Coronary artery vulnerable plaque identification method, storage medium and electronic equipment

Abstract

The invention provides an identification method for coronary artery vulnerable plaques, a storage medium and electronic equipment, and relates to the technical field of biomedical engineering.The identification method comprises the steps that blood vessel segmentation is conducted on a CCTA image; for each blood vessel segment obtained by segmentation, the following processing steps are performed: straightening curved surface reconstruction is performed to generate an SCPR image, and plaque detection is performed based on the SCPR image to position a first plaque area; analyzing the PCAT region corresponding to the blood vessel segment, and determining a CT value distribution diagram in spatial registration with the first plaque region; performing dynamic re-segmentation processing on the first plaque region according to the CT value distribution diagram to determine a second plaque region; and converting an image corresponding to the second plaque region, and inputting the converted image into the trained model to obtain an identification result. In the application, on the basis of the existing form-based vulnerable plaque identification, PCAT analysis is introduced to carry out joint identification, so that improvement of the overall identification performance is facilitated.

Description

Technical Field

[0001] The present application relates to the technical field of biomedical engineering, and particularly to a method, a storage medium, and an electronic device for identifying vulnerable coronary plaques. Background Art

[0002] Acute coronary syndrome (ACS) has a relatively high mortality rate and accounts for a large proportion of deaths caused by cardiovascular diseases. Research shows that approximately 70% of ACS events are triggered by the rupture of vulnerable plaques. Coronary computed tomography angiography (CCTA) is a widely recognized non-invasive diagnostic tool, which is used for early screening and definitive diagnosis of coronary heart disease, and can identify vulnerable plaques in the coronary arteries by analyzing CCTA images, providing important medical data support for clinical applications.

[0003] In the process of implementing the present invention, the applicant noticed that the prior art mainly focuses on analyzing the morphological characteristics of plaques (such as positive remodeling, low density, etc.) using CCTA images. For example, in the patent "A Method and System for Identifying Vulnerable Coronary Plaques Based on CCTA Images" with publication number CN117455878A, by segmenting the original CCTA images in segments, extracting the centerline of the blood vessels to generate SCPR images, and using a plaque detection model to locate the plaque range, then converting the three-dimensional probe images into a two-dimensional format, and adopting an identification model based on ViT to output the probabilities of four vulnerable signs (positive remodeling, low density, napkin ring, punctate calcification) to identify vulnerable plaques. Although this method improves the efficiency compared with the subjective judgment of doctors, due to relying only on the morphological characteristics of plaques, the identification and segmentation of diffuse lesions (or overlapping plaques) are not good. For example, different sub-regions of a diffuse lesion meet a corresponding morphological characteristic, and the entire region is determined as a vulnerable plaque.

[0004] Therefore, in the implementation of identifying vulnerable plaques, how to make technical improvements to address this defect and improve the comprehensive performance of identification has become an urgent technical problem to be solved. Summary of the Invention

[0005] To overcome at least to some extent the problems existing in the related art, embodiments of the present application provide a method, a storage medium, and an electronic device for identifying vulnerable coronary plaques, and introduce the analysis of perivascular adipose tissue during the identification process to jointly distinguish plaques, which is beneficial to improving the comprehensive performance of vulnerable plaque identification.

[0006] First Aspect Some embodiments of the present application provide a method for identifying vulnerable coronary plaques, and the identification method includes: Performing vascular segmentation and segmentation on the obtained CCTA images to generate multiple vascular segments; The following processing steps are performed separately for each vascular segment: Perform straightening surface reconstruction to generate an SCPR image, and perform plaque detection based on the SCPR image to locate the first plaque region; And analyze the PCAT region corresponding to the vascular segment to determine the CT value distribution map spatially registered with the first plaque region; Perform dynamic re-segmentation processing on the first plaque region according to the CT value distribution map to determine the second plaque region; Convert the image corresponding to the second plaque region, and input the converted image into the trained model to obtain the vulnerable plaque recognition result.

[0007] In a possible implementation manner, the dynamic re-segmentation processing includes: Calculate the three-dimensional gradient field of the CT value distribution map and determine the adaptive threshold; Based on the three-dimensional gradient field, cluster the continuous voxels with gradients greater than the adaptive threshold and within the first plaque region into the second plaque region.

