A three-dimensional object segmentation method in three-dimensional body data
By combining dual-energy CT scanning and connectivity analysis of two-dimensional tomography with region growing and boundary methods, the problem of low object segmentation speed and accuracy in three-dimensional volume data was solved, and fast and accurate three-dimensional object segmentation was achieved.
Patent Information
- Application Number
- CN202211217975.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-30
- Publication Date
- 2026-02-24
- Estimated Expiration
- 2042-09-30
AI Technical Summary
Existing technologies struggle to quickly and accurately segment all objects in 3D volumetric data, especially in CT inspection. Traditional methods such as thresholding, region growing, and boundary methods suffer from limitations and inefficiency.
Multiple two-dimensional slices are obtained by dual-energy CT scanning. Connectivity analysis is performed by utilizing the differences in characteristic parameters of the two-dimensional slices. Combined with region growing and boundary methods, three-dimensional object segmentation is performed, reducing the range of region growing from 8 pixels in two dimensions to 26 pixels in three dimensions, thus achieving fast and accurate object segmentation.
It improves the speed and accuracy of 3D object segmentation, reduces the impact of artifacts, avoids iterative calculations, and achieves faster segmentation time and higher segmentation accuracy.
Smart Images
Figure CN115457060B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of CT detection technology, and in particular to a method for segmenting three-dimensional objects in three-dimensional volume data. Background Technology
[0002] CT (Computed Tomography) inspection is a technique that uses X-rays to inspect materials, workpieces, and equipment without damaging the object being inspected. It plays a crucial role in industries such as aerospace, metallurgy, power, petrochemicals, shipbuilding, spaceflight, railways, and automobiles. Accurate segmentation of 3D objects is a key prerequisite for object identification and location. Image segmentation is currently divided into two main categories: one is based on artificial intelligence technology, which, through training, directly segments one or more 3D objects from a training set. This method generally relies on training samples and is heavily dependent on the shape and texture of objects in the training samples, and cannot segment all 3D objects. The other is traditional segmentation methods, commonly including thresholding, 3D region growing, boundary methods, and artificial intelligence. Thresholding is the most common and simplest segmentation method. Its principle is to draw a grayscale histogram of the image and select a threshold from the histogram to classify the image and obtain the segmentation result. Thresholding requires that the grayscale values of the image be uniformly distributed, and that peaks and valleys be clearly observed in the histogram; otherwise, it cannot accurately segment the specified region, and its limitations are significant. Three-dimensional region growing, typically a region growing method, uses a seed point to merge adjacent pixels or regions with similar properties into a new growth point, repeating this process until no further growth is possible. Boundary methods, on the other hand, identify dissimilar pixels to form boundaries, thus separating different regions. Boundary methods are edge-detection-based segmentation algorithms that utilize the differences in features between regions in an image, employing gradient concepts to detect edges and segment the boundaries of different regions. The accuracy of the three-dimensional boundary segmentation results is crucial.
[0003] With advancements in hardware and the emergence of diverse algorithms, the demands on 3D object segmentation time are increasing. Therefore, how to quickly and accurately segment all objects in 3D volumetric data has become a pressing technical problem that needs to be solved by those skilled in the art. Summary of the Invention
[0004] Based on the above analysis, the embodiments of the present invention aim to provide a method for segmenting three-dimensional objects in three-dimensional volume data, in order to solve the problem of how to quickly and accurately segment all objects in three-dimensional volume data.
[0005] On one hand, embodiments of the present invention provide a method for segmenting three-dimensional objects in three-dimensional volume data, the method comprising:
[0006] S1: Projection data of the scanned object at different angles are obtained by dual-energy CT scanning. The projection data is reconstructed to obtain multiple two-dimensional slices, which are then numbered sequentially.
[0007] S2: Segment the two-dimensional objects in each two-dimensional fault, obtain the features of each two-dimensional object in each two-dimensional fault, and number the two-dimensional objects in each two-dimensional fault in sequence.
[0008] S3: Analyze and determine the connectivity of the three-dimensional volume: Based on the parameter differences between the basic two-dimensional objects in the basic two-dimensional fault and the two-dimensional objects in the two-dimensional faults above the basic two-dimensional fault, conduct connectivity analysis to determine the connected regions and complete the three-dimensional volume connectivity analysis; the parameters include center position, area, average density and average atomic number.
