CT image calcification data generation method and system, terminal and medium

Abstract

The invention provides a CT image calcification data generation method and system, and the method comprises the steps: carrying out the calcification region segmentation of a calcified CT image A, obtaining a calcification region image with calcification artifacts and a calcification region image without calcification artifacts, and building a calcification database; extracting a blood vessel region and a center line thereof in the normal CT image B; randomly extracting a calcification region image in the calcification database, and matching with the target blood vessel region based on the center line to obtain a calcification plaque; and reconstructing the calcified plaques to obtain comparison CT images with calcified artifacts and comparison CT images without calcified artifacts. According to the invention, artifacts generated in real CT scanning of the calcification area can be effectively simulated, and related data of corresponding calcification itself can be obtained at the same time; the problem that real calcification data after artifact removal cannot be obtained through existing clinical scanning can be effectively solved, and a basis is provided for a subsequent deep learning solution.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and specifically, to a method, a system, a terminal, and a medium for generating CT image calcification data. Background Art

[0002] Cardiac computed tomography angiography (CCTA) is used to identify significant stenoses in a patient's coronary arteries and is recommended as a valuable non-invasive alternative for assessing coronary artery disease (CAD). CCTA can acquire high-resolution imaging data. Although it is widely used in the assessment of coronary artery disease, it also has limitations, making the detection of calcified lesions inaccurate. The presence of calcified lesions can lead to the appearance of calcification artifacts (or blooming artifacts, calcium blooming artifact) in CT images, causing doctors to misjudge the degree of lumen stenosis, thus greatly reducing the specificity and positive predictive value (PPV).

[0003] To suppress calcification artifacts, different strategies have been proposed, including improving hardware resolution, using dual-energy CT, improving CT reconstruction algorithms, optimizing image post-processing algorithms, etc., so as to improve the specificity and positive predictive value. In recent years, the development of deep learning has provided new ideas for solving this problem. Compared with the above traditional methods, deep learning has powerful feature extraction capabilities and excellent non-linear fitting capabilities, showing great potential in the field of medical image processing, and convolutional neural networks have been proven to perform well in tasks such as medical image segmentation and classification. However, deep learning is a data-based model. CCTA and CAG (coronary angiography) belong to two generation methods, and CCTA cannot directly obtain the contrast images with and without artifacts.

[0004] In response to the above problems, an existing method is to use a phantom to simulate calcification. First, plaque models with different diameters are created, and the CT attenuation of these models is similar to that of calcification. Then, these models are inserted into a 3D-printed patient coronary artery model to simulate different degrees of coronary artery stenosis. Subsequently, the phantom can be subjected to CT scanning and reconstructed image operations to determine the calcification artifacts. However, the shape and size of the phantom are relatively fixed and are very different from actual clinical calcifications, and the phantom is not easily obtained. Therefore, this method is difficult to generate sufficient data for research. Summary of the Invention

[0005] In view of the above deficiencies in the prior art, the present invention provides a method, a system, a terminal, and a medium for generating CT image calcification data.

[0006] According to one aspect of the present invention, there is provided a method for generating CT image calcification data, including:

[0007] Segment the calcified regions of the CT image A with calcifications to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts, and establish a calcification database;

[0008] Extract the vascular regions and their centerlines from the normal CT image B;

[0009] Randomly select the calcified region images from the calcification database, and match them with the target vascular regions based on the centerlines to obtain calcified plaques;

[0010] Reconstruct the calcified plaques to obtain contrast CT images with and without calcification artifacts respectively, and complete the automatic generation of calcification data.

[0011] Preferably, the segmenting the calcified regions of the CT image A with calcifications to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts, and establishing a calcification database includes:

[0012] Obtain the CT image A with calcifications;

[0013] Use different thresholds to segment the calcified regions of the CT image A to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts respectively;

[0014] Perform data screening on the obtained calcified region images, and label the branches and specific segments where the screened calcified region images are located;

[0015] Correspond one by one the calcified region images obtained by segmentation with different thresholds and having the same calcifications, and establish a calcification database.

[0016] Preferably, the performing data screening on the obtained calcified region images includes:

[0017] Divide the pixel points labeled as calcifications in the calcified region images of a single patient obtained by segmentation with different pixel value thresholds into different connected components to obtain multiple calcified sub-region images;

[0018] Statistically calculate the pixel sizes of each calcified sub-region image, and retain the calcified sub-region images whose number of pixels meets the set requirements to form the required calcified region images, and complete the data screening of the calcified region images.

[0019] Preferably, the extracting the vascular regions and their centerlines from the normal CT image B includes:

[0020] Obtain the normal CT image B;

[0021] Based on the CT image B, extract the cardiac region excluding the external region of the heart;

[0022] Based on the cardiac region, a segmentation threshold for differentiating blood vessels from other tissues is selected, and irrelevant regions are filtered out to obtain a blood vessel region;

[0023] Based on the blood vessel region, using a centerline extraction tool, the RAS coordinates of the centerline control points of the blood vessel region are obtained and converted into pixel coordinates to obtain a centerline result.

