A method for predicting coronary artery calcification score based on chest CT image

Abstract

The present application relates to the technical field of image analysis, in particular to a coronary artery calcification score prediction method based on chest CT images, which comprises: acquiring chest CT images corresponding to different preset layers collected at the current time by a patient to be detected, and performing connected domain extraction on each chest CT image; determining the target similarity between different connected domains in the chest CT images corresponding to each adjacent preset layer; screening out connected domains representing the same tissue structure; performing blood vessel change rule analysis and processing on each connected domain sequence; screening out blood vessel connected domains; based on the gray value, screening out a target calcification region from the region formed by all blood vessel connected domains in each chest CT image, and performing blood vessel calcification score prediction based on the target calcification region. The present application realizes blood vessel calcification region identification by analyzing chest CT images, improves the efficiency of blood vessel calcification region identification, and thus improves the efficiency of blood vessel calcification score prediction.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, specifically to a method for predicting coronary artery calcification scores based on chest CT images. Background Technology

[0002] With the development of technology, image analysis is being applied more and more widely. For example, it can be used to predict vascular calcification scores, thereby assisting doctors in assessing vascular calcification. Since vascular calcification often occurs in coronary arteries, the predicted vascular calcification score can, to some extent, characterize the coronary artery calcification score. Currently, calcification score prediction is often based on calcified regions using the Agatston integral algorithm. Therefore, accurate identification of calcified regions is crucial. Currently, the common method for calcification region identification is to use neural networks to identify calcified regions from images.

[0003] However, when using neural networks to identify vascular calcification regions in chest CT images, the following technical problems often arise:

[0004] Because blood vessels in different people often have subtle variations, and blood vessels with different degrees of calcification often exhibit different characteristics, neural networks for identifying calcified regions of blood vessels often require a large number of different calcified blood vessel images and normal blood vessel images during training. Acquiring these images often takes a lot of time, resulting in poor efficiency in identifying calcified regions of blood vessels, and consequently, poor efficiency in predicting calcification integrals of blood vessels. Summary of the Invention

[0005] To address the technical problem of poor efficiency in predicting vascular calcification scores due to the poor efficiency of vascular calcification region identification, this invention proposes a method for predicting coronary artery calcification scores based on chest CT images.

[0006] In a first aspect, the present invention provides a method for predicting coronary artery calcium scores based on chest CT images, the method comprising:

[0007] Acquire chest CT images of the patient to be tested at different preset layers at the current time, and extract connected components for each chest CT image;

[0008] Based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer, the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer is determined.

[0009] Based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers, connected regions representing the same tissue structure are selected, and connected regions representing the same tissue structure are formed into a connected region sequence.

[0010] The vascular change patterns of each connected component sequence were analyzed to obtain the vascular attribution index corresponding to each connected component sequence.

[0011] Based on the vessel attribution index corresponding to different connected component sequences, the vessel connected components are screened out from the connected components in all connected component sequences;

[0012] Based on grayscale values, target calcified regions are selected from the regions formed by all vascular connected domains in each chest CT image, and vascular calcification scores are predicted based on the target calcified regions in different chest CT images.

[0013] In conjunction with the first aspect above, in one possible implementation, determining the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer includes:

[0014] The chest CT images corresponding to any two adjacent preset layers are respectively designated as the first marker image and the second marker image;

[0015] Any connected component in the first labeled image is determined as the first labeled connected component, and any connected component in the second labeled image is determined as the second labeled connected component;

[0016] The region in the second marked image that is at the same position as the first marked connected region is determined as the reference region;

[0017] Based on the area of ​​the reference region and the area of ​​the second marked connected region, determine the area overlap ratio between the first marked connected region and the second marked connected region;

[0018] Based on the positional distribution between the centroids of the reference region and the second marked connected region, determine the length overlap ratio between the first marked connected region and the second marked connected region;

[0019] The target similarity between the first and second marked connected components is determined based on the area overlap ratio and length overlap ratio between the first and second marked connected components, the area difference between the first and second marked connected components, the distance between the centroid of the reference region and the centroid of the second marked connected component, and the SSIM value between the first and second marked connected components.

[0020] In conjunction with the first aspect above, in one possible implementation, determining the area overlap ratio between the first marked connected region and the second marked connected region based on the areas of the reference region and the second marked connected region includes:

[0021] The area of ​​the intersection of the reference region and the second marked connected region is determined as the representative area of ​​the overlap between the first marked connected region and the second marked connected region;

[0022] The average area of ​​the first marked connected region and the area of ​​the second marked connected region are determined as the relative representative area between the first marked connected region and the second marked connected region.

[0023] The ratio of the overlapping area between the first marked connected region and the second marked connected region to the relative area is determined as the area overlap ratio between the first marked connected region and the second marked connected region.

[0024] In conjunction with the first aspect above, in one possible implementation, determining the length overlap ratio between the first and second marked connected regions based on the positional distribution between the centroids of the reference region and the second marked connected region includes:

[0025] The intersection of the reference region and the second marked connected component is determined as the target intersection region, and the union of the reference region and the second marked connected component is determined as the target union region.