[0009] Preferably, step S3 includes: S3.1: using the k-th two-dimensional object M of the i-th two-dimensional fault. ik Based on this, traverse each two-dimensional object in the (i-1)th two-dimensional fault, and determine whether each two-dimensional object in the (i-1)th two-dimensional fault is related to two-dimensional object M. ik The differences in the center position, area, average density, and average atomic number of the two-dimensional object are considered to meet the threshold requirements. If there is a two-dimensional object in the (i-1)th two-dimensional tomography that meets the threshold requirements, then the two-dimensional objects that meet the threshold requirements are comprehensively scored based on the differences in center position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is the one in the (i-1)th two-dimensional tomography that is the same as two-dimensional object M. ik Connected regions with connectivity; if there is no two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, then traverse each two-dimensional object in the ij-th two-dimensional fault and repeat the above steps; where i≥2, k≥1, i≥j≥2; j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirement in the two-dimensional fault with the current value, j takes a value that is 1 greater than the current value.
[0010] S3.2: Repeat step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed; the two-dimensional objects with connectivity are combined to form a three-dimensional object, thus completing the three-dimensional object segmentation.
[0011] Preferably, the comprehensive score is calculated according to formula (I):
[0012] Formula (I);
[0013] Where s0 is a two-dimensional object M ik The area of the two-dimensional object M, ρ0. ik The average density, Z0 is the density of the two-dimensional object M. ikThe average atomic number, x0 and y0 are the two-dimensional object M ik The x and y coordinates of the center position of the two-dimensional object are: s1, ρ1, Z1, x1, and y1. s1 is the area of the two-dimensional object that meets the threshold requirement; ρ1 is the average density of the two-dimensional object that meets the threshold requirement; Z1 is the average atomic number of the two-dimensional object that meets the threshold requirement; x1 and y1 are the x and y coordinates of the center position of the two-dimensional object that meets the threshold requirement, respectively; and a, b, c, and d are parameters greater than 0.
[0014] Preferably, a two-dimensional fault that does not meet the threshold requirement is defined as an error fault. In step S3.1, the number of error faults that can be traversed is A. When j equals A, if there is no two-dimensional object that meets the threshold requirement in the iA-th two-dimensional fault, then the traversal upwards is stopped. Wherein, A≥1 (e.g., A=1, 2 or 3).
[0015] Preferably, the method further includes: in step S3.1, if there are two-dimensional objects that meet the threshold requirements among the traversed two-dimensional faults, after determining the two-dimensional object with the highest comprehensive score, the two-dimensional object M is... ik The number is modified to be the same as the number of the two-dimensional object with the highest comprehensive score; if no two-dimensional object meeting the threshold requirement is found in the traversed two-dimensional faults, then the maximum number of the connectivity region that has already completed the connectivity analysis is incremented by one, and this number is taken as the two-dimensional object M. ik The number.
[0016] Preferably, in step S3.2, repeating step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed includes:
[0017] S3.2.1: Using the (k+1)th two-dimensional object M of the i-th two-dimensional fault. i（k+1） Based on this, repeat step S3.1 until the connectivity analysis of each two-dimensional object in the i-th two-dimensional fault is completed;
[0018] S3.2.2: The k-th two-dimensional object M of the (i+1)-th two-dimensional fault. （i+1）k Based on this, repeat steps S3.1 and S3.2.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed.
[0019] Preferably, the method further includes: in step S1, after obtaining multiple two-dimensional fractures, performing image preprocessing on the two-dimensional fractures.
[0020] Preferably, the image preprocessing includes image smoothing, image enhancement, image region segmentation, and dilation erosion.
[0021] Preferably, in step S2, the method for segmenting the two-dimensional objects in each two-dimensional fault includes the region growing method and the boundary method.
[0022] Preferably, the region growth method includes: traversing each two-dimensional fault and selecting a seed point in each two-dimensional fault, then traversing the neighborhood of the seed point in the two-dimensional fault, and in the neighborhood, points whose difference from the seed point is less than a threshold are used as growth points for neighborhood growth, thereby completing the segmentation of the two-dimensional object in each two-dimensional fault.
[0023] Preferably, the method further includes: step S4: acquiring the features of each three-dimensional object.
[0024] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:
[0025] 1. This invention uses each two-dimensional object in a two-dimensional tomographic sample as the connectivity assessment object, and finally combines the connected two-dimensional objects to form a three-dimensional object. In other words, this invention is a two-dimensional region growing method. Compared to existing three-dimensional region growing methods, this invention reduces the area to be grown from 26 pixels in three dimensions to 8 pixels in two dimensions. Therefore, the three-dimensional object segmentation method of this invention is faster.