[0024] Preferably, the method of randomly extracting calcified region images from the calcification database and matching them with the target blood vessel region based on the centerline to obtain calcified plaques includes:

[0025] Select target points around the centerline of the target blood vessel region;

[0026] Randomly select calcified region images from the calcification database that are in the same branch and / or segment as the centerline control points, and place the calcified region images at the target points such that at least more than half of the calcified region is inside the target blood vessel region, and the calcified region outside the target blood vessel region is removed to complete the matching of the calcified region with the target blood vessel region and obtain calcified plaques.

[0027] Preferably, the method of reconstructing the calcified plaques to obtain contrast CT images with and without calcification artifacts respectively includes:

[0028] Reconstruct the calcified plaques using different CT reconstruction algorithms respectively to obtain a set of contrast CT images with and without calcification artifacts.

[0029] Preferably, the method of reconstructing the calcified plaques using different CT reconstruction algorithms respectively includes:

[0030] Reconstruct the calcified plaques using a filtered back-projection algorithm to obtain preliminary CT images with and without calcification artifacts;

[0031] Interpolate the CT images with and without calcification artifacts to obtain CT images with and without calcification artifacts.

[0032] According to another aspect of the present invention, a system for generating CT image calcification data is provided, including:

[0033] A calcification segmentation module for segmenting the calcified region of a CT image A with calcification to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts, and establishing a calcification database;

[0034] A blood vessel extraction module for extracting the blood vessel region and its centerline from a normal CT image B;

[0035] A simulated calcification generation module, which is used to randomly extract a calcification region image from the calcification database, and match it with a target blood vessel region based on the center line to obtain a calcified plaque;

[0036] A CT image reconstruction module is used to reconstruct the calcified plaque, obtain contrast CT images with and without calcification artifacts, and complete the automatic generation of calcification data.

[0037] According to a third aspect of the present invention, there is provided a computer terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, can be used to execute any of the above-described methods of the present invention, or to run any of the above-described systems of the present invention.

[0038] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to execute any of the methods described above in the present invention, or to run any of the systems described above in the present invention.

[0039] Due to the adoption of the above technical solution, the present invention has at least one of the following beneficial effects compared with the prior art:

[0040] The method, system, terminal and medium for generating CT image calcification data provided by the present invention realize automatic data generation based on real clinical data, and based on the real CT data acquisition and reconstruction process, can effectively simulate the artifacts generated by the calcification area when scanning the real CT image, and simultaneously obtain the relevant data of the corresponding calcification itself.

[0041] The method, system, terminal and medium for generating calcification data of CT images provided by the present invention can simultaneously generate control data with and without calcification artifacts. The former can effectively restore the effect of calcification artifacts, while the actual size and shape of the corresponding calcification can be obtained from the latter, thereby assisting in evaluating the impact of various solutions on calcification, and can be used for subsequent deep learning decalcification solutions by creating a training data set.

[0042] The method, system, terminal and medium for generating CT image calcification data provided by the present invention can generate a large number of comparative data sets with artifacts and de-artifacted data quickly and at low cost.

[0043] The method, system, terminal and medium for generating CT image calcification data provided by the present invention are based on a large amount of real clinical data, and the results thereof conform to the clinical calcification characteristics.

[0044] The CT image calcification data generation method, system, terminal and medium provided by the present invention can select to use a variety of back-projection algorithms / filters to simulate different CT images generated by more CT manufacturers and related parameters, increasing the generality and robustness of the generated data. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] By reading the following detailed description of the non-limiting embodiments with reference to the accompanying drawings, other features, objects and advantages of the present invention will become more apparent:

[0046] Figure 1 It is a working schematic diagram of the CT image calcification data generation method in an embodiment of the present invention.

[0047] Figure 2 It is a working flowchart of the CT image calcification data generation method in a preferred embodiment of the present invention.

[0048] Figure 3 It is a statistical chart of the number of calcified plaque pixels in a specific application example of the present invention.

[0049] Figure 4 It is a statistical distribution diagram of the generation positions of calcified plaques in a specific application example of the present invention.

[0050] Figure 5 In (a)-(d), it is an example diagram of the filter of the CT image reconstruction back-projection algorithm in a specific application example of the present invention.

[0051] Figure 6 In (a)-(f), it is an example diagram of a partial comparison diagram of generated CT images with and without calcification artifacts in a specific application example of the present invention.

[0052] Figure 7 It is a schematic diagram of the composition structure of the CT image calcification data generation system in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The following details the embodiments of the present invention: This embodiment is implemented on the premise of the technical solution of the present invention, and gives detailed implementation manners and specific operation processes. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can be made, and these all belong to the protection scope of the present invention.