[0008] In a possible implementation manner, based on the following expression, determine the adaptive threshold: Where, represents the adaptive threshold, represents the average value of the gradient field, represents the standard deviation, and k represents the adjustment coefficient positively correlated with the vascular diameter.

[0009] In a possible implementation manner, 4. According to the recognition method described in claim 2, the process of performing PCAT analysis includes: Define an annular analysis region by expanding the length of the corresponding vascular diameter outward with the adventitia as the reference, extract the voxel set with CT values between -190 HU and -30 HU within this region, and generate a three-dimensional CT value distribution map to be registered according to this voxel set.

[0010] In a possible implementation manner, the spatial registration is implemented through a feature pyramid registration network.

[0011] In a possible implementation manner, the process of performing vascular segmentation and segmentation on the obtained CCTA image includes: Use a 3D U-Net network to automatically segment the coronary artery main trunk and branches, and output a three-dimensional vascular mask; Calculate the ratio of the curvature change rate to the local vascular diameter along the center line of the segmented blood vessel, and judge and confirm the mutation point of the blood vessel morphology based on this ratio. Use the mutation point as the segmentation boundary to perform blood vessel segmentation.

[0012] In a possible implementation manner, the process of detecting and locating the first plaque region based on the SCPR image includes: on the SCPR image, plaque detection is performed in the central point area one by one through the developed plaque detection model to obtain the plaques on the blood vessels, and the range of each first plaque region is represented by the positions of the starting and ending central points.

[0013] In a possible implementation manner, the conversion of the image corresponding to the second plaque region is specifically: within the plaque range, the pixel values of each layer of three-dimensional probe image are mapped to the common picture format, and the third dimension is 1; the trained model is constructed based on the ViT model.

[0014] In a second aspect Some embodiments of the present application provide a computer-readable storage medium, on which program code is stored, and when the program code is executed by a processor, the method described in any implementation manner of the above first aspect is implemented.

[0015] In a third aspect Some embodiments of the present application provide an electronic device, including: A memory, on which an executable program is stored; A processor, configured to execute the executable program in the memory to implement the steps of the method described in any implementation manner of the above first aspect.

[0016] The method for identifying vulnerable coronary plaques provided by the embodiments of the present application includes: performing vascular segmentation and segmentation on the acquired CCTA image to generate multiple vascular segments; respectively performing the following processing steps for each vascular segment: performing straightening surface reconstruction to generate an SCPR image, detecting and locating the first plaque region based on the SCPR image; and analyzing the perivascular adipose tissue (PCAT) region corresponding to the vascular segment to determine the CT value distribution map spatially registered with the first plaque region; performing dynamic re-segmentation processing on the first plaque region according to the CT value distribution map to determine the second plaque region; converting the image corresponding to the second plaque region, and inputting the converted image into the trained model to implement vulnerable plaque identification. The technical solution of the present application, on the basis of the existing coronary plaque detection implementation, through combining the CT value distribution map of the perivascular adipose tissue for dynamic re-segmentation processing, makes the identification of diffuse lesions or overlapping plaques more accurate, and the vulnerable plaque identification based on this has better comprehensive identification performance compared with the existing methods.

[0017] Other advantages, objectives, and features of the present application will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. Brief Description of the Drawings

[0018] The drawings are used to provide a further understanding of the technical solutions of the present application or the prior art, and constitute a part of the specification. Among them, the drawings expressing the embodiments of the present application are used together with the embodiments of the present application to explain the technical solutions of the present application, but do not constitute a limitation to the technical solutions of the present application.

[0019] Figure 1 It is a schematic flow chart for the method of identifying vulnerable coronary artery plaques in an embodiment of the present application; Figure 2 It is a schematic flow chart for vascular segmentation and segmentation based on CCTA images in an embodiment of the present application; Figure 3 It is a schematic structural diagram of an electronic device provided in an embodiment of the present application. Detailed Description of the Embodiments

[0020] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described in detail below. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other implementation manners obtained by those of ordinary skill in the art without creative efforts fall within the scope protected by the present application.

