Intelligent measurement method and system of scoliosis angle based on spine CT three-dimensional reconstruction

Abstract

The present application relates to the field of image processing technology, and discloses an intelligent measurement method and system for scoliosis angle based on three-dimensional reconstruction of spine CT. The method includes: performing adaptive threshold segmentation and grayscale optimization processing on the spine CT sequence images to obtain standardized preprocessed images and performing adaptive bone tissue area recognition and boundary enhancement processing to obtain spinal bone segmentation data; performing three-dimensional mesh construction and surface smoothing processing on the spinal bone segmentation data to obtain spinal three-dimensional model data; performing vertebral space recognition and key point positioning processing on the spinal three-dimensional model data to obtain a vertebral feature data set and perform spinal midline curve construction and end vertebral marking processing to obtain Cobb angle data; performing three-dimensional spatial reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report. The present application realizes fully automated processing from CT image acquisition to the generation of the final measurement result report while ensuring measurement accuracy.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to an intelligent measurement method and system for scoliosis angle based on spinal CT three-dimensional reconstruction. Background Art

[0002] At present, the diagnosis and measurement of scoliosis mainly rely on X-ray imaging examinations. The traditional Cobb angle measurement method requires doctors to manually mark the end vertebrae and measure the angle on a two-dimensional X-ray film. With the development of medical imaging technology, CT imaging has been widely used in the diagnosis of spinal diseases due to its advantages such as high resolution and three-dimensional imaging. Existing scoliosis measurement methods based on CT images include manual measurement, semi-automatic measurement, and automatic measurement. Among them, the automatic measurement method uses computer image processing technology to reconstruct the CT image in three dimensions, and performs automated angle measurement based on the reconstruction results. This method has significant advantages in reducing human errors and improving measurement efficiency.

[0003] However, there are still some technical problems with the existing automatic measurement methods. First, the grayscale value difference between bone tissue and soft tissue in CT images is not obvious, resulting in low accuracy of bone tissue segmentation; second, the three-dimensional reconstruction of the spine is prone to problems such as rough surface and mesh deformation; third, in the automatic identification of the end vertebrae, morphological abnormalities caused by pathological deformation will affect the recognition accuracy; finally, when dealing with complex scoliosis cases, especially multi-curve scoliosis, the accuracy and reliability of the measurement results of the existing methods still need to be improved. Summary of the invention

[0004] The present application provides a method and system for intelligent measurement of scoliosis angle based on three-dimensional reconstruction of spinal CT, which is used to achieve fully automated processing from CT image acquisition to final measurement result report generation while ensuring measurement accuracy. In particular, when dealing with complex scoliosis cases, it can accurately identify the position of the end vertebrae and perform reliable angle measurement.

[0005] In the first aspect, the present application provides an intelligent measurement method for scoliosis angle based on three-dimensional reconstruction of spine CT, and the intelligent measurement method for scoliosis angle based on three-dimensional reconstruction of spine CT includes: performing adaptive threshold segmentation and grayscale optimization processing on the acquired spine CT sequence images to obtain a standardized preprocessed image; performing adaptive bone tissue area recognition and boundary enhancement processing on the standardized preprocessed image to obtain spinal bone segmentation data; performing three-dimensional mesh construction and surface smoothing processing on the spinal bone segmentation data to obtain spinal three-dimensional model data; performing vertebral space recognition and key point positioning processing on the spinal three-dimensional model data to obtain a vertebral feature data set; performing spinal midline curve construction and end vertebral marking processing on the vertebral feature data set to obtain Cobb angle data; performing three-dimensional spatial reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report.

[0006] In a second aspect, the present application provides an intelligent measurement system for scoliosis angle based on three-dimensional reconstruction of spine CT, and the intelligent measurement system for scoliosis angle based on three-dimensional reconstruction of spine CT includes:

[0007] A segmentation module is used to perform adaptive threshold segmentation and grayscale optimization processing on the acquired spinal CT sequence images to obtain a standardized pre-processed image;

[0008] A recognition module, used for performing adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data;

[0009] A construction module is used to perform three-dimensional mesh construction and surface smoothing on the spinal bone segmentation data to obtain spinal three-dimensional model data;

[0010] A positioning module, used for performing vertebral space identification and key point positioning processing on the three-dimensional model data of the spine to obtain a vertebral feature data set;

[0011] A marking module, used for constructing a spinal midline curve and performing end vertebra marking processing on the vertebral feature data set to obtain Cobb angle data;

[0012] The projection module is used to perform three-dimensional spatial reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report.

[0013] In the technical solution provided by the present application, by performing adaptive threshold segmentation and grayscale optimization processing on the spinal CT sequence images, the bone tissue area can be accurately extracted, the interference of background noise and artifacts can be reduced, and a clear standardized pre-processed image can be obtained. The segmentation accuracy of the spinal bones is effectively improved by using adaptive bone tissue area recognition and boundary enhancement processing, especially the segmentation effect in complex areas such as the vertebral edge and intervertebral space is significantly improved. Through three-dimensional grid construction and surface smoothing processing, high-quality reconstruction of the spinal three-dimensional model is achieved, the continuity and smoothness of the model surface are guaranteed, and a reliable geometric basis is provided for subsequent measurement and analysis. The vertebral space identification and key point positioning processing are performed on the spinal three-dimensional model data, and the anatomical feature points of the vertebra are accurately marked, including the vertebral center point, the end plate center point and the pedicle point, etc. The precise positioning of these feature points is crucial for the measurement of the Cobb angle. Through the construction of the spinal midline curve and the end vertebral marking processing, the scope and degree of scoliosis are accurately identified, especially in complex scoliosis cases, the end vertebral position of the primary scoliosis and the secondary scoliosis can be reliably determined. Finally, 3D space reconstruction and multi-plane projection processing are used to intuitively display the measurement results, generate standardized measurement reports, and provide comprehensive diagnostic reference information. The entire solution realizes fully automatic processing from image acquisition to final measurement result output, significantly improves measurement efficiency, reduces the subjectivity of manual operation, and ensures the accuracy and repeatability of measurement results. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0015] Figure 1 A schematic diagram of an embodiment of a scoliosis angle intelligent measurement method based on spine CT three-dimensional reconstruction in an embodiment of the present application;

[0016] Figure 2 This is a schematic diagram of an embodiment of a scoliosis angle intelligent measurement system based on spinal CT three-dimensional reconstruction in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The embodiment of the present application provides a method and system for intelligent measurement of scoliosis angle based on three-dimensional reconstruction of spinal CT. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0018] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the intelligent measurement method of scoliosis angle based on spine CT three-dimensional reconstruction includes:

[0019] Step S101, performing adaptive threshold segmentation and grayscale optimization processing on the acquired spinal CT sequence images to obtain a standardized preprocessed image;

[0020] Step S102, performing adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data;

[0021] Step S103, constructing a three-dimensional mesh and performing surface smoothing on the spinal bone segmentation data to obtain spinal three-dimensional model data;

[0022] Step S104, performing vertebral space identification and key point positioning processing on the three-dimensional model data of the spine to obtain a vertebral feature data set;

[0023] Step S105, constructing a spinal midline curve and performing end vertebral marking processing on the vertebral feature data set to obtain Cobb angle data;

[0024] Step S106: Perform three-dimensional space reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report.