[0026] 2. Compared with the threshold method in the prior art, which cannot accurately segment all objects due to artifacts in CT images, the present invention uses two-dimensional region growing, which can accommodate differences and reduce the impact of artifacts. Therefore, the present invention has higher accuracy in segmenting three-dimensional objects.
[0027] 3. Compared with the boundary method of the prior art, which requires first-order or second-order differential calculations and iterative operations, the present invention does not require any iteration. Therefore, the three-dimensional object segmentation method of the present invention is faster.
[0028] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description
[0029] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.
[0030] Figure 1 This is a flowchart of the three-dimensional object segmentation method in three-dimensional volume data according to the present invention;
[0031] Figure 2 This is a schematic diagram showing the changes in the numbering of each two-dimensional object before and after connectivity analysis using the three-dimensional object segmentation method in the three-dimensional volume data of the present invention. Detailed Implementation
[0032] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.
[0033] A specific embodiment of the present invention discloses a method for segmenting three-dimensional objects in three-dimensional volume data, such as... Figure 1 As shown. The method includes:
[0034] S1: Projection data of the scanned object at different angles (e.g., 0 to 360 degrees, with an angle set at 1 degree interval) are obtained by dual-energy CT scanning. The projection data are reconstructed and inverted to obtain high-energy, low-energy, density and atomic number data images, resulting in multiple two-dimensional cross-sections of the three-dimensional volume data. The two-dimensional cross-sections are numbered sequentially.
[0035] Furthermore, after obtaining multiple two-dimensional slices, image preprocessing is performed on the two-dimensional slices to improve the signal-to-noise ratio of the image, making the segmentation result more accurate.
[0036] Specifically, the image preprocessing includes image smoothing, image enhancement, image region segmentation, and dilation erosion.
[0037] S2: Segment the two-dimensional objects in each two-dimensional fault, obtain the features of each two-dimensional object in each two-dimensional fault, and number the two-dimensional objects in each two-dimensional fault in sequence.
[0038] Specifically, the method for segmenting the two-dimensional objects in each two-dimensional fault can be a conventional method in the field, such as the region growing method or the boundary method.
[0039] For example, the region growth method includes: traversing each two-dimensional fault and selecting a seed point in each two-dimensional fault, then traversing the neighborhood of the seed point in the two-dimensional fault, and in the neighborhood, points whose difference from the seed point is less than a threshold are used as growth points for neighborhood growth, thereby completing the segmentation of the two-dimensional object in each two-dimensional fault.
[0040] For example, the difference threshold for seed points in density image segmentation is 0.1.
[0041] The features of the two-dimensional object include mean density, root mean square deviation of density, mean atomic number, root mean square deviation of atomic number, object area, object perimeter, horizontal coordinate of the centroid of the region, vertical coordinate of the centroid of the region, and minimum bounding rectangle.
[0042] S3: Analyze and determine the connectivity of the three-dimensional volume: Based on the parameter differences between the basic two-dimensional objects in the basic two-dimensional fault and the two-dimensional objects in the two-dimensional faults above the basic two-dimensional fault, conduct connectivity analysis to determine the connected regions and complete the three-dimensional volume connectivity analysis; the parameters include center position, area, average density and average atomic number.
[0043] S3.1: The k-th two-dimensional object M of the i-th two-dimensional fault. ik Based on this, traverse each two-dimensional object in the (i-1)th two-dimensional fault, and determine whether each two-dimensional object in the (i-1)th two-dimensional fault is related to two-dimensional object M. ik The differences in the center position, area, average density, and average atomic number of the two-dimensional object are considered to meet the threshold requirements. If there is a two-dimensional object in the (i-1)th two-dimensional tomography that meets the threshold requirements, then the two-dimensional objects that meet the threshold requirements are comprehensively scored based on the differences in center position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is the one in the (i-1)th two-dimensional tomography that is the same as two-dimensional object M. ik Connected regions with connectivity; if there is no two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, then traverse each two-dimensional object in the ij-th two-dimensional fault and repeat the above steps; where i≥2, k≥1, i≥j≥2; j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirement in the two-dimensional fault with the current value, j takes a value that is 1 greater than the current value.
[0044] It should be noted that i≥2, meaning that the connectivity of the first two-dimensional fault does not need to be analyzed, but is directly used as an existing connected region, and the connectivity analysis is started directly based on the two-dimensional objects in the second two-dimensional fault.