[0026] If the centroid of the reference region does not coincide with the centroid of the second marked connected region, then a straight line is drawn through the centroid of the reference region and the centroid of the second marked connected region, and denoted as the marked line. The length of the intersection of the marked line and the target intersection region is determined as the target intersection length. The length of the intersection between the marked line and the target union region is determined as the target union length. The ratio of the target intersection length to the target union length is determined as the length overlap ratio index between the first marked connected region and the second marked connected region.

[0027] If the centroid of the reference region coincides with the centroid of the second marked connected region, then through the centroid of the second marked connected region, and with each preset direction as the straight line direction, a reference straight line is drawn in each preset direction. Based on the reference straight line in each preset direction, the target intersection region, and the target union region, the initial length coincidence factor in each preset direction is determined. The minimum value among all the initial length coincidence factors in all preset directions is determined as the length coincidence ratio index between the first marked connected region and the second marked connected region.

[0028] In conjunction with the first aspect above, in one possible implementation, determining the initial length coincidence factor for each preset direction based on the reference line in each preset direction, the target intersection region, and the target union region includes:

[0029] The length of the intersection of the reference line in each preset direction and the target intersection region is determined as the reference intersection length in each preset direction.

[0030] The length of the intersection between the reference line in each preset direction and the target region is determined as the reference length in each preset direction.

[0031] The ratio of the reference intersection length to the reference union length in each preset direction is determined as the initial length coincidence factor in each preset direction.

[0032] In conjunction with the first aspect above, in one possible implementation, the step of filtering connected regions representing the same tissue structure based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers includes:

[0033] Any preset layer is designated as a marker layer, and the preset layer preceding or following the marker layer is designated as a reference layer.

[0034] Any connected component in the chest CT image corresponding to the marked layer is determined as a candidate connected component, and the chest CT image corresponding to the reference layer is determined as a reference image;

[0035] Connected regions with a target similarity greater than a preset similarity threshold with the candidate connected regions are selected from the reference image and used as reference connected regions;

[0036] From all reference connected components, the reference connected component with the highest target similarity to the candidate connected component is selected to obtain the connected component that has the same organizational structure as the candidate connected component.

[0037] In conjunction with the first aspect above, in one possible implementation, the step of performing vascular change pattern analysis on each connected component sequence to obtain the vascular attribution index corresponding to each connected component sequence includes:

[0038] Determine the target width corresponding to each connected component in each connected component sequence to obtain the target width sequence for each connected component sequence;

[0039] Based on the mean and range of the target width sequence corresponding to each connected component sequence, and the target similarity between adjacent connected components in each connected component sequence, the vessel attribution index corresponding to each connected component sequence is determined.

[0040] In conjunction with the first aspect above, in one possible implementation, determining the target width corresponding to each connected component in each connected component sequence includes:

[0041] Determine any one of the connected components in the sequence of all connected components as the calibration connected component, and draw calibration lines in different directions through the centroid of the calibration connected component;

[0042] The length of the intersection of the calibration connected region and each calibration line is determined as the reference length corresponding to each calibration line;

[0043] The minimum value of the reference length corresponding to all calibration lines is determined as the target width corresponding to the calibration connected region.

[0044] In conjunction with the first aspect above, in one possible implementation, the step of filtering out vascular connected regions from all connected region sequences based on the vessel attribution index corresponding to different connected region sequences includes:

[0045] If the vessel attribution index corresponding to the connected component sequence is greater than the preset vessel attribution threshold, then the connected component in the connected component sequence is determined as a vessel connected component.

[0046] In conjunction with the first aspect above, in one possible implementation, the step of filtering out the target calcified region from the region formed by all vascular connected domains in each chest CT image based on grayscale values ​​includes:

[0047] Any chest CT image is designated as a temporary image. Threshold segmentation is performed on the region formed by all vascular connected domains in the temporary image. The sub-region with the larger gray value among the two sub-regions obtained by threshold segmentation is designated as the target calcified region.

[0048] Secondly, the present invention provides a coronary artery calciumization score prediction system based on chest CT images, the system comprising:

[0049] The acquisition and extraction module is used to acquire chest CT images of the patient to be tested at different preset layers at the current time, and to extract the connected components of each chest CT image.

[0050] The target similarity determination module is used to determine the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer.

[0051] The filtering and construction module is used to filter out connected regions that represent the same tissue structure based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers, and to form a connected region sequence from the connected regions that represent the same tissue structure.

[0052] The vascular change pattern analysis and processing module is used to perform vascular change pattern analysis and processing on each connected component sequence to obtain the vascular attribution index corresponding to each connected component sequence.

[0053] The vascular connected component filtering module is used to filter out vascular connected components from all connected component sequences based on the vascular attribution index corresponding to different connected component sequences.