[0054] An embodiment of the present invention provides a method for generating CT image calcification data. The method extracts the examination area and the center line based on the CT image, then selects the calcification to match the target blood vessel, and finally obtains a CT image with / without calcification artifacts through a reconstruction algorithm, solving the problem that existing clinical scans cannot obtain real calcification data after removing artifacts, and providing a basis for subsequent deep learning solutions.

[0055] As shown Figure 1 in the figure, the method for generating CT image calcification data provided by this embodiment may include:

[0056] S100. Segment the calcified area of the CT image A with calcification to obtain a calcified area image with calcification artifacts and a calcified area image without calcification artifacts, and establish a calcification database;

[0057] S200. Extract the blood vessel area and its centerline in the normal CT image B;

[0058] S300. Randomly extract a calcified area image from the calcification database and match it with the target blood vessel area based on the centerline to obtain a calcified plaque;

[0059] S400. Reconstruct the calcified plaque to obtain contrast CT images with and without calcification artifacts respectively, and complete the automatic generation of calcification data.

[0060] The following further details the technical solution provided by the above embodiment of the present invention in conjunction with a preferred embodiment.

[0061] The method for generating CT image calcification data provided by this preferred embodiment includes the following steps:

[0062] S1. Obtain a CT image with calcification and segment the calcified area;

[0063] S2. Extract the blood vessel area and its centerline in the normal CT image;

[0064] S3. Randomly extract a calcified area and match it with the target blood vessel area based on the centerline;

[0065] S4. Use different reconstruction algorithms to reconstruct the calcified area (i.e., calcified plaque) in the blood vessel to obtain contrast CT images without and with artifacts.

[0066] In some preferred embodiments, the above S1 may further include the following operations:

[0067] S101. Obtain the CT image A with calcification;

[0068] S102. Segment the calcified area of the CT image A using different pixel value thresholds to obtain a calcified area image with and without calcification artifacts respectively; further, it includes: S1021. Several CT images of patients containing calcified areas can be obtained as samples; S1022. Use high and low different pixel value thresholds to extract the calcified area from them respectively to obtain calcification data files (calcified area images) without and with calcification artifacts, so as to establish a calcification database;

[0069] S103. Screen the data of the obtained calcified region image, and label the branch and specific segment where the screened calcified region image is located; further, it includes:

[0070] S1031. Divide the calcification labels (0 or 1) for a single patient obtained by threshold segmentation of low pixel values in S102 into different connected components, then count the size of individual calcified regions, and retain the calcifications with appropriate number of pixels to avoid the problem that it is difficult to match subsequent calcified regions and blood vessel labels.

[0071] S1032. Mark the branch where the calcification is located in the examination area, and label the area where it is located according to the segment where the calcification is located, that is, the proximal segment, middle segment or distal segment.

[0072] S104. Correspond one by one the calcified region images with the same calcification obtained by different threshold segmentations, and establish a calcification database.

[0073] In some preferred embodiments, the above S2 may further include the following operations:

[0074] S201. Obtain a normal CT image B.

[0075] S202. Based on the CT image B, extract the heart region excluding the external region of the heart; further, it includes:

[0076] S2021. Import the normal CT image B and render a 3D model effect.

[0077] S2022. Outline the entire heart region and exclude the external region of the heart.

[0078] S203. Based on the heart region, select a segmentation threshold for distinguishing blood vessels from other tissues, and filter out irrelevant regions to obtain a blood vessel region; further, it includes: select a segmentation threshold for distinguishing blood vessels from soft tissues, and filter out irrelevant regions such as soft tissues to obtain a blood vessel region.

[0079] S204. Based on the blood vessel region, use an open-source tool for extracting the centerline to obtain the RAS coordinates of the centerline control points of the blood vessel region and convert them into pixel coordinates to obtain the centerline result; further, it includes: select the starting and ending points of the branch, use the tool for extracting the centerline to extract the centerline, obtain several resampled control points on the line, obtain their numbers, coordinates, directions and other parameters, and calculate the coordinates of the control points on the centerline.

[0080] In some preferred embodiments, the above S3 may further include the following operations:

[0081] S301. Select target points around the centerline of the target blood vessel region; further, it includes:

[0082] S3031, selecting a calcified region extracted with a low threshold value of a branch in the LAD, LCx, and RCA vascular regions, and calculating the coordinate O of its center point;

[0083] S3032, selecting a blood vessel region and randomly selecting a calcification placement position, and selecting a control point P as a target point;

[0084] S302, randomly selecting a calcified region image in the calcification database whose calcified region is located in the same branch and segment of the inspection region as that in S3031, and placing the calcified region at the target point, so that at least more than half of the calcified region is located inside the target blood vessel region, and removing the calcified region located outside the target blood vessel region, completing the matching of the calcified region with the target blood vessel region, and obtaining a calcified plaque; further comprising:

[0085] S3021, randomly select pixel point Q as the new calcification center within the pixel coordinate range with point P as the center and 3 pixels as the radius, and use the difference between the coordinates of P and Q as the translation distance. Then, perform the same operation on the calcified area obtained by high threshold segmentation to obtain the NifTI files of new calcified plaques S_1 and S_2 respectively;

[0086] S3032, determining whether the obtained new calcified plaque is located in the target blood vessel area and screening it.