[0045] For example, the difference thresholds for the center position, area, average density, and average atomic number are 14, 50, 0.1, and 1, respectively.
[0046] For example, the comprehensive score is calculated according to formula (I):
[0047] Formula (I);
[0048] Where s0 is a two-dimensional object M ik The area of the two-dimensional object M, ρ0. ik The average density, Z0 is the density of the two-dimensional object M. ik The average atomic number, x0 and y0 are the two-dimensional object M ikThe x and y coordinates of the center position of the two-dimensional object are: s1, ρ1, Z1, x1, and y1. s1 is the area of the two-dimensional object that meets the threshold requirement; ρ1 is the average density of the two-dimensional object that meets the threshold requirement; Z1 is the average atomic number of the two-dimensional object that meets the threshold requirement; x1 and y1 are the x and y coordinates of the center position of the two-dimensional object that meets the threshold requirement, respectively; and a, b, c, and d are parameters greater than 0.
[0049] The values of a, b, c, and d are weighted proportions, determined according to the difference threshold of the corresponding features. For example, a is 14 / 4, b is 50 / 4, c is 0.1 / 4, and d is 1 / 4, indicating that the four have the same weight under the above difference threshold.
[0050] In this invention, a two-dimensional fault that does not meet the threshold requirement is defined as an error fault. In step S3.1, the number of error fault layers that can be traversed is A. When j equals A, if there is no two-dimensional object that meets the threshold requirement in the iA-th two-dimensional fault, then the traversal upwards is stopped. Wherein, A≥1 (e.g., A=1, 2 or 3).
[0051] In other words, in step S3.1, if there is a two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, it is comprehensively scored to determine the connected region; if there is no two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, each two-dimensional object in the ij-th two-dimensional fault is traversed; i≥j≥2, j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirement in the two-dimensional fault where the current j value is taken, j takes a value 1 greater than the current value, until a two-dimensional object (connected region) that meets the threshold requirement is found or j=A is found; if there is a two-dimensional object that meets the threshold requirement in the two-dimensional fault where the current j value is taken, the current traversal ends, and if no two-dimensional object that meets the threshold requirement is found, the current traversal ends.
[0052] Furthermore, if any two-dimensional object in the traversed two-dimensional faults meets the threshold requirement, after determining the two-dimensional object with the highest comprehensive score, the two-dimensional object M is selected. ik The number is modified to be the same as the number of the two-dimensional object with the highest comprehensive score; if no two-dimensional object meeting the threshold requirement is found in the traversed two-dimensional faults, then the maximum number of the connectivity region that has already completed the connectivity analysis is incremented by one, and this number is taken as the two-dimensional object M. ik The number.
[0053] It should be noted that each two-dimensional object in the first fault can be regarded as a connected region that has already undergone connectivity analysis.
[0054] S3.2: Repeat step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional section is completed; connected two-dimensional objects combine to form a three-dimensional object, completing the three-dimensional object segmentation. This includes:
[0055] S3.2.1: Using the (k+1)th two-dimensional object M of the i-th two-dimensional fault. i（k+1） Based on this, repeat step S3.1 until the connectivity analysis of each two-dimensional object in the i-th two-dimensional fault is completed;
[0056] S3.2.2: The k-th two-dimensional object M of the (i+1)-th two-dimensional fault. （i+1）k Based on this, repeat steps S3.1 and S3.2.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed.
[0057] To further clarify the analysis of connectivity in steps S1-S3, Figure 2 Let's take an example to illustrate.
[0058] S1: Four two-dimensional tomographic images were obtained through dual-energy CT scanning (DEXC). Figure 2 Only four two-dimensional faults are shown for ease of explanation, and the four two-dimensional faults are numbered from top to bottom as 1, 2, 3, and 4.
[0059] S2: Segment the two-dimensional objects in each two-dimensional fault and obtain the features of each two-dimensional object in each two-dimensional fault. Number the two-dimensional objects in the first two-dimensional fault as 1, 2, 3, 4; number the two-dimensional objects in the second two-dimensional fault as 1, 2, 3; number the two-dimensional objects in the third two-dimensional fault as 1, 2; and number the two-dimensional objects in the fourth two-dimensional fault as 1, 2, 3.