[0054] The screening and vascular calcification score prediction module is used to screen out target calcified regions from the region composed of all vascular connected domains in each chest CT image based on gray values, and to predict vascular calcification scores based on target calcified regions in different chest CT images.

[0055] Thirdly, a server is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, causing the device to perform the methods of the first aspect or any possible implementation thereof.

[0056] Fourthly, a computer program product is provided, comprising: computer program code, which, when run on a computer, causes the computer to perform the methods described in the first aspect or any possible implementation thereof.

[0057] Fifthly, a computer-readable storage medium is provided that stores computer program code, which, when executed on a computer, causes the computer to perform the methods described in the first aspect or any possible implementation thereof.

[0058] The present invention has the following beneficial effects:

[0059] This invention provides a method for predicting coronary artery calcium scores based on chest CT images. By analyzing chest CT images, it achieves vascular calcification region identification, solving the technical problem of poor vascular calcification region identification leading to poor vascular calcification score prediction efficiency, and improving the efficiency of vascular calcification region identification, thereby improving the efficiency of vascular calcification score prediction. Specifically, compared to vascular calcification region identification through neural networks, this invention achieves vascular calcification region identification by analyzing chest CT images, thus avoiding the need for training set acquisition, reducing time waste to a certain extent, and improving the efficiency of vascular calcification region identification and vascular calcification score prediction. Secondly, when identifying vascular calcification regions, this invention analyzes the vascular change patterns of connected domain sequences representing the same tissue structure, quantifies the vascular attribution index corresponding to each connected domain sequence, thereby screening out vascular connected domains, and based on grayscale values, filters out target calcification regions from the region formed by all vascular connected domains in the chest CT image, improving the accuracy of vascular calcification region identification, and thus improving the accuracy of vascular calcification score prediction. Attached Figure Description

[0060] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a flowchart of a method for predicting coronary artery calcification scores based on chest CT images according to the present invention.

[0062] Figure 2 This is a schematic diagram of the composition structure of a coronary artery calcification score prediction system based on chest CT images according to the present invention.

[0063] Figure 3 This is a schematic diagram of the structure of a computer device according to the present invention. Detailed Implementation

[0064] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the specific implementation methods, structures, features, and effects of the technical solution proposed according to the present invention are described in detail below with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0066] Vascular calcification occurs when calcium and phosphorus products accumulate on the walls of blood vessels. The continuous buildup of calcified plaques can harden blood vessels, affecting the oxygen and blood supply to related organs and easily leading to many diseases. Coronary artery calcification (CAC) is a major cause of coronary heart disease, and early detection of CAC is crucial for the prevention and treatment of coronary heart disease and other cardiovascular diseases.

[0067] Currently, coronary artery calcification is mainly detected through computed tomography (CT) imaging. Doctors use the obtained CT images to check for areas of coronary artery calcification and quantify the risk of cardiovascular disease based on the Coronary Artery Calcification Score (CACS).

[0068] Currently, one method for region identification is based on differences in grayscale values. However, when identifying vascular calcification regions in chest CT images solely based on grayscale values, the presence of other tissues with high calcium content, such as bone, within the chest region can lead to inaccuracies in the identification of vascular calcification pixels. This is because the grayscale values ​​of these tissues are often similar to those of the vascular calcification regions. Therefore, relying solely on grayscale values ​​for vascular calcification region identification in chest CT images may result in poor accuracy in identifying vascular calcification pixels, thus leading to inaccurate vascular calcification region identification.

[0069] In identifying vascular calcification regions, this invention analyzes the vascular change patterns of connected domain sequences representing the same tissue structure, quantifies the vascular attribution index corresponding to each connected domain sequence, thereby screening out vascular connected domains. Based on grayscale values, it selects target calcification regions from the regions composed of all vascular connected domains in chest CT images, improving the accuracy of vascular calcification region identification and thus improving the accuracy of vascular calcification score prediction.

[0070] refer to Figure 1 The flowchart illustrates some embodiments of a method for predicting coronary artery calcium scores based on chest CT images according to the present invention. This method for predicting coronary artery calcium scores based on chest CT images includes the following steps:

[0071] Step S1: Obtain chest CT images of the patient to be tested at different preset layers at the current time, and extract connected components for each chest CT image.

[0072] The patient to be tested can be one undergoing calcification score prediction. The preset layer can be a pre-defined scanning layer with a layer thickness of 5 mm. Chest CT images corresponding to different preset layers can represent cross-sectional imaging at different heights in the vertical direction. The chest CT image can be a CT image of the patient's chest. CT images are tomographic imaging, which uses X-ray beams to scan layers of the human body at specific thicknesses.

[0073] As an example, this step may include the following steps:

[0074] The first step is to use a CT scanner to acquire chest CT images of the patient at different preset slices at the current time. The chest CT images can be CT images after median filtering and noise reduction.

[0075] The second step is to extract the connected components of each chest CT image using a connected component extraction algorithm.

[0076] Step S2: Determine the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer.