[0087] In some preferred embodiments, the above S4 may further include the following operations:

[0088] Reconstructing the calcified plaques using different CT reconstruction algorithms to obtain a set of contrast CT images with and without calcification artifacts; further comprising:

[0089] S401, for the combination of the original normal CT image and the calcified plaques S_1 and S_2, a filtered back projection reconstruction algorithm is used to obtain preliminary CT images without calcification artifacts and with calcification artifacts; wherein the high pixel value threshold calcification image is used to generate the calcification artifact-free image; and the low pixel value threshold calcification image is used to generate the calcification artifact-within-one image;

[0090] S402: interpolating the preliminary CT images with and without calcification artifacts obtained in S401 to obtain smooth CT images with and without calcification artifacts

[0091] The method for generating calcification data of CT images provided in the above embodiments of the present invention generates real calcification data after removing artifacts, so as to realize data support for deep learning and low-cost and rapid evaluation of calcification artifact solutions. This method can effectively restore the effect of calcification artifacts, and at the same time obtain the actual size and shape of the corresponding calcifications, thereby helping researchers evaluate the impact of various solutions on calcifications and creating a training dataset for subsequent deep learning de-calcification solutions.

[0092] The following further details the technical solution provided in the above embodiments of the present invention in conjunction with a specific application example. In this specific application example, the coronary artery is used as the examination area, and the calcification data of coronary CT images is generated.

[0093] As Figure 2 shown, the method for generating calcification data of coronary CT images adopted in this specific application example includes the following steps:

[0094] S1, acquisition and processing of coronary CT images with calcifications;

[0095] S1-1, acquisition of coronary CT images with calcifications

[0096] In this specific application example, coronary computed tomography angiography (CCTA) images can be obtained using a 128-row multi-slice CT (Definition AS+, Siemens Healthineers, Germany) instrument from Siemens Corporation, Germany, and stored in the format of DICOM image sequences with a resolution of 512×512.

[0097] S1-2, segmentation of calcified regions

[0098] Coronary CT images of several patients containing calcified regions can be obtained as samples, and the calcified regions can be extracted from them using threshold segmentation to obtain a NIfTI-format calcification annotation file, and a calcification database is established.

[0099] Two different thresholds T 1 , T 2 (T 1 > T 2 ) are used for segmentation to obtain a NIfTI annotation file of the calcified region. The size of the image data contained therein is 512×512×number of slices, where the pixel value of 1 represents the calcified region and 0 represents other regions. According to the principle of the threshold segmentation algorithm, it can be known that the number of pixels (i.e., voxels) contained in the calcification annotation (denoted as N 1 ) segmented from T 1 is less than T 2The number of pixels in the segmented calcification annotation (denoted as N 2 ), and the former is included in the latter. In this embodiment, according to the statistical results of the pixel values of the calcification regions in the CT images of several real cases, and according to the clinical diagnosis criteria, the CT value > 130 HU is set as the threshold T 1 , and the corresponding plaque is N 1 .

[0100] In coronary CT, since the density of the calcification region is higher than that of the surrounding soft tissues, partial volume effect will occur during imaging, resulting in a halo-like artifact around it and making it look larger than the actual size. This is due to the CT imaging principle and it is impossible to completely restore the actual size of the calcification region by improving the precision of the CT instrument, improving the image reconstruction algorithm or using image post-processing means. Therefore, under the guidance of clinical physicians with many years of practice, N 1 can be regarded as calcification without calcification artifacts, while N 2 is regarded as calcification with calcification artifacts.

[0101] S1-3, Calcification data screening

[0102] The calcification annotations for a single patient segmented from T in S1-2 1 are divided into different connected components according to connectivity, that is, separate calcification regions are obtained. The sizes of all separate calcification regions are counted, and the calcification regions with too many and too few pixels are excluded. Because too large calcification regions often have individuality and particularity and cannot represent the shape characteristics of most real calcifications, while too small calcification regions are difficult to distinguish in imaging. In this embodiment, according to the statistical distribution of the actually extracted calcifications, the calcifications with the number of pixels in the range of [15, 2000] are retained, so as to avoid the problem that it is difficult to match the subsequent calcification regions and coronary artery vessel annotations. The distribution of calcification voxels is as Figure 3 shown.