[0021] As described in the background art, acute coronary syndrome (ACS) has a relatively high mortality rate and accounts for a relatively large proportion of deaths caused by cardiovascular diseases. Research shows that approximately 70% of ACS events are triggered by the rupture of vulnerable plaques. Coronary computed tomography angiography (CCTA) is a widely recognized non-invasive diagnostic tool, which is used for early screening and clear diagnosis of coronary heart disease, and can identify vulnerable plaques in the coronary arteries by analyzing CCTA images, providing important medical data support for clinical applications.

[0022] In the process of implementing the present invention, the applicant noticed that the prior art mainly focuses on analyzing the morphological characteristics of plaques (such as positive remodeling, low density, etc.) using CCTA images. For example, in the patent "A method and system for identifying vulnerable coronary plaques based on CCTA images" with publication number CN117455878A, by segmenting the original CCTA images in segments, extracting the centerline of the blood vessel to generate an SCPR image, and using a plaque detection model to locate the plaque range, then converting the three-dimensional probe image into a two-dimensional format, and adopting an identification model based on ViT to output the probabilities of four vulnerable signs (positive remodeling, low density, napkin ring, punctate calcification) to identify vulnerable plaques. Although this method improves the efficiency compared to the subjective judgment of doctors, due to relying only on the morphological characteristics of plaques, the identification and segmentation of diffuse lesions (or overlapping plaques) are not good. For example, different sub-regions of a diffuse lesion meet a corresponding morphological characteristic, and the entire region is determined as a vulnerable plaque.

[0023] According to relevant medical research, the inflammatory state of pericoronary adipose tissue (PCAT) is highly correlated with plaque vulnerability. The fat attenuation index (FAI) of the coronary artery is a newly proposed parameter in the industry, which is defined as the average CT value of PCAT within the range where the radial distance from the outer wall of the coronary artery is equal to the average diameter of the coronary artery, and it can be used as a reference for evaluating the inflammatory state.

[0024] Based on this, according to the above research results, the technical solution of this application introduces the analysis of FCAT in the implementation of vulnerable plaque identification to jointly achieve the discrimination of vulnerable plaques, so as to overcome the defects existing in the prior art to a certain extent.

[0025] In one embodiment, as Figure 1 shown, the method for identifying vulnerable coronary plaques proposed in this application includes: A segmentation and segmentation step of performing blood vessel segmentation and segmentation on the obtained CCTA image to generate multiple blood vessel segments.

[0026] Those skilled in the art can easily understand that this process involves accurately identifying and separating each independent blood vessel structure, and dividing the entire coronary artery into multiple continuous but independent blood vessel segments according to its anatomical characteristics and location. This can define the starting and ending positions of each blood vessel segment, providing a basis for more accurate analysis of each segment of the blood vessel in the follow-up (such as facilitating the generation of SCPR images after segmentation), and helping to improve the accuracy and reliability of the overall identification method. In practice, existing related implementation methods can be used to implement this step, such as using an existing coronary artery segmentation and segmentation model to perform segmentation and segmentation, etc.

[0027] Subsequently, the following processing steps are respectively performed for each vascular segment: Step S110: Perform straightened curved planar reconstruction to generate an SCPR image, and perform plaque detection based on the SCPR image to locate the first plaque region.

[0028] In this step, based on the CPR curved planar reconstruction technology, straightened curved planar reconstruction (SCPR) is performed on these vascular segments to generate an SCPR image. This process involves transforming the curved vascular path into a straight or nearly straight form for display, so that the vascular regions that are difficult to directly observe and analyze due to the viewing angle or anatomical structure complexity become easy to evaluate; furthermore, the generated SCPR image can be used to locate the plaque region by using relevant plaque detection methods. For example, in this application, plaque detection is performed in the center point region one by one on the SCPR image through the developed plaque detection model to obtain the plaques on the blood vessel, and the range of each first plaque region is represented by the positions of the starting and ending center points. This implementation method belongs to the mature implementation methods in the existing related technologies. Here, this application does not elaborate on it in detail, but only briefly introduces its implementation principle. In this implementation method, a series of center points distributed along the vascular path are first determined, and these points are used as the centers of the detection windows to ensure coverage of the entire vascular segment. For each center point region, the plaque detection model will analyze the image features in this local region, including but not limited to morphological features such as edge information and texture, to identify possible plaques; once a plaque is detected, the range of each first plaque region is represented by recording the positions of its starting and ending center points.