[0025] It is understandable that the execution subject of the present application can be a scoliosis angle intelligent measurement system based on spinal CT three-dimensional reconstruction, or a terminal or a server, which is not limited here. The present application embodiment is described by taking the server as the execution subject as an example.

[0026] Specifically, the acquired spinal CT sequence images were preprocessed, the original image data was obtained through DICOM format analysis, and the pixel resolution was standardized, and the resolution of all CT images was unified to 1mm×1mm. Appropriate window width and window position parameters were set for the unified images, with the window width value set to 2000HU and the window position value set to 400HU to highlight the bone tissue structure. The bilateral filter was used to reduce the noise of the image, with the spatial domain standard deviation of the filter set to 3 and the gray value domain standard deviation set to 30, which effectively suppressed the noise while maintaining the edge information. Subsequently, bone tissue area recognition and boundary enhancement were performed. First, preliminary segmentation was performed based on the CT value range, and the area with CT values ​​between 200HU and 2000HU was marked as the candidate bone tissue area. Edge detection was performed on the marked area, and the image gradient was calculated using the Sobel operator to extract the bone tissue contour. The stray areas with an area less than 100 square millimeters were eliminated through connected domain analysis, and the remaining areas were subjected to morphological processing, including a dilation operation with a radius of 2 pixels and an erosion operation with a radius of 1 pixel, to optimize the segmentation boundary.

[0027] A three-dimensional mesh model was constructed based on the segmentation results. The two-dimensional segmentation results were reorganized in three dimensions according to the CT layer thickness information, and the initial three-dimensional mesh was generated by voxelization. The mesh was assigned a coordinate system, and a local coordinate system with the center point of the first lumbar vertebra as the origin was established. The surface point cloud data was extracted, and the point cloud was uniformly sampled, with the sampling density set to 2 points per square millimeter. The surface normal vector and local curvature features were calculated, and the mesh was adaptively simplified based on the curvature value, maintaining high accuracy in areas with large curvature and appropriately reducing accuracy in areas with small curvature. The vertebral gap was identified by spatial gradient analysis, and the gradient values ​​of each point on the mesh surface were calculated. The continuous areas with large gradient values ​​were marked as candidate gap areas. Density clustering analysis was performed on the candidate areas, with the cluster radius set to 3 mm and the minimum number of points set to 50 points. The clustering results were used as vertebral separation markers. The end plate surface of each vertebra was extracted, the coordinates of the geometric center point were calculated, and the location of the pedicle area was identified. Based on these feature points, a vertebral posture feature vector was constructed, which contained information such as spatial position and inclination angle.

[0028] The spinal midline curve is constructed based on the vertebral feature data, and the vertebral center points are smoothly connected using the cubic spline interpolation method. The curvature value of each point on the curve is calculated, and the curvature extreme point is extracted as the scoliosis vertex candidate. The curve is segmented into regions to identify the primary and secondary scoliosis regions. The end vertebral position is determined in each scoliosis region, the end plate inclination is calculated, and the Cobb angle is calculated by the vector projection method. Finally, multi-plane reconstruction is performed, and the three-dimensional reconstruction results are projected to the sagittal plane, coronal plane and transverse section, and the end vertebral position and measurement angle are marked on each plane. A standardized measurement report is generated, which contains patient information, image parameters, measurement results, etc.

[0029] For example, a set of 1mm-thick and 200-layer thoracic and lumbar CT images were processed, and 12 vertebral regions were initially obtained through threshold segmentation. After boundary optimization, the precise vertebral contour was obtained. After three-dimensional reconstruction, 240 feature points were extracted, including 12 pairs of pedicle points and 24 endplate center points. A major scoliosis was identified through curvature analysis, with the top vertebra located at the 10th thoracic vertebra, the upper end vertebra at the 8th thoracic vertebra (endplate inclination 26.4 degrees), and the lower end vertebra at the 12th thoracic vertebra (endplate inclination 23.8 degrees). The Cobb angle of the main scoliosis was calculated to be 50.2 degrees. The measurement results were intuitively displayed through the three-dimensional reconstruction model and multi-plane projection images, and fully recorded in the measurement report.

[0030] In the implementation of this application, by performing adaptive threshold segmentation and grayscale optimization processing on the spinal CT sequence images, the bone tissue area can be accurately extracted, the interference of background noise and artifacts can be reduced, and a clear standardized pre-processed image can be obtained. The adaptive bone tissue area recognition and boundary enhancement processing are used to effectively improve the segmentation accuracy of the spinal bones, especially the segmentation effect in complex areas such as the vertebral edge and intervertebral space. Through three-dimensional grid construction and surface smoothing, high-quality spinal three-dimensional model reconstruction is achieved, the continuity and smoothness of the model surface are guaranteed, and a reliable geometric basis is provided for subsequent measurement and analysis. The vertebral space identification and key point positioning processing are performed on the spinal three-dimensional model data, and the anatomical feature points of the vertebra are accurately marked, including the vertebral center point, the end plate center point and the pedicle point, etc. The precise positioning of these feature points is crucial for the measurement of Cobb angles. Through the construction of the spinal midline curve and the end vertebral marking processing, the range and degree of scoliosis are accurately identified, especially in complex scoliosis cases, the end vertebral position of the primary scoliosis and secondary scoliosis can be reliably determined. Finally, 3D space reconstruction and multi-plane projection processing are used to intuitively display the measurement results, generate standardized measurement reports, and provide comprehensive diagnostic reference information. The entire solution realizes fully automatic processing from image acquisition to final measurement result output, significantly improves measurement efficiency, reduces the subjectivity of manual operation, and ensures the accuracy and repeatability of measurement results.