[0077] It should be noted that the obtained CT images are tomographic images, which use X-ray beams to scan layers of the human body at specific thicknesses. The thickness between layers is relatively low, and the same human tissue structure in adjacent layers often shows a certain degree of similarity in CT images. Therefore, determining the target similarity between different connected regions in chest CT images corresponding to adjacent preset layers can facilitate subsequent determination of whether different connected regions in chest CT images corresponding to adjacent preset layers belong to the same tissue.

[0078] As an example, this step may include the following steps:

[0079] The first step is to designate any two adjacent chest CT images corresponding to preset layers as the first marker image and the second marker image, respectively.

[0080] The second step is to determine any one of the connected components in the first labeled image as the first labeled connected component, and to determine any one of the connected components in the second labeled image as the second labeled connected component.

[0081] The third step is to identify the regions in the second marked image that are in the same position as the first marked connected region as reference regions.

[0082] The reference region can be a region in the second marked image that has the same location and shape as the connected region of the first marked area.

[0083] The fourth step, determining the area overlap ratio between the first and second marked connected regions based on the areas of the aforementioned reference region and the second marked connected region, may include the following sub-steps:

[0084] The first sub-step involves determining the area of ​​the intersection of the aforementioned reference region and the aforementioned second marked connected region as the representative area of ​​the overlap between the aforementioned first marked connected region and the aforementioned second marked connected region.

[0085] The second sub-step involves determining the average area of ​​the first marked connected region and the area of ​​the second marked connected region as the relative representative area between the first marked connected region and the second marked connected region.

[0086] The third sub-step is to determine the ratio of the overlapping representative area to the relative representative area between the first marked connected region and the second marked connected region as the area overlap ratio index between the first marked connected region and the second marked connected region.

[0087] For example, the formula for determining the area overlap ratio between the first-labeled connected component and the second-labeled connected component can be:

[0088] ;

[0089] ;

[0090] Where B is the percentage of overlap between the first and second labeled connected regions. A is the representative overlapping area between the first and second labeled connected regions, which is the area of ​​the intersection of the reference region and the second labeled connected region. S is the relative representative area between the first and second labeled connected regions. It is the numerical value corresponding to the area of ​​the first labeled connected region. It is the numerical value corresponding to the area of ​​the second-marked connected region.

[0091] It should be noted that in practice, if two connected regions belong to the same tissue and are obtained from adjacent CT scans, their corresponding areas in the CT images often overlap to some extent. A larger value for B generally indicates a greater degree of overlap in the areas of the first and second labeled connected regions within the same coordinate system, and also suggests a relatively smaller morphological variation between the two regions.

[0092] The fifth step, based on the positional distribution between the centroids of the aforementioned reference region and the aforementioned second-marked connected region, determines the length overlap ratio between the aforementioned first-marked connected region and the aforementioned second-marked connected region, which may include the following sub-steps:

[0093] The first sub-step is to determine the intersection of the above-mentioned reference region and the above-mentioned second-marked connected region as the target intersection region, and to determine the union of the above-mentioned reference region and the above-mentioned second-marked connected region as the target union region.

[0094] The second sub-step, if the centroid of the aforementioned reference region does not coincide with the centroid of the aforementioned second-marked connected region, then determining the length overlap ratio between the aforementioned first-marked connected region and the aforementioned second-marked connected region may include the following steps:

[0095] First, draw a straight line through the centroid of the aforementioned reference region and the centroid of the aforementioned second-marked connected region, and denote it as the marked straight line.

[0096] The marked line can be a line passing through the centroid of the reference region and the centroid of the second marked connected region.

[0097] Next, the length of the intersection of the marked line and the target intersection area is determined as the target intersection length.

[0098] It should be noted that the intersection of a straight line and a region is often a line segment.

[0099] Then, the length of the intersection between the marked line and the target region is determined as the target length.

[0100] Finally, the ratio of the intersection length of the above targets to the union length of the above targets is determined as the length overlap ratio index between the above first marked connected region and the above second marked connected region.

[0101] For example, if the centroid of the reference region does not coincide with the centroid of the second labeled connected region, the formula for determining the percentage of length overlap between the first and second labeled connected regions can be:

[0102] ;

[0103] Where d is the percentage of overlap in length between the first and second labeled connected components. v is the target intersection length, which is the length of the intersection between the labeled line and the target region. u is the target union length, which is the length of the intersection between the labeled line and the target union region.

[0104] It should be noted that d can characterize the positional change between the first labeled connected component and the second labeled connected component. The larger the value, the smaller the positional change between the first labeled connected component and the second labeled connected component.

[0105] The third sub-step, if the centroid of the aforementioned reference region coincides with the centroid of the aforementioned second-marked connected region, then determining the length overlap ratio between the aforementioned first-marked connected region and the aforementioned second-marked connected region may include the following steps:

[0106] First, through the centroid of the second marked connected region, and taking each preset direction as the straight line direction, draw a reference straight line under each preset direction.