[0029] Different from the prior art, on the basis of performing step S110, this application further performs step S120 to analyze the PCAT region corresponding to the vascular segment and determine the CT value distribution map spatially registered with the first plaque region; as described in the previous introduction parts about PCAT and FAI, the state of the periapical adipose tissue PCAT is related to the presence of plaques. In this step here, the PCAT situation corresponding to the blood vessel is analyzed based on the spatial correspondence relationship, independent of the blood vessel itself.

[0030] Specifically, the process of performing PCAT analysis here includes: referring to the definition of FAI, defining an annular analysis region by extending the length of the corresponding blood vessel diameter outward with the adventitia as the reference, extracting the voxel set with CT values ranging from -190 HU to -30 HU within this region, and generating a three-dimensional CT value distribution map to be registered based on this voxel set; it is easy to understand here that the CT values from -190 HU to -30 HU correspond to adipose tissue, so based on the limitations of spatial position and CT value, the PCAT region corresponding to the blood vessel can be effectively determined, and further the CT value distribution information required for subsequent processing can be determined.

[0031] On this basis, registration processing with the first plaque region also needs to be performed to enable subsequent related joint analysis. In the related technologies of medical image processing, registration processing is a basic but extremely important operation, which is to find the best spatial correspondence relationship between different images through computational methods. This integrates image information from multiple sources into a common coordinate system, so that these images can be directly compared and analyzed. For the PCAT analysis of this application, the main purpose of registration is to accurately align the three-dimensional CT value distribution information of the plaque region of interest with the surrounding tissues. For example, as a specific implementation method, the spatial registration here is achieved through a feature pyramid registration network to finally determine the CT value distribution map that is spatially registered with the first plaque region.

[0032] In the specific implementation of this spatial registration method, the multi-scale surface projection features of the SCPR image are first extracted. This step is crucial for accurately identifying and locating the plaque area. In actual implementation, the improved FPN (Feature Pyramid Network) structure can be used for projection feature extraction. The backbone network in this network uses 3D ResNet-34, a deep learning model that is particularly suitable for processing three-dimensional medical imaging data, which is used to extract rich underlying features from the original SCPR image. After the feature extraction stage, the number of channels is adjusted using the lateral connection of the 1×1×1 convolution kernel to ensure that the information between feature maps at different levels can be effectively fused. Compared with the traditional deconvolution operation, the trilinear interpolation technology is used in this implementation for upsampling layer processing, thereby avoiding the possible chessboard effect; then, in order to align the three-dimensional voxel space corresponding to the three-dimensional CT value distribution map to be registered, a deformable convolution layer is introduced. This convolution layer has stronger spatial adaptability and can more accurately capture the complex deformation in the target area. In the registration process, in order to evaluate and optimize the registration accuracy, a composite loss function is constructed based on the Chamfer Distance and the Gradient Consistency Loss. It is easy for those skilled in the art to understand that the Chamfer Distance is an effective measure of the similarity between two point sets, while the Gradient Consistency Loss helps to ensure the matching degree of the edge details of the registered images. The two work together to improve the accuracy and robustness of the entire registration process.

[0033] Continue back Figure 1 After step S120, step S130 may be performed to dynamically re-segment the first plaque region according to the CT value distribution map to determine the second plaque region; Specifically, the dynamic re-segmentation processing here includes: calculating the three-dimensional gradient field of the CT value distribution map and determining the adaptive threshold; this process involves performing spatial differential operations on the data corresponding to the CT value distribution map, thereby obtaining the rate of change along three dimensions (X, Y, Z) at each voxel point. Through this operation, the degree of PCAT differentiation can be effectively highlighted, thereby reflecting the differences between different regions within the first plaque area; in the technical scenario of this application, this difference can be used to distinguish different sub-regions of diffuse lesions, and then based on the three-dimensional gradient field, the continuous voxels with a gradient greater than the adaptive threshold and located in the first plaque area are clustered into the second plaque area, thereby realizing re-segmentation processing.