[0031] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0032] (1) Perform DICOM format parsing on the spinal CT sequence images to obtain the original image data, and perform pixel resolution unification on the original image data to obtain uniform resolution image data;

[0033] (2) performing window width and window position parameter setting processing on the uniform resolution image data to obtain bone tissue enhanced image data, and performing noise suppression processing on the bone tissue enhanced image data through bilateral filtering to obtain denoised image data;

[0034] (3) Perform grayscale histogram analysis on the denoised image data to obtain grayscale distribution feature data, and perform grayscale interval division on the grayscale distribution feature data to obtain CT value segmentation data;

[0035] (4) Performing bone tissue region extraction processing on the CT value segmentation data to obtain preliminary segmentation image data, and performing morphological operation processing on the preliminary segmentation image data to obtain boundary optimized image data;

[0036] (5) Perform gamma curve correction on the boundary optimized image data to obtain a standardized preprocessed image.

[0037] Specifically, the DICOM format file is parsed. As a standard format for medical imaging, the DICOM format contains image data and scanning parameter information. The parsing process extracts image matrix data and key parameters, including pixel spacing, layer thickness, image position information, etc. The original image data is stored in the form of a 16-bit integer matrix with a value range of -1024 to 3072 Hounsfield units, which covers the CT values ​​of all tissues in the human body. The extracted original image data is processed to unify the pixel resolution, and the images acquired by different devices and different scanning parameters are converted to standard resolution. The image is resampled by the bilinear interpolation method to achieve unified resolution while maintaining image quality. This step ensures that the data processed subsequently has the same spatial resolution, which is convenient for three-dimensional reconstruction and measurement.

[0038] After unifying the resolution, the window width and window position parameters are adjusted, which is a key step to enhance the display effect of bone tissue. The CT value of bone tissue is generally distributed between 200 and 2000 Hounsfield units. Setting appropriate window width and window position parameters can highlight the bone tissue structure. The window width controls the width of the display range, and the window position determines the center value of the display range. For bone tissue imaging, the window width is set to 2000 Hounsfield units and the window position is set to 400 Hounsfield units, so that the bone tissue structure is clearly visible in the image. Bilateral filtering is used to reduce noise in the enhanced image. The bilateral filter maintains edge information while reducing noise by considering the spatial position relationship and gray value difference of pixels at the same time. During the filtering process, the standard deviation of the spatial domain is set to 3 pixels, and the standard deviation of the gray value domain is set to 30 gray levels. This set of parameters has shown good noise reduction and edge retention capabilities in practice.

[0039] The grayscale histogram of the denoised image is analyzed to count the pixel distribution of different grayscale values. The grayscale demarcation point of bone tissue and soft tissue is determined by analyzing the peak and valley characteristics of the histogram. The grayscale interval is divided based on the statistical analysis results, focusing on the grayscale range corresponding to the bone tissue, and CT value segmentation data is generated. The segmentation process uses an adaptive threshold method to dynamically adjust the segmentation threshold according to image features. The bone tissue area is extracted based on the CT value segmentation results to generate a preliminary segmented image. The segmentation results are morphologically processed, including opening and closing operations. The opening operation is used to remove small isolated areas, and the closing operation is used to fill small holes near the boundary. The morphological operation uses a circular structure element with a radius of 2 pixels, which is large enough to handle most noise and boundary discontinuity problems.

[0040] Finally, the image after boundary optimization is subjected to Gamma curve correction to adjust the overall grayscale distribution of the image. Gamma correction enhances the local contrast of the image through nonlinear transformation, making the bone tissue boundary clearer. The correction parameters are dynamically adjusted according to the image features to ensure that the processed image not only maintains the original anatomical structure information but also has a good visual effect.

[0041] Taking the CT data of a scoliosis patient as an example, the original scanning sequence contains 180 cross-sectional images, each with a size of 512×512 pixels, a pixel pitch of 0.8 mm, and a layer thickness of 1.0 mm. Through resolution unification processing, the image size is adjusted to 410×410 pixels, and the resolution is unified to 1 mm×1 mm. After setting the window width to 2000 Hounsfield units and the window position to 400 Hounsfield units, the CT value of bone tissue is mapped from the original range of -100 to 1900 Hounsfield units to the display grayscale range of 0 to 255. After bilateral filtering and denoising, the standard deviation of the image is reduced from the original 45.6 to 12.8, while the edge sharpness remains at the original level. The threshold of the bone tissue CT value is determined to be 225 Hounsfield units through histogram analysis, and a continuous and complete bone tissue contour is obtained after morphological processing. Finally, after Gamma correction, a standardized preprocessed image is obtained, laying the foundation for subsequent accurate segmentation and three-dimensional reconstruction.

[0042] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0043] (1) Perform bone tissue density threshold analysis on the standardized preprocessed images to obtain bone density feature data, and perform bone tissue region classification on the bone density feature data to obtain preliminary bone tissue labeling data;

[0044] (2) Performing edge detection processing on the preliminary bone tissue labeling data to obtain bone tissue contour data, and performing connected domain labeling processing on the bone tissue contour data to obtain bone tissue region data;

[0045] (3) Performing vertebral contour feature extraction processing on the bone tissue region data to obtain vertebral region feature data, and performing spatial position constraint processing on the vertebral region feature data to obtain vertebral position data;

[0046] (4) Performing bone tissue morphological analysis on the vertebral position data to obtain bone tissue morphological feature data, and performing boundary enhancement processing on the bone tissue morphological feature data to obtain boundary enhanced image data;

[0047] (5) Performing region growing segmentation processing on the boundary enhanced image data to obtain target region data, and performing morphological smoothing processing on the target region data to obtain spinal bone segmentation data.

[0048] Specifically, the bone tissue density threshold analysis is performed on the standardized preprocessed image. Since bone tissue appears as a high-density area in the CT image, its CT value is generally between 200 and 2000 Heinz units. According to the image histogram characteristics, 225 Heinz units are selected as the initial threshold to extract bone density features from the image. The density feature data contains the CT value of each pixel and its neighborhood statistical characteristics, such as local mean, standard deviation and other information. Based on these features, regional classification is performed to divide the image into bone tissue areas and non-bone tissue areas, and a preliminary bone tissue labeling map is generated. Edge detection is performed on the preliminary marked bone tissue area, and the Sobel operator is used for gradient calculation. The Sobel operator includes gradient operators in the horizontal and vertical directions, and the gradient values ​​in the x and y directions are calculated respectively, and then the edge strength and direction information are obtained comprehensively. After edge detection, the contour data of the bone tissue is obtained, and the connected domain analysis is performed based on the contour data to mark all connected bone tissue areas. The connected domain labeling adopts the 8-neighborhood connectivity criterion to classify adjacent bone tissue pixels into the same area and assign a unique identification number to each area.