[0060] S3: Analyze and determine the connectivity of the 3D volume:
[0061] S3.1: First analyze the connectivity of the second two-dimensional fault, taking the first two-dimensional object M of the second two-dimensional fault as an example. 21 Based on this, we iterate through each two-dimensional object in the first two-dimensional fault. If the first two-dimensional object in the first two-dimensional fault satisfies the threshold requirement and has the highest overall score, then the first two-dimensional object in the first two-dimensional fault is two-dimensional object M. 21 The connected region of the two-dimensional object M 21 The numbering is changed to be the same as the numbering of the first two-dimensional object in the first two-dimensional fault (their numbers are both 1, so no change is needed here); with the second two-dimensional object M in the second two-dimensional fault... 22Based on this, we iterate through each two-dimensional object in the first two-dimensional fault. If the third two-dimensional object in the first two-dimensional fault satisfies the threshold requirement and has the highest overall score, then the third two-dimensional object in the first two-dimensional fault is two-dimensional object M. 22 The connected region of the two-dimensional object M 22 The numbering was changed to be the same as the numbering of the third two-dimensional object in the first two-dimensional fault, that is, two-dimensional object M 22 The number is changed to 3; the third two-dimensional object M of the second two-dimensional fault is used. 23 Based on this, we iterate through each two-dimensional object in the first two-dimensional fault. If there are no two-dimensional objects in the first two-dimensional fault that meet the threshold requirement, and the maximum number of two-dimensional objects in the first two-dimensional fault is 4, then we set two-dimensional object M... 23 The number was changed to 5; the connectivity analysis of the second two-dimensional fault was completed.
[0062] Next, analyze the connectivity of the third two-dimensional fault, taking the first two-dimensional object M of the third two-dimensional fault as an example. 31 Based on this, we iterate through each two-dimensional object in the second two-dimensional fracture. If no two-dimensional object in the second two-dimensional fracture meets the threshold requirement, we then iterate through each two-dimensional object in the first two-dimensional fracture. The second two-dimensional object in the first two-dimensional fracture meets the threshold requirement and has the highest overall score; that is, the second two-dimensional object in the first two-dimensional fracture is two-dimensional object M. 31 The connected region of the two-dimensional object M 31 The numbering of object M is changed to be the same as that of the second two-dimensional object in the first two-dimensional fault, that is, two-dimensional object M 31 The number is changed to 2; the second two-dimensional object M of the third two-dimensional fault is used. 32 Based on this, we iterate through each two-dimensional object in the second two-dimensional fracture. The two-dimensional object numbered 5 in the second two-dimensional fracture meets the threshold requirement and has the highest comprehensive score. That is, the two-dimensional object numbered 5 in the second two-dimensional fracture is two-dimensional object M. 32 Connected regions, two-dimensional object M 32 The number was changed to 5; the connectivity analysis of the third two-dimensional fault was completed.
[0063] Next, we analyze the connectivity of the fourth two-dimensional fault, taking the first two-dimensional object M of the fourth two-dimensional fault as an example. 41Based on this, we iterate through each two-dimensional object in the third two-dimensional fault. If no two-dimensional object in the third two-dimensional fault meets the threshold requirement, we iterate through each two-dimensional object in the second two-dimensional fault. If no two-dimensional object in the second two-dimensional fault meets the threshold requirement, we iterate through each two-dimensional object in the first two-dimensional fault. If no two-dimensional object in the first two-dimensional fault meets the threshold requirement, we determine that no two-dimensional object meeting the threshold requirement has been found, and the maximum number of the previously found connected region is 5. Then, we set the two-dimensional object M to... 41 The numbering was changed to 6; the second two-dimensional object M of the fourth two-dimensional fault was used. 42 Based on this, we iterate through each two-dimensional object in the third two-dimensional fault. If no two-dimensional object in the third two-dimensional fault meets the threshold requirement, we iterate through each two-dimensional object in the second two-dimensional fault. If no two-dimensional object in the second two-dimensional fault meets the threshold requirement, we iterate through each two-dimensional object in the first two-dimensional fault. If no two-dimensional object in the first two-dimensional fault meets the threshold requirement, we determine that no two-dimensional object meeting the threshold requirement has been found, and the maximum number of the previously found connected regions is 6. Then, we set the two-dimensional object M to the next level. 42 The numbering was changed to 7; the third two-dimensional object M of the fourth two-dimensional fault was used. 43 Based on this, we iterate through each two-dimensional object in the third two-dimensional fault. The two-dimensional object numbered 5 in the third two-dimensional fault satisfies the threshold requirement and has the highest overall score. Therefore, we select two-dimensional object M. 43 The number is changed to 5. That is, the connectivity analysis of all two-dimensional objects in all two-dimensional faults is completed.