[0034] In the above process, an adaptive threshold implementation method is adopted, and the threshold size is dynamically adjusted according to the characteristics of the PCAT corresponding to the corresponding first plaque region, ensuring that different sub-regions of diffuse lesions can be accurately distinguished even in the case of high background noise or small plaque density differences, which has higher flexibility and accuracy compared with the traditional fixed threshold method. As a specific implementation method, based on the following expression (1), the adaptive threshold is determined: (1) In expression (1), represents the adaptive threshold, represents the average value of the gradient field, represents the standard deviation, and k represents a regulation coefficient that is positively correlated with the blood vessel diameter.

[0035] For ease of implementation, based on the statistical analysis of the structural characteristics of the human coronary artery, the calculation of the above regulation coefficient k satisfies k = 0.15 + 0.03d, where d represents the diameter of the current blood vessel segment (unit: mm), and the value range of k is limited to 0.2 ≤ k ≤ 0.5. In this way, the regulation coefficient k is dynamically adjusted according to the diameter of the specific blood vessel segment, making the processing process more suitable for individual differences. This method takes into account the natural variations of blood vessels in different parts and among different patients, thereby improving the accuracy and reliability of the processing results; while limiting the value range, it avoids underprocessing caused by too small a coefficient and prevents overprocessing caused by too large a coefficient. Such a design is beneficial to ensuring the stability and applicability of the algorithm in various application scenarios.

[0036] After the determination of the second plaque region in step S130, step S140 is continued. The image corresponding to the second plaque region is converted, and the converted image is input into the trained model to obtain the vulnerable plaque recognition result. In step S140, the conversion of the image corresponding to the second plaque region is specifically: within the plaque range, the pixel values of each layer of the three-dimensional probe image are mapped to the ordinary picture format, and the third dimension is 1; the above-mentioned trained model is constructed based on the ViT model.

[0037] It is easy to understand that the implementation method of the above step S140 is the prior art for vulnerable plaque recognition. By adopting this method or other related technical means for vulnerable plaque recognition based on morphological features, better recognition performance can be achieved on the basis of the dynamic re-segmentation of the present application.

[0038] For the convenience of understanding the technical solution, the implementation manner of S140 is briefly introduced here. In step S140, the trained model is a coronary vulnerable plaque recognition model based on the ViT model. As is known to those skilled in the art, ViT is the abbreviation of Vision Transformer, that is, Visual Transformer, which is a model that applies the Transformer architecture to the field of computer vision. Specifically in the technical scenario of this application, similar to the prior art, the model structure includes the following structures: Input encoder: It is used to divide the input image into a set of fixed-size image patches, and then extract features from each image patch through a convolutional neural network CNN (such as Figure 2 ) to obtain a set of vector representations as the input sequence. Since the length of each plaque is different and the number of SCPR probe images obtained is also different, linear interpolation is first performed to fix the number of images before encoding to N, and then the N two-dimensional images pass through the encoder to obtain N×M-dimensional image vectors.

[0039] Position encoder: It is used to add position embedding vectors to represent the relative and absolute position information of each position in the input sequence.

[0040] Encoder: It consists of multiple encoding layers. Each encoding layer contains a multi-head self-attention mechanism and a feed-forward fully connected network. The self-attention mechanism can simultaneously focus on the dependencies between different positions in the sequence, and through the multi-head mechanism, feature extraction can be performed from different subspaces; the feed-forward fully connected network is responsible for performing non-linear transformation on the features.

[0041] Normalization layer: It is used to perform normalization processing on the output of the encoding layer, such as Layer Normalization or Batch Normalization.

[0042] Classifier: A linear layer is connected after the encoding layer to map the output of the encoding layer to the prediction probabilities of four signs.

[0043] Based on the initial model with the above structure, during the pre-training process of the model, first, the constructed vulnerable plaque recognition model is pre-trained and fine-tuned through a natural image dataset, and then the model is initialized with the trained parameters and retrained on the vulnerable plaque dataset to obtain a model that can be used in practice.

[0044] When the model is implemented and applied in step S140, the converted image needs to be input into the trained model to obtain the vulnerable plaque recognition result. It should be noted that the output of the model is the probability of each sign (positive remodeling, low density, napkin ring, and punctate calcification) of the vulnerable plaque. Based on the output of the model, a probability threshold is set for different signs. If the probability output by the model for a certain sign is greater than or equal to the threshold, it is considered that the model recognizes this sign. If the model recognizes two or more signs, it is considered that the plaque is a vulnerable plaque; otherwise, it is a non-vulnerable plaque.