[0049] The vertebral contour features are extracted from the marked bone tissue area, and the geometric features of each connected area are analyzed, including parameters such as area, circumference, and circularity. The vertebra presents a characteristic shape in the cross section, and the vertebral area is identified by shape features. The extracted regional feature data is combined with spatial position information to establish a spatial constraint relationship and determine the spatial connection between adjacent vertebrae. The position constraint is based on anatomical knowledge, considering the normal arrangement of vertebrae, and screening areas that meet the position constraint conditions. In the bone tissue morphological analysis stage, the morphological features of each vertebral area are calculated, including the main axis direction of the area, axial length ratio, convex hull features, etc. The morphological feature data reflects the three-dimensional structural characteristics of the vertebra and provides a basis for boundary enhancement. The boundary enhancement process adopts a gradient-based enhancement method to enhance the contrast and continuity of the boundary while maintaining the original boundary position.

[0050] The region growing algorithm is used to accurately segment the image after boundary enhancement. The center point of the vertebral region is selected as the seed point, and the region is gradually expanded according to the grayscale similarity and spatial continuity criteria. During the region growing process, the similarity threshold is dynamically adjusted according to the statistical characteristics of the local region to ensure the accuracy of the segmentation. After obtaining the target region, morphological smoothing is performed to remove the small bumps and bumps on the boundary to obtain a smooth and continuous spinal bone segmentation result.

[0051] For example, after preprocessing, the CT sequence of a patient with scoliosis was segmented with an initial bone density threshold of 225 Hounsfield units, and a preliminary marking result containing 12 vertebrae was obtained. In the contour data obtained by edge detection, each vertebra contained an average of about 2000 boundary points. Connected domain analysis marked 15 candidate regions, of which 12 were vertebral regions and 3 were rib regions. The average area of ​​the vertebral region was 785 square millimeters, the circumference was about 108 millimeters, and the circularity was between 0.82 and 0.89. Spatial position constraints removed two candidate regions that did not conform to the anatomical position. Morphological analysis showed that the axial length ratio of normal vertebrae was between 1.2 and 1.4, while the axial length ratio of scoliotic vertebrae reached 1.8. During region growing segmentation, the initial seed point was set at the center of the region with a CT value greater than 300 Hounsfield units, and the similarity threshold was set to ±15% of the local mean. The final segmentation result accurately identified the boundaries of each vertebra.

[0052] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0053] (1) Perform voxel grid division on the spinal bone segmentation data to obtain basic grid data, and perform coordinate system conversion on the basic grid data to obtain unified coordinate grid data;

[0054] (2) Performing surface point cloud extraction processing on the unified coordinate grid data to obtain surface feature point data, and performing density equalization processing on the surface feature point data to obtain uniform point cloud data;

[0055] (3) Calculate the surface normal vector of the uniform point cloud data to obtain surface directional feature data, and perform curvature analysis on the surface directional feature data to obtain local curvature data;

[0056] (4) Reconstruct the local curvature data to obtain triangular patch data, and extract geometric features from the triangular patch data to obtain surface topology data;

[0057] (5) Performing surface smoothing interpolation processing on the surface topology data to obtain initial reconstruction data, and performing boundary constraint processing on the initial reconstruction data to obtain the three-dimensional model data of the spine.

[0058] Specifically, the spinal bone segmentation data is divided into voxel grids. Voxels are the basic units in three-dimensional space, similar to pixels in two-dimensional images. During the division process, the continuous two-dimensional segmentation results are organized into a three-dimensional grid structure according to the layer thickness and voxel size of the CT image. Each voxel carries position coordinates and grayscale value information, and the voxel size is set to 1 mm × 1 mm × 1 mm to ensure reconstruction accuracy. The basic grid data needs to be converted to a unified three-dimensional coordinate system. The coordinate transformation uses a rigid body transformation method to align CT images at different levels to the same coordinate system. The center point of the first lumbar vertebra is selected as the transformation reference point, and a local coordinate system with this point as the origin is established. The converted grid data is represented in a unified coordinate system, which is convenient for subsequent feature extraction and analysis.

[0059] Surface point clouds are extracted from the unified grid data. Surface point clouds are a set of discrete sampling points on the surface of an object. The extraction process uses gradient information to identify the surface position of the voxel grid. The center of each surface voxel is used as a sampling point, and its three-dimensional coordinate value is recorded. The density of the initially extracted point cloud is uneven, and density equalization is required. The equalization process uses the voxel downsampling method to retain a representative point in each cubic area of ​​2 mm to ensure that the point cloud is evenly distributed. The surface normal vector is calculated for the homogenized point cloud data. The normal vector represents the vertical direction of the surface at the point. The local neighborhood of the point is considered in the calculation, and the point set in a spherical area with a radius of 3 mm is selected. The direction of the normal vector is determined by principal component analysis. After obtaining the surface directional features, local curvature analysis is performed. The curvature reflects the degree of curvature of the surface. The principal curvature and Gaussian curvature are calculated by fitting the local point set to describe the geometric characteristics of the surface.

[0060] Mesh reconstruction is performed based on point cloud and curvature information, and triangular facets are used to represent the surface of the object. The reconstruction process uses the spherical azimuth method to determine the connection relationship of each point according to its normal vector and generate a triangular facet mesh. Each triangular facet contains three vertex coordinates and a normal vector. Geometric features are extracted from the generated triangular facets, including the area, normal vector, adjacency relationship, etc. of the facet, to construct the surface topology structure. The surface topology data is processed by surface smoothing interpolation, and the mesh is locally smoothed using the moving least squares method. The smoothing process maintains the overall shape of the surface while reducing local noise. Finally, boundary constraint processing is performed to ensure that the reconstructed surface conforms to anatomical characteristics. The constraint conditions include the continuity and smoothness of the vertebral boundary.