[0064] The present invention also includes step S4: acquiring the features of each three-dimensional object. These features are used to determine the performance of the three-dimensional objects.
[0065] The features of the three-dimensional object include spatial location, volume, average density, average atomic number, root mean square deviation of density, and root mean square deviation of atomic number. The corresponding features of the three-dimensional object are as follows:
[0066] ;
[0067] ;
[0068] ;
[0069] ;
[0070] ;
[0071] Among them, S i Voxel represents the area corresponding to the i-th pixel of a 3D object, and Voxel represents the pixel size of that pixel. and These represent the density value and atomic number value corresponding to the pixel, respectively; V, , , , These represent the volume, average density, average atomic number, root mean square deviation of density, and root mean square deviation of atomic number of a three-dimensional object, respectively.
[0072] To quickly obtain the root mean square error characteristics, the density root mean square error and the atomic number root mean square error can be simplified as follows: and :
[0073] ;
[0074] Among them, S k,i , , These represent the area, density, and atomic number of the i-th pixel in the k-th (k=1-K) layer slice, respectively.
[0075] Compared to existing technologies, this invention uses each two-dimensional object in a two-dimensional tomographic image as the connectivity assessment object, and finally combines the connected two-dimensional objects to form a three-dimensional object; that is, this invention is a two-dimensional region growing method. Compared to the existing three-dimensional region growing method, this invention reduces the area to be grown from 26 pixels in three dimensions to 8 pixels in two dimensions. Therefore, the three-dimensional object segmentation method of this invention is faster. Compared to the existing thresholding method, which cannot accurately segment all objects due to artifacts in CT images, this invention uses two-dimensional region growing, which can accommodate differences and reduce the impact of artifacts. Therefore, this invention has higher accuracy in three-dimensional object segmentation. Compared to the existing boundary method, which requires first- or second-order differential calculations and iterative operations, this invention does not require any iterations. Therefore, the three-dimensional object segmentation method of this invention is faster.
[0076] The method for segmenting three-dimensional objects in three-dimensional volume data according to the present invention is further illustrated below through specific embodiments.
[0077] Example 1
[0078] A method for segmenting 3D objects from 3D volume data, comprising:
[0079] S1: Projection data of the scanned object at different angles (0 to 360 degrees, with one angle set at 1 degree intervals) are obtained by dual-energy CT scanning. The projection data are reconstructed and inverted to obtain high-energy, low-energy, density and atomic number data images, thereby obtaining multiple two-dimensional cross-sections of the three-dimensional volume data. The two-dimensional cross-sections are preprocessed, including image smoothing, image enhancement, image region segmentation and dilatation erosion. The two-dimensional cross-sections are numbered sequentially.
[0080] S2: The two-dimensional objects in each two-dimensional fault are segmented using the region growing method, and the characteristics of each two-dimensional object in each two-dimensional fault are obtained, including the mean density, the root mean square of density, the mean atomic number, the root mean square of atomic number, the object area, the object perimeter, the abscissa of the region centroid, the ordinate of the region centroid, and the minimum bounding rectangle; the two-dimensional objects in each two-dimensional fault are numbered sequentially.
[0081] S3: Analyze and determine the connectivity of the 3D volume:
[0082] S3.1: The k-th two-dimensional object M of the i-th two-dimensional fault. ik Based on this, traverse each two-dimensional object in the (i-1)th two-dimensional fault, and determine whether each two-dimensional object in the (i-1)th two-dimensional fault is related to two-dimensional object M. ik The differences in the center position, area, average density, and average atomic number of the two-dimensional object M are compared to the threshold requirements (e.g., thresholds of 14, 50, 0.1, and 1 for the center position, area, average density, and average atomic number, respectively). If there is a two-dimensional object in the (i-1)th two-dimensional tomography that meets the threshold requirements, then the two-dimensional object that meets the threshold requirements is comprehensively scored based on the differences in center position, area, average density, and average atomic number. The smaller the difference, the higher the score. The two-dimensional object with the highest comprehensive score is the one in the (i-1)th two-dimensional tomography that is the same as two-dimensional object M. ik Connected regions with connectivity; if there is no two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, then traverse each two-dimensional object in the ij-th two-dimensional fault and repeat the above steps; where i≥2, k≥1, i≥j≥2; j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirement in the two-dimensional fault with the current value, j takes a value that is 1 greater than the current value.
[0083] The comprehensive score is calculated according to formula (I): Formula (I);
[0084] Where a is 14 / 4, b is 50 / 4, c is 0.1 / 4, and d is 1 / 4.
