structure and repairing any missing contour portions to obtain a continuous and complete sample contour includes: Background segmentation is performed on the original sample data to obtain the foreground region and the background noise region; A cylindrical three-dimensional structural element matching the scale of the tubular structure is used to perform a small-scale core erosion operation on the foreground region to remove contour impurities, and then a large-scale core expansion operation is performed to initially connect the contour fracture area. The original sample data is extracted by traversing the original sample data through a three-dimensional square window of a preset size along the long axis of the tubular structure. By fitting the missing points of the original sample data contour using B-spline basis functions, a continuous and complete sample contour is obtained.

3. The method of tubular structure three-dimensional atlas registration of claim 1, wherein, The process of obtaining the morphological transformation relationship based on the sample centerline and the template centerline includes: At each pair of corresponding points of the sample centerline and the template centerline, a local right-handed coordinate system is constructed with the corresponding point as the origin and the tangent direction of the corresponding point as the Z-axis. For each pair of corresponding points, calculate the homogeneous transformation matrix from the local right-handed coordinate system of the sample centerline to the local right-handed coordinate system of the template centerline; Based on the homogeneous transformation matrix of all corresponding points, the morphological transformation relationship from the original sample data to the standard spectrum is obtained.

4. The three-dimensional map registration method for tubular structures as described in claim 3, characterized in that, The step of performing morphological correction on the original sample data according to the morphological transformation relationship to obtain corrected sample data includes: Define a dense sampling grid in the three-dimensional space of the original sample data; Determine the local right-handed coordinate system to which each sampling point belongs, and transform the coordinates of the sampling point to the local coordinate system space of the standard atlas using the homogeneous transformation matrix corresponding to the coordinate system to which the sampling point belongs; The intensity values ​​of the original sample are resampled at the transformed spatial location to obtain the corrected sample data.

5. The three-dimensional map registration method for tubular structures as described in claim 1, characterized in that, The process involves calculating the positive deformation field from the corrected sample data to the standard spectrum based on rigid and non-rigid registration transformations, and then performing an inverse transformation on the positive deformation field to obtain the inverse deformation field, including: Rigid registration is performed on the corrected sample data to obtain rigid transformation parameters; Based on the rigid registration results, non-rigid registration is performed to obtain non-rigid transformation parameters; Based on the rigid transformation parameters and the non-rigid transformation parameters, the positive deformation field of the corrected sample data to the standard spectrum is calculated; The positive deformation field is interpolated into a dense positive dense deformation field based on the spline interpolation method; The inverse operation of the forward dense deformation field is performed using the Newton-Raphson iteration method to obtain the dense inverse dense deformation field; The inverse dense deformation field is reparameterized into a spline control point network, and the inverse deformation field is obtained by fitting it using the least squares method.

6. The three-dimensional map registration method for tubular structures as described in claim 1, characterized in that, The process of mapping the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship includes: Based on the inverse deformation field, the standard spectrum is mapped to the corrected sample data space through resampling transformation; Based on the inverse transformation relationship of the morphological transformation relationship, the standard spectrum mapped to the corrected sample data space is restored to the three-dimensional coordinate space of the original sample data, so as to realize the mapping from the standard spectrum to the original sample data.

7. The three-dimensional map registration method for tubular structures as described in claim 1, characterized in that, The step of performing contour-guided upsampling processing on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum includes: A boundary search algorithm is used to extract the three-dimensional contour surface of the mapped standard atlas; A high-precision interpolation algorithm is used to smoothly fit the three-dimensional contour surface, and the contour points are densified according to the resolution of the original sample to generate a high-resolution contour surface. The pixels enclosed by the high-resolution contour surface are filled and rasterized according to the values ​​of the corresponding regions in the standard atlas to obtain the final high-resolution registration map.

8. A three-dimensional map registration device with a tubular structure, characterized in that, include: The contour repair module is used to detect the contour of the original sample data of the tubular structure and repair the missing parts of the contour to obtain a continuous and complete sample contour. The centerline extraction module is used to extract the sample centerline of the original sample data and the template centerline of the standard spectrum corresponding to the original sample. The morphology correction module is used to obtain the morphology transformation relationship based on the sample centerline and the template centerline, and to perform morphology correction on the original sample data according to the morphology transformation relationship to obtain the corrected sample data. The inverse mapping module is used to calculate the forward deformation field of the corrected sample data to the standard spectrum based on rigid registration transformation and non-rigid registration transformation, perform inverse transformation on the forward deformation field to obtain the inverse deformation field, and map the standard spectrum to the original sample data based on the inverse deformation field and the morphological transformation relationship. The upsampling module is used to perform contour-guided upsampling processing on the standard spectrum mapped to the original sample data to obtain a high-resolution final registration spectrum.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program; Wherein, when the processor executes the computer program, it implements the three-dimensional map registration method for tubular structures as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the three-dimensional map registration method for tubular structures as described in any one of claims 1 to 7.

Images