[0061] For example, the CT data of a patient with scoliosis was reconstructed in three dimensions. The original segmented data contained 180 sections, and the resolution of each layer was 1 mm × 1 mm. After voxelization, about 2.8 million voxel points were obtained, of which about 420,000 were surface voxels. The center of the L1 vertebra was selected as the origin for the coordinate system transformation, and the transformation matrix was calculated by least squares registration. The surface point cloud extraction initially obtained 420,000 points, and 150,000 uniformly distributed feature points were retained after density equalization. When calculating the normal vector, an average of 25 neighboring points were considered for each point to obtain a reliable direction estimate. Curvature analysis showed that the principal curvature of the normal vertebral surface was between 0.05 and 0.15, while the curvature value at the scoliosis reached 0.25. Mesh reconstruction generated about 300,000 triangular facets with an average side length of 2.5 mm. The root mean square error after surface smoothing was less than 0.3 mm, while maintaining the key features of the anatomical structure. The final reconstructed three-dimensional model data accurately reflects the spatial morphology of the spine, providing a high-quality geometric basis for subsequent measurement and analysis.

[0062] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0063] (1) Perform spatial gradient calculation on the three-dimensional model data of the spine to obtain surface gradient data, and extract candidate vertebral gap regions from the surface gradient data to obtain a gap feature point set;

[0064] (2) Performing density clustering analysis on the gap feature point set to obtain vertebral separation mark data, and performing vertebral boundary positioning processing on the vertebral separation mark data to obtain vertebral boundary contour data;

[0065] (3) The vertebral body boundary contour data is processed by vertebral end plate surface extraction to obtain end plate feature data, and the vertebral body geometric center is calculated on the end plate feature data to obtain center point coordinate data;

[0066] (4) Performing pedicle area positioning processing on the center point coordinate data to obtain pedicle marking data, and performing key anatomical point extraction processing on the pedicle marking data to obtain anatomical feature point data;

[0067] (5) Perform vertebral posture analysis on the anatomical feature point data to obtain posture parameter data, and construct feature vectors on the posture parameter data to obtain a vertebral feature data set.

[0068] Specifically, the spatial gradient calculation is first performed on the three-dimensional model data of the spine to calculate the spatial gradient value of each point. The gradient value reflects the rate of change of the surface morphology. In three-dimensional space, the gradient calculation takes into account the changes in the three directions of x, y, and z to generate surface gradient data. In the surface gradient data, the vertebral gap area appears as a continuous area with abnormal gradient values. By setting the gradient threshold, the gap candidate area is screened out to obtain the gap feature point set. The gap feature point set is subjected to density clustering analysis, and the density-based spatial clustering method is used to divide similar points into the same category. The clustering parameters include a clustering radius of 3 mm and a minimum number of 50 points to ensure the continuity of the clustering results. The vertebral separation marker data obtained by clustering is used to define the boundary line of adjacent vertebrae, and the complete vertebral boundary contour is extracted by the boundary tracking algorithm.

[0069] The end plate surface features are extracted from the vertebral boundary contour data. The end plate surface is the upper and lower flatter surface of the vertebral body. The extraction process is based on the curvature feature, and the continuous area with a small curvature value is identified as the candidate area of ​​the end plate surface. The geometric center of the identified end plate surface is calculated using the following formula:

[0070]

[0071] in are the geometric center coordinates, is the number of points on the final board surface, is the coordinate of the ith point, is the weight coefficient, reflecting the reliability of the point.

[0072] The center point coordinate data is used to locate the pedicle area. The pedicle is the two protruding structures at the back of the vertebral body. The positioning process uses the following positioning formula:

[0073]

[0074] in The pedicle area, is the point to be determined, The center of the vertebral body, is the maximum distance threshold, is the posterior vector of the cone, Represents the vector (pC) and the posterior vector of the vertebra The angle between and is the angle threshold.

[0075] Vertebral posture analysis is performed based on anatomical feature point data. The posture analysis uses the following feature calculation formula:

[0076]

[0077] in is the attitude matrix, , , are the front-back, left-right, and up-down direction vectors of the vertebral body, respectively. , , ] are the center point coordinates in the local coordinate system.

[0078] For example, the three-dimensional model contains 12 vertebrae, and the spatial gradient calculation shows that the average gradient value of the intervertebral disc area is 0.35, which is significantly higher than the 0.12 of the vertebral surface. Density clustering analysis identified 11 intervertebral disc areas, each containing an average of 180 feature points. When the end plate surface is extracted, the upper end plate surface contains an average of 850 points, and the lower end plate surface contains an average of 820 points. When calculating the geometric center, the weight coefficient is determined according to the distance from the point to the edge of the end plate surface. Pedicle positioning parameter settings: is 20 mm, is 150 degrees, The angle between the anterior-posterior vector and the coronal plane of a normal vertebra is 85-95 degrees, while the angle deviation of the scoliotic vertebra is 15 degrees. The vertebral feature data set finally generated contains complete feature information such as the position, direction, and size of each vertebra, providing an accurate data basis for subsequent Cobb angle measurement.

[0079] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0080] (1) Perform center point connection processing on the vertebral feature data set to obtain initial spinal curve data, and perform curve smoothing interpolation processing on the initial spinal curve data to obtain continuous curve data;

[0081] (2) Calculate the curvature of the continuous curve data to obtain the curve curvature distribution data, and extract the extreme points of the curve curvature distribution data to obtain the curvature extreme point set;

[0082] (3) Performing regional segmentation processing on the curvature extreme value point set to obtain curve partition data, and performing scoliosis area definition processing on the curve partition data to obtain scoliosis interval data;

[0083] (4) The end vertebra position is calculated and processed for the scoliosis interval data to obtain candidate end vertebra data, and the end plate direction is analyzed and processed for the candidate end vertebra data to obtain end plate inclination angle data;

[0084] (5) Performing angle projection calculation on the end plate inclination data to obtain projection angle data, and performing primary and secondary curve distinction processing on the projection angle data to obtain primary curve angle data;

[0085] (6) The compensation coefficient of the main curve angle data is calculated to obtain error compensation data, and the error compensation data is corrected to obtain Cobb angle data.

[0086] Specifically, the vertebral center points in the vertebral feature data set are connected to obtain the initial curve describing the spinal shape. To ensure the continuity and smoothness of the curve, the cubic spline interpolation method is used for smoothing. The interpolation process adds control points between adjacent vertebral center points, and the control point spacing is set to 1 mm to generate continuous and smooth spinal curve data. The curvature of the continuous curve is calculated, and the curvature reflects the degree of curvature of the curve. The local curvature value is calculated for each point on the curve to obtain the curvature distribution along the curve. The extreme points in the curvature distribution data correspond to the positions where the degree of spinal curvature changes most significantly. By finding the local maximum and minimum values ​​of the curvature value, the curvature extreme point set is extracted.