[0103] S1-4, Location indication of calcification data

[0104] The separate calcification annotations extracted in S1-3 are loaded into the original coronary CT image for joint observation. First, determine the coronary artery branch (LAD, LCx, RCA) where the calcification plaque is located. Then, referring to the 18-segment coronary artery segmentation system released by SCCT in 2014, judge whether the calcification is located in the proximal (p) / middle (m) / distal (d) segment of the coronary artery branch according to the segment where the calcification is located.

[0105] The proximal RCA (pRCA) is the length from the opening of the right coronary artery to half of the bend; the middle RCA (mRCA) is the length from the end of the proximal RCA to the bend; and the distal RCA (dRCA) is the length from the end of the middle RCA to the opening of the posterior descending artery (PDA).

[0106] The proximal LAD (pLAD) is from the end of the left main trunk to the first major septal branch or the first diagonal branch (diameter greater than 1.5 mm), whichever is closest; the middle LAD (mLAD) is half the length from the end of the proximal LAD to the apex; the distal LAD (dLAD) is the part from the end of the middle LAD to the end of the anterior descending branch.

[0107] The proximal LCx segment (pCx) is from the end of the left main trunk to the origin of the first obtuse marginal branch (OM1); the mid-distal LCx segment (LCx) is from the opening of the first obtuse marginal branch to the end of the circumflex artery. In order to facilitate unified calculation with LAD and LCx, in this embodiment, the LCx is divided into the mid-LCx segment and the distal LCx segment.

[0108] S1-5, two groups of calcification data obtained by segmentation with different thresholds correspond to each other

[0109] Will use T 1 ,T 2 The calcification annotations obtained by segmentation are imported into the corresponding original CT images. 1 The segmented calcification C 1 With T 2 The obtained calcification C 2 The position and shape of the NIfTI annotation file are consistent, and the result of reading the NIfTI annotation file data is consistent with The two are marked as the same calcification for subsequent operations.

[0110] S2, obtain the vascular region and extract the centerline based on the normal coronary CT segmentation;

[0111] S2-1, blood vessel region segmentation

[0112] With the help of the open source software 3D Slicer, calcification-free coronary CT images were imported and rendered into a 3D model. First, the entire heart area was outlined and the area outside the heart was eliminated. Then, based on the CT image, a threshold that could well distinguish between blood vessels and soft tissue was selected. The pixel value range was determined to preliminarily select the area where the blood vessels were located, while the area where the soft tissue was located was discarded.

[0113] Use the 3D brush to outline the three branches along the direction of the coronary artery as three sets of non-intersecting marks. After completing the preliminary outline, use the "Crop" function to further refine the vascular annotation for irregular and uneven vascular wall parts.

[0114] S2-2, Centerline Extraction

[0115] Manually select the starting and ending points of the coronary artery branches, and use 3D Slicer to extract the centerline to obtain a continuous and smooth curve and several control points evenly distributed on the line. Resample the control points to make their number 40, and obtain their relevant parameters, including control point numbers (starting from 0 to 39), coordinates, directions and other parameters. The control point coordinates use the RAS coordinate system (Right, Anterior, Superior).

[0116] S2-3, Calculate the Coordinates of Control Points on the Centerline

[0117] Convert the RAS coordinates (original coordinates, unit: mm) to pixel coordinates (coordinates of NIfTI files as matrix data, unit: pixel), that is, IJK coordinates.

[0118] The RAS coordinates refer to the anatomical space, that is, the patient coordinate system, which includes an axial plane parallel to the ground and dividing the head (superior) and feet (inferior), a coronal plane perpendicular to the ground and dividing the front (anterior) and back (posterior), and a sagittal plane dividing the left and right. The meaning of the RAS coordinates used by the 3D Slicer software is as follows:

[0119]

[0120] The pixel coordinates refer to the image coordinate system. The medical scanning instrument establishes a regular pixel point array from the upper left corner. The i-axis increases to the right, the j-axis increases downward, and the k-axis increases backward. The image information includes the pixel value of each voxel point (i, j, k), the position (origin) corresponding to the starting point (0, 0, 0), the physical distance (spacing) between voxels, the transformation matrix (matrix), etc., and satisfies the following relationship:

[0121] IJK coordinates=matrix -1 *(RAScoordinates-origin) / spacing

[0122] For the control points of the blood vessel centerline, the nibabel library in Python can be used to read the NIfTI file to obtain the affine attribute, that is, the 3×3 transformation matrix A, and the pixel coordinates can be calculated using the following formula:

[0123] (i,j,k)=inv(A)*[RAS1]′

[0124] S3, randomly select calcification and blood vessels for matching, and obtain the coronary branches containing calcification;

[0125] S3-1, according to the statistical distribution, select a coronary artery branch (LAD left anterior descending branch, LCx circumflex branch, RCA right main trunk) with T 1 For the calcification extracted by the threshold, calculate the coordinates O of its center point.