[0045] It is easy to understand that after such identification processing is performed on each vascular segment (steps S110 to S140), the identification results corresponding to each vascular segment can be obtained. Then, the obtained identification processing results are summarized and displayed. For example, based on actual needs, the number and location area of the identified vulnerable plaques are visually output, so as to provide a reference basis for the relevant diagnosis and discrimination of doctors.

[0046] Based on the existing coronary artery plaque detection implementation, the technical solution of this application performs dynamic re-segmentation processing by combining the CT value distribution map of the perivascular adipose tissue, making the identification of diffuse lesions or overlapping plaques more accurate. On this basis, the identification of vulnerable plaques has better comprehensive identification performance compared with the existing methods.

[0047] All the defects existing in the above solutions are the results obtained by the inventors after practice and careful research. Therefore, the discovery process of the above problems and the solutions proposed by this application for the above problems in the following text should be the contributions made by the inventors to this application during the process of this application.

[0048] Based on the above embodiments, as Figure 2 shown, as a specific implementation manner, in one embodiment, this application also adopts a specific implementation manner for vascular segmentation and segmentation. Specifically, the process of performing vascular segmentation and segmentation on the obtained CCTA image includes: Step S210, using a 3D U-Net network to automatically segment the coronary artery main trunk and branches and output a three-dimensional vascular mask. In the actual implementation of this step, the network model needs to be trained with a large amount of labeled data so that it can effectively identify and separate the vascular structure, thereby generating a detailed three-dimensional vascular mask. This mask accurately depicts the contour and position of the blood vessel in three-dimensional space, facilitating segmentation from the original CCTA image and providing key basic information for subsequent analysis.

[0049] Step S220: Calculate the ratio of the curvature change rate to the local blood vessel diameter along the centerline of the segmented blood vessel, and based on this, determine and confirm the mutation points of the blood vessel morphology. Use the mutation points as the segmentation boundaries to segment the blood vessel. In this step, on the basis of completing the blood vessel segmentation, further calculate the ratio of the curvature change rate to the local blood vessel diameter along the centerline of the segmented blood vessel. By this method, the change of the blood vessel morphology can be quantitatively evaluated. According to the calculated ratio and based on the specific method configuration, such as comparing the ratio with relevant thresholds, the method execution entity can intelligently identify the mutation points on the blood vessel morphology that distinguish the main trunk from the branches. Furthermore, based on the determined mutation points, segment the main trunk and branches of the blood vessel, divide the blood vessel into multiple sub-segments, and each sub-segment represents a relatively stable and feature-consistent part of the blood vessel. This segmentation strategy helps to analyze the overall anatomical structure of the blood vessel more carefully and also helps to reduce the demand for computing power in the overall recognition process.

[0050] In the technical solution of the present application, compared with the prior art, by combining the CT value distribution map of the adipose tissue around the blood vessel for dynamic re-segmentation processing, the recognition of complex lesion conditions (such as diffuse lesions or overlapping plaques) is more accurate. Compared with the traditional method that only relies on morphological features, this method can more accurately locate and segment the plaque area. Use an adaptive threshold to calculate the three-dimensional gradient field and cluster continuous voxels into the second plaque area. This process has a high degree of automation and reduces the need for manual intervention, thereby improving the efficiency of the overall detection process. By quantitatively analyzing the PCAT area corresponding to the blood vessel segment and using the feature pyramid registration network to achieve spatial registration, the consistency and reliability of the diagnosis results among different cases are ensured, and the uncertainty caused by the subjective judgment differences of doctors is avoided. Map the pixel values of the three-dimensional probe image to the ordinary picture format, and build a trained recognition model based on the ViT model, which effectively improves the model's understanding ability of the input image, and further improves the accuracy of vulnerable plaque recognition. Use the 3D U-Net network to automatically segment the main trunk and branches of the coronary artery, and determine the mutation points as the segmentation boundaries according to the ratio of the curvature change rate to the local blood vessel diameter, which not only improves the segmentation accuracy but also better adapts to the complex blood vessel structure, facilitating the precise execution of subsequent plaque detection.