[0085] In this invention, a two-dimensional fault that does not meet the threshold requirement is defined as an error fault. In step S3.1, the number of error faults that can be traversed is A. When j equals A, if there is still no two-dimensional object that meets the threshold requirement in the iA-th two-dimensional fault, then the traversal upwards is stopped; where A=3.
[0086] If there is a two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, then a comprehensive score is given to determine the connected region; if there is no two-dimensional object that meets the threshold requirement in the (i-1)th two-dimensional fault, then each two-dimensional object in the ij-th two-dimensional fault is traversed; i≥j≥2, j takes values from small to large, and only when there is no two-dimensional object that meets the threshold requirement in the two-dimensional fault with the current value of j, will j take a value 1 greater than the current value, until a two-dimensional object that meets the threshold requirement (connected region) is found or j=A; if there is a two-dimensional object that meets the threshold requirement in the two-dimensional fault with the current value of j, then the traversal stops.
[0087] If any two-dimensional object meets the threshold requirement among the traversed two-dimensional faults, after determining the two-dimensional object with the highest comprehensive score, select two-dimensional object M. ik The number is modified to be the same as the number of the two-dimensional object with the highest comprehensive score; if no two-dimensional object meeting the threshold requirement is found in the traversed two-dimensional faults, then the maximum number of the connectivity region that has already completed the connectivity analysis is incremented by one, and this number is taken as the two-dimensional object M. ik The number.
[0088] Each two-dimensional object in the first fault is considered as a connected region for which connectivity analysis has been completed.
[0089] S3.2: Repeat step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional section is completed; connected two-dimensional objects combine to form a three-dimensional object, completing the three-dimensional object segmentation. This includes:
[0090] S3.2.1: Using the (k+1)th two-dimensional object M of the i-th two-dimensional fault. i（k+1） Based on this, repeat step S3.1 until the connectivity analysis of each two-dimensional object in the i-th two-dimensional fault is completed;
[0091] S3.2.2: The k-th two-dimensional object M of the (i+1)-th two-dimensional fault. （i+1）k Based on this, repeat steps S3.1 and S3.2.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed.
[0092] S4: Obtain the features of each 3D object. Used to determine the performance of 3D objects.
[0093] In this embodiment, on a 2.4GHz computer, for a The three-dimensional volume data was used to perform three-dimensional object segmentation. The three-dimensional object segmentation was completed in 5 seconds, and 11 types of three-dimensional objects were segmented.
[0094] Comparative Example 1
[0095] The same scanned document as in Example 1 was scanned, and the three-dimensional objects in the three-dimensional volume data of the scanned document were segmented using the existing three-dimensional region growing method. On a 2.4GHz computer, for one The three-dimensional volume data and the region growth process takes about 50 seconds to complete.
[0096] Comparative Example 2
[0097] The same scanned document as in Example 1 was scanned, and the three-dimensional objects in the three-dimensional volume data of the scanned document were segmented using the thresholding method in the prior art. On a 2.4 GHz computer, for one With the three-dimensional volume data, the thresholding method can only segment six types of three-dimensional objects.
[0098] Comparative Example 3
[0099] The same scanned document as in Example 1 was scanned, and the three-dimensional objects in the three-dimensional volume data of the scanned document were segmented using the boundary method in the prior art. On a 2.4 GHz computer, for one The three-dimensional volume data was obtained using the boundary method in 45 seconds.
[0100] A comparison of the results of Example 1 with those of Comparative Examples 1 and 3 shows that, compared with existing 3D region growing methods and boundary methods, the method of the present invention can complete 3D object segmentation more quickly; a comparison of the results of Example 1 with those of Comparative Example 2 shows that, compared with existing threshold methods, the method of the present invention has higher accuracy in segmenting 3D objects.
[0101] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.