[0126] The calcified plaque is regarded as a combination of several pixel points, and the average value of the IJK coordinates of all pixel points is taken as the coordinate of the center point of the calcification.

[0127] S3-2, select the vessel and randomly choose the placement of the calcification.

[0128] Select the normal blood vessel annotation file that belongs to the same coronary branch as the calcification in S3-1, and randomly select a control point P on the same segment as the calcification in S3-1 (the control point number is [0,15] for the near segment, [16,27] for the middle segment, and [28,39] for the far segment) to obtain the slice where point P is located. Use pixel coordinates to determine the coordinates of the points inside the blood vessel on the slice with point P as the center and 3 pixels as the radius as an alternative. The segmented statistical distribution is as follows Figure 4 shown.

[0129] S3-3, randomly select a point Q within the pixel coordinate range obtained in S3-2 as the center point of the new calcification, and place the calcification selected in S3-1. Calculate the coordinate difference between O and Q as the translation distance, and obtain the corresponding new calcification plaque S 1 For NIfTI files with T 2 The same operation is performed on the calcification at the same position extracted by the threshold to obtain the corresponding new calcified plaque S 2 NIfTI file.

[0130] S3-4, determine whether it is located in a blood vessel and screen it. 1 , Statistics S 1 Whether the included pixels are located inside the coronary branch. If the number of pixels inside the blood vessel is N in Greater than or equal to S 1 1 / 3 of the total number of pixels and meets the size requirement of calcified plaques, that is, N mentioned in S1-3 in ∈[15,2000], the calcified plaque is retained, and all pixels outside the blood vessels are removed and only the pixels inside the blood vessels are retained to form the final calcified plaque S′ 1 , similarly retain the corresponding S 2 All pixels inside the blood vessel form another calcified plaque S′ 2 ; Otherwise, return to S3-3 to regenerate the calcified plaque.

[0131] S4. Reconstruct the results of S3 using two different CT reconstruction algorithms to obtain a set of comparative CT images with / without calcification artifacts.

[0132] S4-1. For the original normal CT image and the combination of S 1 and S 2 respectively use the filtered backprojection (FBP) reconstruction algorithm to obtain CT images with and without calcification artifacts.

[0133] The imaging process of CT images can be summarized as the Radon Transform and the inverse Radon Transform. The former projects the object into the sinogram domain through X-ray scanning, and the latter reconstructs the projection data to restore the object image.

[0134] Perform the Radon transform on the combination of the original normal CT image and S 1 and map it to the sinogram domain. The formula is as follows:

[0135]

[0136] where δ(x) is the Dirac function, θ is the detector rotation angle, and t = rsinθ is the position of the trajectory left by the detector point source.

[0137] The inverse Radon transform for image reconstruction is mathematically the two-dimensional inverse Fourier transform. Since the two-dimensional inverse Fourier transform requires a large amount of calculation, the backprojection method is used to implement the inverse Radon transform. The backprojection algorithm backprojects the projection signal for each angle, averages the projection signal to each two-dimensional spatial point, and then superimposes the backprojection images of all angles to infer the original image. Due to the discrete and finite projection angles, the image superimposed by the backprojection method will have "star artifacts", resulting in blurred image edges.

[0138] The FBP algorithm is an improvement of the backprojection method. First, filter the projection for each projection angle and then perform backprojection. Commonly used filters include the Ram-Lak filter, Shepp-Logan filter, cosine filter, and Hann filter, etc., as shown in Figure 5 (a) - (d) below.

[0139] The specific steps of the FBP algorithm are as follows:

[0140] 1. Pad the projection function p(t, θ) with 0 to get p′(t, θ), and then perform one-dimensional Fourier transform on t to obtain P(ω, θ);

[0141] 2. Select a suitable filter, at θi The original projection p(t, θ) will be obtained at an angle of i ) is convolved and filtered to obtain the filtered projection Q(ω, θ i );

[0142] 3. The filtered projection Q(ω, θ i ) is subjected to one-dimensional inverse Fourier transform to be restored to the time domain, and the density q(t, θ) of the original image in the direction of t = rcos(θ - θ i ) is obtained;

[0143] 4. Back-projection is performed using q(t, θ), and the reconstructed image is obtained after superposition.

[0144] S4-2. Interpolation is performed on the CT image of S4-1 to make the reconstructed image more continuous and smoother. Common interpolation methods include nearest neighbor interpolation and bilinear interpolation, etc.

[0145] The partial comparison diagrams of the obtained parts with and without calcification artifacts are as Figure 6 shown in (a)-(f) therein.