[0051] In one embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable medium stores program code, and when the program code is executed by a processor and runs on a computer, the method described in the above embodiment is implemented.

[0052] In addition, the present application also proposes an electronic device Figure 3 which is a schematic structural diagram of the electronic device provided in an embodiment of the present application, as Figure 3 shown. The electronic device 300 includes: A memory 301, on which an executable program is stored; A processor 302, configured to execute the executable program in the memory 301 to implement the steps of the above method.

[0053] Regarding the electronic device 300 in the above embodiments, the specific manner in which its processor 302 executes the program in the memory 301 has been described in detail in the embodiments related to the method, and will not be elaborated here.

[0054] As described above, the above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0055] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be seen in the same or similar content of other embodiments.

[0056] It should be noted that in the description of the present application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present application, unless otherwise specified, the meaning of "a plurality" refers to at least two.

[0057] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0058] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for identifying vulnerable coronary plaques, characterized in that, Comprising: Performing vascular segmentation and segmentation on the acquired CCTA images to generate a plurality of vascular segments; Performing the following processing steps for each vascular segment respectively: Performing straightening surface reconstruction to generate an SCPR image, and performing plaque detection on the SCPR image to locate the first plaque region; And analyzing the PCAT region corresponding to the vascular segment to determine a CT value distribution map spatially registered with the first plaque region; Performing dynamic re-segmentation processing on the first plaque region according to the CT value distribution map to determine a second plaque region; Converting the image corresponding to the second plaque region, and inputting the converted image into a trained model to obtain a vulnerable plaque recognition result.

2. The method for identifying vulnerable coronary artery plaques according to claim 1, wherein, The dynamic re-segmentation processing includes: Calculating a three-dimensional gradient field of the CT value distribution map and determining an adaptive threshold; Based on the three-dimensional gradient field, clustering continuous voxels with gradients greater than the adaptive threshold and within the first plaque region into a second plaque region.

3. The method for identifying vulnerable coronary artery plaques according to claim 2, wherein, Determining the adaptive threshold based on the following expression: Among them, represents the adaptive threshold,[[]]END]] represents the average value of the gradient field,[[]]END]] represents the standard deviation, and k represents the adjustment coefficient that is positively correlated with the blood vessel diameter.[[]]END]] 4. The method for identifying vulnerable coronary artery plaques according to claim 2, wherein, The process of performing PCAT analysis includes: Defining an annular analysis region by expanding the length of the corresponding vascular diameter outward with the adventitia as a reference, extracting a set of voxels with CT values between -190 HU and -30 HU within this region, and generating a three-dimensional CT value distribution map to be registered based on this set of voxels.

5. The method for identifying vulnerable coronary artery plaques according to claim 2, wherein The spatial registration is implemented through a feature pyramid registration network.

6. The method for identifying vulnerable coronary artery plaques according to claim 1, wherein, The process of performing vascular segmentation and segmentation on the acquired CCTA images includes: Using a 3D U-Net network to automatically segment the coronary artery main trunk and branches, and outputting a vascular three-dimensional mask; Calculating the ratio of the curvature change rate to the local vascular diameter along the center line of the segmented blood vessel, and judging and confirming the mutation points of the blood vessel morphology based on this, and using the mutation points as segmentation boundaries to perform vascular segmentation.

7. The method for identifying vulnerable coronary artery plaques according to claim 1, wherein The process of performing plaque detection on the SCPR image to locate the first plaque region includes: on the SCPR image, performing plaque detection within the center point area one by one through a developed plaque detection model to obtain plaques on the blood vessel, and representing the range of each first plaque region by the positions of the starting and ending center points.

8. The method for identifying vulnerable coronary artery plaques according to claim 1, wherein, The conversion of the image corresponding to the second plaque region specifically is: within the plaque range, mapping the pixel values of each layer of three-dimensional probe image to a common picture format, and the third dimension is 1; the trained model is constructed based on the ViT model.

9. A computer-readable storage medium having program code stored thereon, characterized in that, When the program code is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

10. An electronic device, characterized in that, Comprising: A memory having an executable program stored thereon; A processor for executing the executable program in the memory to implement the steps of the method according to any one of claims 1-8.

Images