[0107] The preset direction can be a pre-defined straight line direction. The number of preset directions can be a pre-defined number, which can be equal to 180. The angle corresponding to the preset direction can range from {0°, 1°, 2°, ..., 179°}. The reference line under the preset direction can be a straight line passing through the centroid of the second marked connected region and whose direction is the preset direction.

[0108] Next, based on the reference line in each preset direction, the aforementioned target intersection region, and the aforementioned target union region, determining the initial length coincidence factor in each preset direction may include: determining the length of the intersection between the reference line in each preset direction and the aforementioned target intersection region as the reference intersection length in each preset direction; determining the length of the intersection between the reference line in each preset direction and the aforementioned target union region as the reference union length in each preset direction; and determining the ratio of the reference intersection length to the reference union length in each preset direction as the initial length coincidence factor in each preset direction.

[0109] Finally, the minimum value among the initial length overlap factors under all preset directions is determined as the length overlap ratio index between the first marked connected region and the second marked connected region.

[0110] For example, if the centroid of the reference region coincides with the centroid of the second labeled connected region, then the formula for determining the length overlap ratio between the first and second labeled connected regions can be:

[0111] ;

[0112] ;

[0113] Where d is the percentage of length overlap between the first-labeled connected component and the second-labeled connected component. It is a function that takes the minimum value. It is the initial length coincidence factor under the first preset direction. It is the initial length coincidence factor under the second preset direction. It is the initial length coincidence factor under the i-th preset direction. i is the index of the preset direction. It is the initial length coincidence factor under the nth preset direction. n is the number of preset directions. It is the reference intersection length under the i-th preset direction, which is the length of the intersection area between the reference line under the i-th preset direction and the target intersection area. It is the reference length under the i-th preset direction, which is the length of the intersection between the reference line under the i-th preset direction and the target intersection region.

[0114] It should be noted that when A larger value generally indicates a smaller positional change between the first and second labeled connected components in the i-th preset direction. Therefore, To a certain extent, it can characterize the positional changes between the first-labeled connected component and the second-labeled connected component.

[0115] Step 6: Determine the target similarity between the first and second labeled connected regions based on the area overlap ratio and length overlap ratio between the first and second labeled connected regions, the area difference between the first and second labeled connected regions, the distance between the centroid of the reference region and the centroid of the second labeled connected region, and the SSIM (Structural Similarity) value between the first and second labeled connected regions.

[0116] For example, the formula for determining the target similarity between the first-labeled connected component and the second-labeled connected component can be:

[0117] ;

[0118] Where Y is the target similarity between the first labeled connected component and the second labeled connected component. is the normalization function. B is the area overlap ratio between the first-labeled and second-labeled connected components. d is the length overlap ratio between the first-labeled and second-labeled connected components. It is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0, and its value can be 0.0001. It is an absolute value function. It is the numerical value corresponding to the area difference between the first-labeled connected region and the second-labeled connected region. It is the numerical value corresponding to the area of ​​the first labeled connected region. is the area corresponding to the second labeled connected region. D is the distance between the centroid of the reference region and the centroid of the second labeled connected region. h is the SSIM value between the first and second labeled connected regions.

[0119] It should be noted that a larger B value generally indicates a greater overlap in the areas of the first and second labeled connected regions in the same coordinate system, and consequently, a smaller morphological variation between them. d can characterize the positional variation between the first and second labeled connected regions; a larger d value generally indicates a smaller positional variation between them. When... The smaller the value of h, the more similar the areas of the first and second labeled connected regions tend to be, and the more similar their shapes are likely to be. When h is larger, the first and second labeled connected regions tend to be relatively similar. When D is smaller, the first and second labeled connected regions tend to be more geographically close. Therefore, when Y is larger, the first and second labeled connected regions tend to be relatively similar, and the more likely they are to represent the same tissue.

[0120] Step S3: Based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers, connected regions representing the same tissue structure are selected, and connected regions representing the same tissue structure are formed into a connected region sequence.

[0121] In this sequence, different connected components come from different chest CT images, and the order of the connected components in the sequence can be the same as the order of the preset layers corresponding to the chest CT images to which the connected components belong.

[0122] As an example, this step may include the following steps:

[0123] The first step is to designate any preset layer as the marker layer, and designate the preset layer preceding or following the marker layer as the reference layer.

[0124] The second step is to determine any connected component in the chest CT image corresponding to the above-mentioned labeled layer as a candidate connected component, and to determine the chest CT image corresponding to the above-mentioned reference layer as a reference image.

[0125] The third step is to select from the above reference images the connected regions whose target similarity with the above candidate connected regions is greater than a preset similarity threshold, and use them as reference connected regions.

[0126] The preset similarity threshold can be a pre-set threshold, which can be 0.7.

[0127] The fourth step is to select the reference connected domain with the highest target similarity from all the reference connected domains and obtain the connected domain that represents the same tissue structure as the above candidate connected domains. In this way, the matching of connected domains representing the same tissue structure in chest CT images corresponding to adjacent preset layers can be achieved, thereby obtaining the connected domains corresponding to the same tissue structure in different chest CT images.