[0146] An embodiment of the present invention provides a system for generating CT image calcification data, as Figure 7 shown, the system provided by this embodiment may include the following modules:

[0147] Calcification segmentation module, which is used to segment the calcified area of the CT image A with calcification to obtain the calcified area image with calcification artifacts and the calcified area image without calcification artifacts, and establish a calcification database;

[0148] Vessel extraction module, which is used to extract the vessel area and its centerline in the normal CT image B;

[0149] Simulation calcification generation module, which is used to randomly extract the calcified area image from the calcification database and match it based on the centerline with the target vessel area to obtain calcified plaques;

[0150] CT image reconstruction module, which is used to reconstruct the calcified plaques to obtain the comparison CT images with and without calcification artifacts respectively, and complete the automatic generation of calcification data.

[0151] In some preferred embodiments, the above-mentioned calcification segmentation module first collects CT images and stores them in the format of DICOM image sequences. Then, it segments the calcification regions. It can obtain the CT images of several patients containing calcification regions as samples, and extract the calcification regions from them using two different high and low thresholds respectively to obtain calcification data files, thereby establishing a calcification database. It screens the calcification data, divides the calcification annotations for a single patient obtained by low-threshold segmentation into different connected components, then counts the sizes of individual calcification regions, and retains the calcifications with appropriate numbers of pixels to avoid problems in subsequent matching between calcification regions and vessel annotations. It marks the positions of the calcification data, and annotates the regions where the calcifications are located according to the segments, namely the proximal segment, the distal segment, and the middle segment. Finally, the two sets of calcification data obtained by different-threshold segmentation are made to correspond one by one.

[0152] In some preferred embodiments, the above-mentioned vessel extraction module first performs vessel region segmentation. It imports CT images without calcification and renders a 3D model effect, then outlines the entire heart region, and then selects a threshold for distinguishing vessels and soft tissues to select the vessel region. Then, it can perform centerline extraction, select the start and end points of the branches, and use tools to extract the centerline to obtain several resampled control points on the line, and obtain parameters such as their numbers, coordinates, and directions. Finally, it calculates the coordinates of the control points on the centerline.

[0153] In some preferred embodiments, the above-mentioned simulated calcification generation module selects the calcifications extracted with a low threshold from a certain branch among LAD, LCX, and RCA, and calculates the coordinates of their center points. It selects a vessel and randomly selects a placement position for the calcification, and selects a control point P. It randomly selects a pixel point Q within the obtained pixel coordinate range as the center point of the calcification, and uses the difference between the coordinates of P and Q as the translation distance. Then, it performs the same operation on the calcifications segmented with a high threshold to obtain NifTI files of new calcification plaques S1 and S2 respectively. Finally, it determines whether the new calcifications are located in the vessels and performs screening.

[0154] In some preferred embodiments, the above-mentioned CT image reconstruction module, for the combination of the original normal CT image and S1, uses the filtered back-projection reconstruction algorithm to obtain a CT image without calcification artifacts; for the combination of the original CT image and S2, uses the back-projection reconstruction algorithm to obtain a CT image with calcification artifacts.

[0155] It should be noted that the steps in the method provided by the present invention can be implemented by corresponding modules, devices, units, etc. in the system. Those skilled in the art can refer to the technical solutions of the method to implement the composition of the system. That is, the embodiments in the method can be understood as preferred examples for constructing the system, which will not be elaborated here.

[0156] An embodiment of the present invention provides a computer terminal, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it can be used to execute the method of any one of the above embodiments of the present invention, or, run the system of any one of the above embodiments of the present invention.

[0157] Optionally, the memory is used to store programs; the memory may include volatile memory (English: volatile memory), such as random access memory (English: random-access memory, abbreviation: RAM), such as static random access memory (English: static random-access memory, abbreviation: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviation: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory is used to store computer programs (such as application programs and functional modules for implementing the above methods), computer instructions, etc. The above computer programs, computer instructions, etc. can be stored in partitions in one or more memories. And the above computer programs, computer instructions, data, etc. can be called by the processor.

[0158] The above computer programs, computer instructions, etc. can be stored in partitions in one or more memories. And the above computer programs, computer instructions, data, etc. can be called by the processor.

[0159] The processor is used to execute the computer program stored in the memory to implement each step in the method or each module in the system involved in the above embodiments. For specific details, please refer to the relevant descriptions in the previous method and system embodiments.

[0160] The processor and the memory can be independent structures or integrated structures. When the processor and the memory are independent structures, the memory and the processor can be coupled and connected through a bus.

[0161] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor, it can be used to execute the method of any one of the above embodiments of the present invention, or, run the system of any one of the above embodiments of the present invention.

[0162] The method and system for generating CT image calcification data provided by the above embodiments of the present invention realize automatic data generation based on real clinical data. Based on the real CT acquisition and reconstruction process, it can effectively simulate the artifacts generated by the calcified area during real CT scanning and obtain the relevant data corresponding to the calcification itself at the same time. The method and system for generating CT image calcification data provided by the above embodiments of the present invention can effectively solve the problem that the existing clinical scans cannot obtain the real calcification data after removing artifacts, and provide a basis for subsequent deep learning solutions.