[0087] The curvature extreme point set is segmented into regions, and the region between adjacent extreme points is regarded as a curve segment. The segmented data reflects the curvature characteristics of different regions of the spine. By analyzing the curvature change trend of each segment, the scoliosis region is identified. The scoliosis interval is defined based on the direction and amplitude of the curvature value change to determine the range of each scoliosis.

[0088] The position of the end vertebra is calculated using the following formula:

[0089]

[0090] in is the end vertebral body collection, is the i-th vertebra, For the j The curvature value of a point, k The neighborhood range considered is usually set to 2.

[0091] The end plate direction of the candidate vertebral body is analyzed and the end plate inclination angle is calculated. The angle projection calculation uses the following formula:

[0092]

[0093] in is the projection angle, , is the normal vector of the upper and lower end plates, h is the vertebral height, d is the vertebral distance. The final Cobb angle calculation takes into account the projection error compensation, and the compensation calculation formula is:

[0094]

[0095] in is the compensated Cobb angle, is the surface curvature coefficient, is the side bending angle, is the baseline correction value.

[0096] For example, the vertebral feature dataset contains the coordinates of the center points of 12 vertebrae. The center points are connected to generate an initial curve with a length of 320 mm. The curve smoothing interpolation adds 320 control points with an average spacing of 1 mm. Curvature calculation shows that the curvature value of the normal area is between 0.002 and 0.005, while the curvature value of the scoliotic area reaches 0.015. The curvature extreme value analysis found two significant scoliosis, the main scoliosis is located in the thoracic segment, and the secondary scoliosis is located in the lumbar segment. The upper end vertebra of the main scoliotic area is located at the 8th thoracic vertebra, the lower end vertebra is located at the 12th thoracic vertebra, and the top vertebra is the 10th thoracic vertebra. When calculating the end vertebra position, the neighborhood range k=2 is taken to ensure that sufficient curvature change information is covered. The inclination angle of the upper end vertebra end plate is 26.4 degrees, and the inclination angle of the lower end vertebra end plate is 23.8 degrees. The projection angle calculation takes into account the influence of vertebral height (average 13 mm) and vertebral spacing (average 8 mm). In the compensation calculation, the surface curvature coefficient α is 0.15, the lateral bending direction angle β is 75 degrees, and the baseline correction value γ is 2 degrees. The final calculated primary lateral bending Cobb angle is 50.2 degrees, and the secondary lateral bending Cobb angle is 32.5 degrees.

[0097] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0098] (1) Performing three-dimensional coordinate system conversion processing on the Cobb angle data to obtain standardized spatial data, and performing three-dimensional spatial reconstruction processing on the standardized spatial data to obtain spatial reconstruction data;

[0099] (2) performing sagittal plane projection processing on the spatial reconstruction data to obtain sagittal plane projection data, and performing marker point drawing processing on the sagittal plane projection data to obtain sagittal plane marker data;

[0100] (3) performing coronal plane projection processing on the sagittal plane marking data to obtain coronal plane projection data, and performing curve drawing processing on the coronal plane projection data to obtain coronal plane display data;

[0101] (4) performing cross-sectional slicing processing on the coronal display data to obtain cross-sectional image data, and performing axial rotation processing on the cross-sectional image data to obtain rotation view data;

[0102] (5) performing end vertebra region labeling processing on the rotation view data to obtain labeled image data, and performing angle labeling processing on the labeled image data to obtain angle labeling data;

[0103] (6) performing image synthesis processing on the angle mark data to obtain multi-plane synthesis data, and performing interface layout processing on the multi-plane synthesis data to obtain display layout data;

[0104] (7) Performing statistical processing on the display layout data to obtain statistical analysis data, and performing report template filling processing on the statistical analysis data to obtain a measurement result report.

[0105] Specifically, the Cobb angle data was transformed into a three-dimensional coordinate system, and the data in the local coordinate system was transformed into a standard coordinate system. The standard coordinate system takes the center of the first lumbar vertebra as the origin, and a global reference system is established. The data after coordinate transformation was reconstructed in three-dimensional space. The reconstruction process adopted the voxelization method, and the voxel size was set to 1 mm × 1 mm × 1 mm to ensure the display accuracy. The reconstructed spatial data was projected in the sagittal plane, which is the plane that cuts the human body from front to back. The projection process retains the depth information, and the maximum intensity projection method is adopted. The value of each projection point is the maximum voxel value in that direction. Key anatomical points are marked on the sagittal projection map, including the vertebral center point, the endplate center point, and the pedicle point. The marked points are distinguished by different colors and shapes.

[0106] The sagittal marker data is projected onto the coronal plane, which is a plane that cuts the human body longitudinally from left to right. The projection data draws the spinal midline curve, and the curve uses cubic spline interpolation to achieve smoothness and continuity. The curve is drawn with a dotted line to represent the spinal axis and a solid line to represent the vertebral boundary. Different colors distinguish different scoliosis segments. The coronal display data is sliced ​​cross-sectionally at the specified position. The cross section is a section perpendicular to the spinal axis. The slicing position includes the central layer and endplate layer of each vertebra, and a cross-sectional image sequence is generated. The cross-sectional image is rotated axially to make the anatomical position of the vertebra consistent with the standard position. The rotation angle is determined based on the main axis direction of the vertebra.

[0107] The rotated view data is annotated with the end vertebra area, including the vertebral number, endplate position, and vector direction. The angle mark uses a fan-shaped area to represent the Cobb angle and annotates the specific angle value. The annotation uses a semi-transparent effect to avoid blocking the underlying anatomical structure. The marked image is synthesized in multiple planes, using a 2×2 layout. The upper left is a three-dimensional reconstructed view, the upper right is a sagittal projection, the lower left is a coronal projection, and the lower right is a key cross-sectional sequence. The layout design takes into account clinical observation habits, and the views are coordinated and consistent.

[0108] The layout data is measured and statistically analyzed to generate numerical reports, including parameters such as the inclination angle and rotation angle of each vertebra. The report adopts a standard template format and integrates patient information, image parameters, measurement results, and doctor's advice.