[0128] Step S4: Perform vascular change pattern analysis on each connected component sequence to obtain the vascular attribution index corresponding to each connected component sequence.

[0129] As an example, this step may include the following steps:

[0130] The first step is to determine the target width corresponding to each connected component in each connected component sequence, thus obtaining the target width sequence for each connected component sequence.

[0131] For example, determining the target width for each connected component in each sequence of connected components may include the following sub-steps:

[0132] The first sub-step involves identifying any one of the connected components in the sequence as the calibration connected component, and then drawing calibration lines in different directions through the centroid of the calibration connected component.

[0133] The number of calibration lines can be a preset number, which can be equal to 180, and the range of the angles corresponding to the calibration lines can be {0°, 1°, 2°, ..., 179°}.

[0134] It should be noted that the more calibration lines in different directions there are, the better the target width of subsequent quantization can characterize the width of the calibration connected region.

[0135] The second sub-step involves determining the length of the intersection of the above-mentioned calibrated connected region and each calibrated line as the reference length corresponding to each calibrated line.

[0136] The third sub-step is to determine the minimum value of the reference length corresponding to all the calibration lines as the target width corresponding to the above-mentioned calibration connected region.

[0137] The second step is to determine the vessel attribution index corresponding to each connected component sequence based on the mean and range of the target width sequence corresponding to each connected component sequence, as well as the target similarity between adjacent connected components in each connected component sequence.

[0138] For example, the formula for determining the vessel attribution index corresponding to a connected component sequence can be:

[0139] ;

[0140] ;

[0141] in, It is the vessel attribution index corresponding to the j-th connected component sequence. j is the sequence number of the connected component sequence. It is a normalization function. It is the mean of the target similarity among all adjacent connected components in the j-th connected component sequence. It is the average of all target widths in the target width sequence corresponding to the j-th connected component sequence. It is the range of the target width sequence corresponding to the j-th connected component sequence. It is a pre-set factor greater than 0, mainly used to prevent the denominator from being 0, and its value can be 0.0001. is the number of connected components in the j-th connected component sequence. a is the index of the connected component in the j-th connected component sequence. It is the target similarity between the a-th connected component and the (a+1)-th connected component in the j-th connected component sequence.

[0142] It should be noted that, in reality, compared to other human tissues in the chest area, blood vessels appear relatively uniform in thickness and narrow in width across different CT images, and often undergo slight changes with heartbeats. The smaller the value, the smaller the relative width of different connected components in the j-th connected component sequence, and the more likely the j-th connected component sequence conforms to the rule of relatively small blood vessel width. When A larger value often indicates a more uniform width variation among different connected components in the j-th connected component sequence, and thus suggests that the j-th connected component sequence is more likely to conform to the characteristic of uniform blood vessel width variation. When the value is relatively small, it often indicates that there is a certain degree of variation between adjacent connected regions in the j-th connected region sequence representing the same tissue, and that the j-th connected region sequence is more likely to conform to the vascular characteristics that undergo small-amplitude changes with heartbeats. Therefore, when The larger the value, the more likely the j-th connected component sequence is to represent a blood vessel.

[0143] Step S5: Based on the vessel attribution index corresponding to different connected region sequences, filter out the vascular connected regions from all connected region sequences.

[0144] As an example, if the vessel attribution index corresponding to the connected component sequence is greater than a preset vessel attribution threshold, then the connected components in the connected component sequence can be identified as vessel connected components. The preset vessel attribution threshold can be a pre-set threshold, such as 0.8.

[0145] Step S6: Based on grayscale values, target calcified regions are selected from the regions formed by all vascular connected domains in each chest CT image, and vascular calcification integral prediction is performed based on the target calcified regions in different chest CT images.

[0146] As an example, this step may include the following steps:

[0147] The first step is to screen out the target calcified region from the region formed by all vascular connected domains in each chest CT image based on grayscale values.

[0148] For example, any chest CT image can be designated as a temporary image. Threshold segmentation can be performed on the region formed by all vascular connected domains in the temporary image, and the sub-region with the larger gray value among the two sub-regions obtained by threshold segmentation can be designated as the target calcified region.

[0149] It should be noted that calcified regions have a higher density and tend to have a higher absorption rate of X-rays. Therefore, calcified regions often have a higher grayscale value in CT images compared to other regions. Thus, the areas with higher grayscale values ​​obtained from threshold segmentation of calcified vessels are often calcified regions. Secondly, directly identifying vascular calcification regions within the area formed by all connected vascular domains in chest CT images can, to some extent, avoid interference from bone tissue with high calcium content.

[0150] The second step involves using the Agatston integral calculation method to predict the vascular calcification score for each chest CT image based on the target calcification region in each chest CT image. The mean of the calcification scores for all chest CT images is then used as the final predicted vascular calcification score.