[0102] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A three-dimensional object segmentation method in three-dimensional volume data, characterized by, The method includes: S1: Projection data of the scanned object at different angles are obtained by dual-energy CT scanning. The projection data is reconstructed to obtain multiple two-dimensional slices, which are then numbered sequentially. S2: Segment the two-dimensional objects in each two-dimensional fault, obtain the features of each two-dimensional object in each two-dimensional fault, and number the two-dimensional objects in each two-dimensional fault in sequence. S3: Analyze and determine the connectivity of the three-dimensional volume: Based on the parameter differences between the basic two-dimensional object in the basic two-dimensional fault and the two-dimensional object in the two-dimensional fault above the basic two-dimensional fault, perform connectivity analysis to determine the connected regions and complete the three-dimensional volume connectivity analysis; the parameters include center position, area, average density, and average atomic number; wherein, step S3 includes: S3.1: the kth two-dimensional object M in the ith two-dimensional slice ik , traverse each two-dimensional object in the i-1th two-dimensional slice, and determine whether the difference between the center position, area, average density, and average atomic number of each two-dimensional object in the i-1th two-dimensional slice and the two-dimensional object M ik meets the threshold requirement; if there is a two-dimensional object in the i-1th two-dimensional slice that meets the threshold requirement, then according to the differences in center position, area, average density, and average atomic number, a comprehensive score is given to the two-dimensional object that meets the threshold requirement, the smaller the difference, the higher the score, and the two-dimensional object with the highest comprehensive score is the two-dimensional object in the i-1th two-dimensional slice that is most similar to the two-dimensional object M ik ; if there is no two-dimensional object in the i-1th two-dimensional slice that meets the threshold requirement, then traverse each two-dimensional object in the i-jth two-dimensional slice and repeat the above steps; where i≥2, k≥1, i≥j≥2; j takes values from small to large, and only when the current two-dimensional slice does not have a two-dimensional object that meets the threshold requirement, j takes a value that is 1 greater than the current value. S3.2: Repeat step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed; the two-dimensional objects with connectivity are combined to form a three-dimensional object, thus completing the three-dimensional object segmentation.
2. The method of claim 1, wherein, The comprehensive score is calculated according to formula (I): Formula (I); wherein s0 is the area of the two-dimensional object M ik , p0 is the average density of the two-dimensional object M ik , Z0 is the average atomic number of the two-dimensional object M ik , x0 and y0 are the horizontal and vertical coordinates of the center position of the two-dimensional object M ik , s1 is the area of the two-dimensional object satisfying the threshold requirement, p1 is the average density of the two-dimensional object satisfying the threshold requirement, Z1 is the average atomic number of the two-dimensional object satisfying the threshold requirement, x1 and y1 are the horizontal and vertical coordinates of the center position of the two-dimensional object satisfying the threshold requirement, and a, b, c, and d are weight parameters greater than 0, and the specific values are: d = center position difference threshold / 4, a = area difference threshold / 4, b = average density difference threshold / 4, c = average atomic number difference threshold / 4, used to balance the influence of the differences of each parameter on connectivity judgment.
3. The method of claim 1, wherein, Two-dimensional faults that do not meet the threshold requirements are defined as error faults. In step S3.1, the number of error faults that can be traversed is A. When j equals A, if there are no two-dimensional objects that meet the threshold requirements in the iA-th two-dimensional fault, then the traversal upwards is stopped; where A≥1.
4. The method of claim 3, wherein, The method further comprises: in step S3.1, if there is a two-dimensional object satisfying the threshold requirement in the traversed two-dimensional fault, after determining the two-dimensional object with the highest comprehensive score, modifying the number of the two-dimensional object M ik to the same number as the two-dimensional object with the highest comprehensive score; if no two-dimensional object satisfying the threshold requirement is found in the traversed two-dimensional fault all the time, adding one to the maximum number of the connectivity region which has completed connectivity analysis to serve as the number of the two-dimensional object M ik .
5. The method of claim 1, wherein, In step S3.2, repeating step S3.1 until the connectivity analysis of each two-dimensional object in each two-dimensional fault is completed includes: S3.2.1 : with the (k+1)th two-dimensional object M of the ith two- dimensional fault i（k+1） As a basis, step S3.1 is repeated until the connectivity analysis of each two- dimensional object in the ith two-dimensional fault is complete; S3.2.2: with the (k+1)th two-dimensional object M of the (i+1)th two-dimensional slice （i+1）k S3.1 and step S3.2.1 are repeated until the connectivity analysis of each two- dimensional object in each two-dimensional slice is completed.
6. The method of claim 1, wherein, The method further includes: in step S1, after obtaining multiple two-dimensional fractures, performing image preprocessing on the two-dimensional fractures.
7. The method of claim 6, wherein, The image preprocessing includes image smoothing, image enhancement, image region segmentation, and dilation erosion.
8. The method according to claim 1, characterized in that, In step S2, the methods for segmenting the two-dimensional objects in each two-dimensional fault include the region growing method and the boundary method.
9. The method according to any one of claims 1-8, characterized in that, The method further includes: step S4: obtaining the features of each three-dimensional object.