[0163] Those skilled in the art know that in addition to implementing the system and its various devices provided by the present invention in the form of pure computer-readable program code, the method steps can be logically programmed to enable the system and its various devices provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same functions. Therefore, the system and its various devices provided by the present invention can be regarded as a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component; the devices for implementing various functions can also be regarded as either software modules for implementing the method or structures within the hardware component.

[0164] All matters not described in detail in the above embodiments of the present invention are well-known technologies in the art.

[0165] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various deformations or modifications within the scope of the claims, which does not affect the essence of the present invention.

Claims

1. A method for generating CT image calcification data, characterized in that, it includes: Segment the calcified regions of the CT image A with calcification to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts, and establish a calcification database; Extract the blood vessel regions and their centerlines from the normal CT image B; Randomly select the calcified region images in the calcification database, and match them with the target blood vessel region based on the centerline to obtain calcified plaques; Reconstruct the calcified plaques to obtain contrast CT images with and without calcification artifacts respectively, and complete the automatic generation of calcification data.

2. The method for generating CT image calcification data according to claim 1, characterized in that, The segmenting the calcified regions of the CT image A with calcification to obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts, and establishing a calcification database includes: Obtain the CT image A with calcification; Use different pixel value thresholds to segment the CT image A for calcified regions, and obtain a calcified region image with calcification artifacts and a calcified region image without calcification artifacts respectively; Perform data screening on the obtained calcified region images, and label the branches and specific segments where the screened calcified region images are located; Correspond one by one the calcified region images with the same calcification segmented by different thresholds, and establish a calcification database.

3. The method for generating CT image calcification data according to claim 2, characterized in that, The performing data screening on the obtained calcified region images includes: Divide the pixel points labeled as calcification in the calcified region images of a single patient segmented by different pixel value thresholds into different connected components to obtain multiple calcified sub-region images; Count the pixel sizes of each calcified sub-region image, and retain the calcified sub-region images whose number of pixels meets the set requirements to form the required calcified region images, and complete the data screening of the calcified region images.

4. The method for generating CT image calcification data according to claim 1, characterized in that, The extracting the blood vessel regions and their centerlines from the normal CT image B includes: Obtain the normal CT image B; Based on the CT image B, extract the heart region excluding the external region of the heart; Based on the heart region, select a segmentation threshold for distinguishing blood vessels from other tissues, and filter out irrelevant regions to obtain the blood vessel region; Based on the blood vessel region, use the centerline extraction tool to obtain the RAS coordinates of the centerline control points of the blood vessel region and convert them into pixel coordinates to obtain the centerline result.

5. The method for generating CT image calcification data according to claim 1, characterized in that, The randomly selecting the calcified region images in the calcification database, and matching them with the target blood vessel region based on the centerline to obtain calcified plaques includes: Select target points around the centerline of the target blood vessel region; A calcified region image in the calcification database that is located in the same branch and segment as the centerline control point is randomly selected, and the calcified region image is placed at the target point, so that at least more than half of the calcified region is located inside the target blood vessel region, and the calcified region outside the target blood vessel region is removed, and the calcified region is matched with the target blood vessel region to obtain a calcified plaque.

6. The method for generating CT image calcification data according to claim 1, It is characterized in that The reconstructing the calcified plaque to obtain contrast CT images with and without calcification artifacts includes: Different CT reconstruction algorithms are used to reconstruct the calcified plaques to obtain a set of contrast CT images with and without calcification artifacts.

7. The method for generating CT image calcification data according to claim 6, It is characterized in that The reconstructing the calcified plaque by using different CT reconstruction algorithms respectively includes: Reconstructing the calcified plaque using a filtered back projection algorithm to obtain preliminary CT images with and without calcification artifacts; The CT images without calcification artifacts and with calcification artifacts are interpolated to obtain CT images without calcification artifacts and with calcification artifacts.

8. A system for generating calcification data of CT images, It is characterized in that include: A calcification segmentation module, which is used to segment the calcification region of the CT image A with calcification, obtain calcification region images with calcification artifacts and calcification region images without calcification artifacts, and establish a calcification database; A blood vessel extraction module, which is used to extract the blood vessel region and its center line in a normal CT image B; A simulated calcification generation module, which is used to randomly extract a calcification region image from the calcification database, and match it with a target blood vessel region based on the center line to obtain a calcified plaque; A CT image reconstruction module is used to reconstruct the calcified plaque, obtain contrast CT images with and without calcification artifacts, and complete the automatic generation of calcification data.

9. A computer terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the computer program, it can be used to perform the method described in any one of claims 1 to 7, or to run the system described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, it can be used to perform the method described in any one of claims 1 to 7, or to run the system described in claim 8.

Images