[0109] For example: 3D reconstruction data is displayed in a standard coordinate system with a spatial resolution of 1 mm. The sagittal projection image is 512×512 pixels in size and 48 key points are marked, including 12 pairs of vertebral center points and 24 endplate center points. The length of the spinal curve drawn by the coronal projection is 320 mm. The curve is red for the main scoliosis segment (T8-T12) and blue for the secondary scoliosis segment (L1-L4). The cross-sectional slices generate images at 13 key positions, each with a size of 256×256 pixels and a rotation angle ranging from -15 degrees to +12 degrees. The end vertebrae are marked with a yellow semi-transparent area, the main scoliosis Cobb angle of 50.2 degrees is represented by a red sector, and the secondary scoliosis of 32.5 degrees is represented by a blue sector. The synthesis interface uses a high-definition display resolution of 1920×1080 pixels, and the proportions of each view are coordinated to clearly display the measurement results. The measurement report contains a complete data statistics table, records the detailed parameters of each vertebra, and generates a standardized inspection report document.

[0110] The above describes the intelligent measurement method of scoliosis angle based on spine CT three-dimensional reconstruction in the embodiment of the present application. The following describes the intelligent measurement system of scoliosis angle based on spine CT three-dimensional reconstruction in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the scoliosis angle intelligent measurement system based on spine CT three-dimensional reconstruction includes:

[0111] A segmentation module is used to perform adaptive threshold segmentation and grayscale optimization processing on the acquired spinal CT sequence images to obtain a standardized pre-processed image;

[0112] A recognition module is used to perform adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data;

[0113] A construction module is used to perform three-dimensional mesh construction and surface smoothing on the spinal bone segmentation data to obtain spinal three-dimensional model data;

[0114] A positioning module, used for performing vertebral space identification and key point positioning processing on the three-dimensional model data of the spine to obtain a vertebral feature data set;

[0115] A marking module, used for constructing a spinal midline curve and performing end vertebra marking processing on the vertebral feature data set to obtain Cobb angle data;

[0116] The projection module is used to perform three-dimensional spatial reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report.

[0117] Through the cooperation of the above components, the bone tissue area can be accurately extracted by adaptive threshold segmentation and grayscale optimization processing of the spinal CT sequence images, the interference of background noise and artifacts can be reduced, and clear standardized pre-processed images can be obtained. The segmentation accuracy of the spinal bones is effectively improved by the adaptive bone tissue area recognition and boundary enhancement processing, especially the segmentation effect in complex areas such as the vertebral edge and intervertebral space. Through three-dimensional mesh construction and surface smoothing processing, high-quality spinal three-dimensional model reconstruction is achieved, the continuity and smoothness of the model surface are guaranteed, and a reliable geometric basis is provided for subsequent measurement and analysis. The intervertebral space recognition and key point positioning processing of the spinal three-dimensional model data are carried out, and the anatomical feature points of the vertebra are accurately marked, including the vertebral center point, the end plate center point and the pedicle point, etc. The precise positioning of these feature points is crucial for the measurement of Cobb angles. Through the construction of the spinal midline curve and the end vertebral marking processing, the range and degree of scoliosis are accurately identified, especially in complex scoliosis cases, the end vertebral position of the primary and secondary scoliosis can be reliably determined. Finally, 3D space reconstruction and multi-plane projection processing are used to intuitively display the measurement results, generate standardized measurement reports, and provide comprehensive diagnostic reference information. The entire solution realizes fully automatic processing from image acquisition to final measurement result output, significantly improves measurement efficiency, reduces the subjectivity of manual operation, and ensures the accuracy and repeatability of measurement results.

[0118] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent method for measuring scoliosis angle based on three-dimensional reconstruction of spine CT, characterized in that: The intelligent measurement method for scoliosis angle based on spine CT three-dimensional reconstruction includes: The acquired spinal CT sequence images are subjected to adaptive threshold segmentation and grayscale optimization processing to obtain standardized preprocessed images; Performing adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data; Performing three-dimensional mesh construction and surface smoothing processing on the spinal bone segmentation data to obtain spinal three-dimensional model data; The spine three-dimensional model data is subjected to vertebral space identification and key point positioning processing to obtain a vertebral feature data set, including: performing spatial gradient calculation processing on the spine three-dimensional model data to obtain surface gradient data, and performing vertebral space candidate region extraction processing on the surface gradient data to obtain a gap feature point set; performing density clustering analysis processing on the gap feature point set to obtain vertebral separation mark data, and performing vertebral boundary positioning processing on the vertebral separation mark data to obtain vertebral boundary contour data; performing vertebral end plate surface extraction processing on the vertebral boundary contour data to obtain end plate feature data, and performing vertebral geometric center calculation processing on the end plate feature data to obtain center point coordinate data; performing pedicle region positioning processing on the center point coordinate data to obtain pedicle mark data, and performing key anatomical point extraction processing on the pedicle mark data to obtain anatomical feature point data; performing vertebral posture analysis processing on the anatomical feature point data to obtain posture parameter data, and performing feature vector construction processing on the posture parameter data to obtain a vertebral feature data set; Constructing a spinal midline curve and performing end vertebral marking processing on the vertebral feature data set to obtain Cobb angle data; The Cobb angle data is subjected to three-dimensional spatial reconstruction and multi-plane projection processing to obtain a measurement result report.

2. The intelligent method for measuring scoliosis angle based on spine CT three-dimensional reconstruction according to claim 1 is characterized in that: The method of performing adaptive threshold segmentation and grayscale optimization processing on the acquired spinal CT sequence images to obtain standardized preprocessed images includes: Performing DICOM format parsing processing on the spine CT sequence images to obtain original image data, and performing pixel resolution unification processing on the original image data to obtain uniform resolution image data; Performing window width and window position parameter setting processing on the uniform resolution image data to obtain bone tissue enhanced image data, and performing noise suppression processing on the bone tissue enhanced image data through bilateral filtering to obtain denoised image data; Performing grayscale histogram analysis on the denoised image data to obtain grayscale distribution feature data, and performing grayscale interval division processing on the grayscale distribution feature data to obtain CT value segmentation data; Performing bone tissue region extraction processing on the CT value segmented data to obtain preliminary segmented image data, and performing morphological operation processing on the preliminary segmented image data to obtain boundary optimized image data; Gamma curve correction processing is performed on the boundary optimized image data to obtain a standardized pre-processed image.