[0151] It should be noted that the calcification integral prediction of the calcified region using the Agatston integral calculation method can include: assigning the average CT value of the calcified region as the calcification density score; if the calcification density score is less than 130, then update the calcification density score to 0; if the calcification density score belongs to [130, 199], then update the calcification density score to 1; if the calcification density score belongs to [200, 299], then update the calcification density score to 2; if the calcification density score belongs to [300, 399], then update the calcification density score to 3; if the calcification density score is greater than or equal to 400, then update the calcification density score to 4; and multiplying the updated calcification density score by the area of ​​the calcified region to determine the predicted calcification integral value.

[0152] refer to Figure 2Based on the same inventive concept as the above-described method embodiments, this invention provides a coronary artery calcium score prediction system based on chest CT images. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of a coronary artery calcium score prediction method based on chest CT images, specifically including:

[0153] The acquisition and extraction module 201 is used to acquire chest CT images of different preset layers of the patient to be tested at the current time, and to extract the connected components of each chest CT image.

[0154] The target similarity determination module 202 is used to determine the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer.

[0155] The filtering and construction module 203 is used to filter out connected regions representing the same tissue structure based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers, and to form a connected region sequence from the connected regions representing the same tissue structure.

[0156] The vascular change pattern analysis and processing module 204 is used to perform vascular change pattern analysis and processing on each connected component sequence to obtain the vascular attribution index corresponding to each connected component sequence.

[0157] The vascular connected domain filtering module 205 is used to filter out vascular connected domains from all connected domain sequences based on the vascular attribution index corresponding to different connected domain sequences.

[0158] The screening and vascular calcification score prediction module 206 is used to screen out the target calcification region from the region composed of all vascular connected domains in each chest CT image based on gray values, and to predict the vascular calcification score based on the target calcification region in different chest CT images.

[0159] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. For example, as shown... Figure 3 As shown, the computer device 300 includes: a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302. When the processor 302 executes the computer program 303, the computer device can execute any of the coronary artery calcium score prediction methods based on chest CT images described above.

[0160] Based on the same inventive concept as the above-described method embodiments, the present invention provides a server, including a memory and a processor. The memory stores executable program code, and the processor calls and runs the executable program code from the memory, causing the device to execute any of the above-described methods for predicting coronary artery calcium scores based on chest CT images.

[0161] Based on the same inventive concept as the above-described method embodiments, the present invention provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to execute any of the above-described methods for predicting coronary artery calcium scores based on chest CT images.

[0162] Based on the same inventive concept as the above-described method embodiments, the present invention provides a computer-readable storage medium storing computer program code, which, when executed on a computer, causes the computer to perform any of the above-described methods for predicting coronary artery calcium scores based on chest CT images.

[0163] In summary, compared to identifying vascular calcification regions using neural networks, this invention achieves vascular calcification region identification by analyzing chest CT images. This avoids the need for training set acquisition, reducing time waste and improving the efficiency of vascular calcification region identification and vascular calcification score prediction. Secondly, when identifying vascular calcification regions, this invention analyzes the vascular change patterns of connected domain sequences representing the same tissue structure, quantifies the vascular attribution index corresponding to each connected domain sequence, and thus filters out vascular connected domains. Based on grayscale values, it selects target calcification regions from the regions formed by all vascular connected domains in chest CT images, improving the accuracy of vascular calcification region identification and consequently improving the accuracy of vascular calcification score prediction.

[0164] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting coronary artery calcium scores based on chest CT images, characterized in that, Includes the following steps: Acquire chest CT images of the patient to be tested at different preset layers at the current time, and extract connected components for each chest CT image; Based on the multi-feature similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer, the target similarity between different connected regions in the chest CT images corresponding to each adjacent preset layer is determined. Based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers, connected regions representing the same tissue structure are selected, and connected regions representing the same tissue structure are formed into a connected region sequence. The vascular change patterns of each connected component sequence were analyzed to obtain the vascular attribution index corresponding to each connected component sequence. Based on the vessel attribution index corresponding to different connected component sequences, the vessel connected components are screened out from the connected components in all connected component sequences; Based on grayscale values, target calcified regions are selected from the regions formed by all vascular connected domains in each chest CT image, and vascular calcification integrals are predicted based on the target calcified regions in different chest CT images. The step of determining the target similarity between different connected regions in chest CT images corresponding to each adjacent preset layer based on the multi-feature similarity between different connected regions in chest CT images corresponding to each adjacent preset layer includes: determining any two adjacent chest CT images corresponding to preset layers as a first labeled image and a second labeled image, respectively; determining any connected region in the first labeled image as a first labeled connected region, and determining any connected region in the second labeled image as a second labeled connected region; determining the region in the second labeled image that is at the same position as the first labeled connected region as a reference region; and determining the area of ​​the first labeled connected region and the area of ​​the second labeled connected region based on the area of ​​the reference region and the area of ​​the second labeled connected region. The area overlap ratio between the second marked connected regions; the length overlap ratio between the first marked connected regions and the second marked connected regions is determined based on the positional distribution between the centroids of the reference region and the second marked connected regions; the target similarity between the first marked connected regions and the second marked connected regions is determined based on the area overlap ratio and length overlap ratio between the first marked connected regions and the second marked connected regions, the area difference between the first marked connected regions and the second marked connected regions, the distance between the centroid of the reference region and the centroid of the second marked connected regions, and the SSIM value between the first marked connected regions and the second marked connected regions.

2. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The step of determining the area overlap ratio between the first marked connected region and the second marked connected region based on the area of ​​the reference region and the second marked connected region includes: The area of ​​the intersection of the reference region and the second marked connected region is determined as the representative area of ​​the overlap between the first marked connected region and the second marked connected region; The average area of ​​the first marked connected region and the area of ​​the second marked connected region are determined as the relative representative area between the first marked connected region and the second marked connected region. The ratio of the overlapping area between the first marked connected region and the second marked connected region to the relative area is determined as the area overlap ratio between the first marked connected region and the second marked connected region.

3. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The step of determining the length overlap ratio between the first and second marked connected regions based on the positional distribution between the centroids of the reference region and the second marked connected region includes: The intersection of the reference region and the second marked connected component is determined as the target intersection region, and the union of the reference region and the second marked connected component is determined as the target union region. If the centroid of the reference region does not coincide with the centroid of the second marked connected region, then a straight line is drawn through the centroid of the reference region and the centroid of the second marked connected region, and denoted as the marked line. The length of the intersection of the marked line and the target intersection region is determined as the target intersection length. The length of the intersection between the marked line and the target union region is determined as the target union length. The ratio of the target intersection length to the target union length is determined as the length overlap ratio index between the first marked connected region and the second marked connected region. If the centroid of the reference region coincides with the centroid of the second marked connected region, then through the centroid of the second marked connected region, and with each preset direction as the straight line direction, a reference straight line is drawn in each preset direction. Based on the reference straight line in each preset direction, the target intersection region, and the target union region, the initial length coincidence factor in each preset direction is determined. The minimum value among all the initial length coincidence factors in all preset directions is determined as the length coincidence ratio index between the first marked connected region and the second marked connected region.

4. The method for predicting coronary artery calcium scores based on chest CT images according to claim 3, characterized in that, The step of determining the initial length coincidence factor for each preset direction based on the reference line, the target intersection region, and the target union region for each preset direction includes: The length of the intersection of the reference line in each preset direction and the target intersection region is determined as the reference intersection length in each preset direction. The length of the intersection between the reference line in each preset direction and the target region is determined as the reference length in each preset direction. The ratio of the reference intersection length to the reference union length in each preset direction is determined as the initial length coincidence factor in each preset direction.

5. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The step of filtering connected regions representing the same tissue structure based on the target similarity between different connected regions in chest CT images corresponding to all adjacent preset layers includes: Any preset layer is designated as a marker layer, and the preset layer preceding or following the marker layer is designated as a reference layer. Any connected component in the chest CT image corresponding to the marked layer is determined as a candidate connected component, and the chest CT image corresponding to the reference layer is determined as a reference image; Connected regions with a target similarity greater than a preset similarity threshold with the candidate connected regions are selected from the reference image and used as reference connected regions; From all reference connected components, the reference connected component with the highest target similarity to the candidate connected component is selected to obtain the connected component that has the same organizational structure as the candidate connected component.

6. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The process of analyzing vascular change patterns in each connected component sequence yields vascular attribution indices corresponding to each connected component sequence, including: Determine the target width corresponding to each connected component in each connected component sequence to obtain the target width sequence for each connected component sequence; Based on the mean and range of the target width sequence corresponding to each connected component sequence, and the target similarity between adjacent connected components in each connected component sequence, the vessel attribution index corresponding to each connected component sequence is determined.

7. The method for predicting coronary artery calcium scores based on chest CT images according to claim 6, characterized in that, Determining the target width corresponding to each connected component in each connected component sequence includes: Determine any one of the connected components in the sequence of all connected components as the calibration connected component, and draw calibration lines in different directions through the centroid of the calibration connected component; The length of the intersection of the calibration connected region and each calibration line is determined as the reference length corresponding to each calibration line; The minimum value of the reference length corresponding to all calibration lines is determined as the target width corresponding to the calibration connected region.

8. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The step of filtering vascular connected regions from all connected region sequences based on the vessel attribution index corresponding to different connected region sequences includes: If the vessel attribution index corresponding to the connected component sequence is greater than the preset vessel attribution threshold, then the connected component in the connected component sequence is determined as a vessel connected component.

9. The method for predicting coronary artery calcium scores based on chest CT images according to claim 1, characterized in that, The process of selecting target calcified regions from the region comprised of all vascular connected domains in each chest CT image based on grayscale values ​​includes: Any chest CT image is designated as a temporary image. Threshold segmentation is performed on the region formed by all vascular connected domains in the temporary image. The sub-region with the larger gray value among the two sub-regions obtained by threshold segmentation is designated as the target calcified region.

Images