3. The intelligent method for measuring scoliosis angle based on spine CT three-dimensional reconstruction according to claim 1 is characterized in that: The step of performing adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data includes: Performing bone tissue density threshold analysis on the standardized preprocessed image to obtain bone density feature data, and performing bone tissue region classification on the bone density feature data to obtain preliminary bone tissue labeling data; Performing edge detection processing on the preliminary bone tissue labeling data to obtain bone tissue contour data, and performing connected domain labeling processing on the bone tissue contour data to obtain bone tissue region data; Performing vertebral contour feature extraction processing on the bone tissue region data to obtain vertebral region feature data, and performing spatial position constraint processing on the vertebral region feature data to obtain vertebral position data; Performing bone tissue morphology analysis processing on the vertebral position data to obtain bone tissue morphology feature data, and performing boundary enhancement processing on the bone tissue morphology feature data to obtain boundary enhanced image data; The boundary enhanced image data is subjected to region growing segmentation processing to obtain target region data, and the target region data is subjected to morphological smoothing processing to obtain spinal bone segmentation data.

4. The intelligent method for measuring scoliosis angle based on spine CT three-dimensional reconstruction according to claim 1, characterized in that: The three-dimensional mesh construction and surface smoothing processing of the spinal bone segmentation data to obtain the three-dimensional model data of the spinal column include: Performing voxel grid division processing on the spinal bone segmentation data to obtain basic grid data, and performing coordinate system conversion processing on the basic grid data to obtain unified coordinate grid data; Performing surface point cloud extraction processing on the unified coordinate grid data to obtain surface feature point data, and performing density equalization processing on the surface feature point data to obtain uniform point cloud data; Performing surface normal vector calculation processing on the uniform point cloud data to obtain surface directional feature data, and performing curvature analysis processing on the surface directional feature data to obtain local curvature data; Performing mesh reconstruction processing on the local curvature data to obtain triangular patch data, and performing geometric feature extraction processing on the triangular patch data to obtain surface topology data; The surface topology data is processed by surface smoothing interpolation to obtain initial reconstruction data, and the initial reconstruction data is processed by boundary constraint to obtain three-dimensional spine model data.

5. The intelligent measurement method of scoliosis angle based on spine CT three-dimensional reconstruction according to claim 1 is characterized in that: The step of constructing a spinal midline curve and performing end vertebral marking processing on the vertebral feature data set to obtain Cobb angle data includes: Performing center point connection processing on the vertebral feature data set to obtain initial spinal curve data, and performing curve smoothing interpolation processing on the initial spinal curve data to obtain continuous curve data; Performing curvature calculation processing on the continuous curve data to obtain curve curvature distribution data, and performing extreme point extraction processing on the curve curvature distribution data to obtain a curvature extreme point set; Performing regional segmentation processing on the curvature extreme value point set to obtain curve partition data, and performing scoliosis region definition processing on the curve partition data to obtain scoliosis interval data; Performing end vertebra position calculation processing on the scoliosis interval data to obtain candidate end vertebra data, and performing end plate direction analysis processing on the candidate end vertebra data to obtain end plate inclination angle data; Performing angle projection calculation processing on the end plate inclination angle data to obtain projection angle data, and performing primary and secondary curve distinction processing on the projection angle data to obtain primary curve angle data; The compensation coefficient calculation process is performed on the main curve angle data to obtain error compensation data, and the error compensation data is subjected to angle correction process to obtain Cobb angle data.

6. The intelligent measurement method of scoliosis angle based on spine CT three-dimensional reconstruction according to claim 1 is characterized in that: The three-dimensional space reconstruction and multi-plane projection processing of the Cobb angle data are performed to obtain a measurement result report, including: Performing a three-dimensional coordinate system conversion process on the Cobb angle data to obtain standardized spatial data, and performing a three-dimensional spatial reconstruction process on the standardized spatial data to obtain spatial reconstruction data; Performing sagittal plane projection processing on the spatial reconstruction data to obtain sagittal plane projection data, and performing marker point drawing processing on the sagittal plane projection data to obtain sagittal plane marker data; Performing coronal plane projection processing on the sagittal plane marking data to obtain coronal plane projection data, and performing curve drawing processing on the coronal plane projection data to obtain coronal plane display data; Performing cross-sectional slicing processing on the coronal plane display data to obtain cross-sectional image data, and performing axial rotation processing on the cross-sectional image data to obtain rotation view data; Performing end vertebra region labeling processing on the rotation view data to obtain labeled image data, and performing angle labeling processing on the labeled image data to obtain angle labeling data; Performing image synthesis processing on the angle mark data to obtain multi-plane synthesis data, and performing interface layout processing on the multi-plane synthesis data to obtain display layout data; The display layout data is subjected to measurement data statistical processing to obtain statistical analysis data, and the statistical analysis data is subjected to report template filling processing to obtain a measurement result report.

7. An intelligent scoliosis angle measurement system based on spine CT three-dimensional reconstruction, used to implement the intelligent scoliosis angle measurement method based on spine CT three-dimensional reconstruction as described in any one of claims 1 to 6, characterized in that: The scoliosis angle intelligent measurement system based on spine CT three-dimensional reconstruction includes: A segmentation module is used to perform adaptive threshold segmentation and grayscale optimization processing on the acquired spinal CT sequence images to obtain a standardized pre-processed image; A recognition module, used for performing adaptive bone tissue region recognition and boundary enhancement processing on the standardized pre-processed image to obtain spinal bone segmentation data; A construction module is used to perform three-dimensional mesh construction and surface smoothing on the spinal bone segmentation data to obtain spinal three-dimensional model data; A positioning module is used to perform vertebral space identification and key point positioning processing on the three-dimensional model data of the spine to obtain a vertebral feature data set, including: performing spatial gradient calculation processing on the three-dimensional model data of the spine to obtain surface gradient data, and performing vertebral space candidate area extraction processing on the surface gradient data to obtain a gap feature point set; performing density clustering analysis processing on the gap feature point set to obtain vertebral separation mark data, and performing vertebral boundary positioning processing on the vertebral separation mark data to obtain vertebral boundary contour data; performing vertebral end plate surface extraction processing on the vertebral boundary contour data to obtain end plate feature data, and performing vertebral geometric center calculation processing on the end plate feature data to obtain center point coordinate data; performing pedicle region positioning processing on the center point coordinate data to obtain pedicle mark data, and performing key anatomical point extraction processing on the pedicle mark data to obtain anatomical feature point data; performing vertebral posture analysis processing on the anatomical feature point data to obtain posture parameter data, and performing feature vector construction processing on the posture parameter data to obtain a vertebral feature data set; A marking module, used for constructing a spinal midline curve and performing end vertebra marking processing on the vertebral feature data set to obtain Cobb angle data; The projection module is used to perform three-dimensional spatial reconstruction and multi-plane projection processing on the Cobb angle data to obtain a measurement result report.

Images