Carotid artery plaque stenosis category detection method and device and electronic equipment
By segmenting blood vessels and plaques in head and neck CTA images and combining sparse convolution and convolution feature extraction, the problem of inaccurate manual diagnosis of carotid artery plaque stenosis categories is solved, achieving more accurate plaque stenosis category detection and reducing hardware costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INFERVISION MEDICAL TECH CO LTD
- Filing Date
- 2022-11-30
- Publication Date
- 2026-07-07
Smart Images
Figure CN115937127B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of medical image processing, specifically to a method and apparatus for detecting carotid plaque stenosis categories, an electronic device, and a computer-readable storage medium. Background Technology
[0002] Different types of plaques can often appear in the carotid arteries. The plaque stenosis category (i.e., the different degrees of stenosis of the plaque) is closely related to the abnormality of the carotid arteries. Therefore, it is crucial to detect the carotid artery plaque stenosis category.
[0003] Currently, manual diagnosis of carotid artery plaque stenosis type is often required in clinical practice. However, manual diagnosis relies too heavily on the clinical experience of doctors, and the test results of different doctors also vary greatly. In other words, there is a problem of inaccurate plaque stenosis type test results. Summary of the Invention
[0004] In view of this, embodiments of this application provide a method and apparatus for detecting carotid plaque stenosis category, an electronic device, and a computer-readable storage medium, to solve the problem of inaccurate detection results of carotid plaque stenosis category by manual diagnosis.
[0005] According to a first aspect of the embodiments of this application, a method for detecting carotid artery plaque stenosis category is provided, comprising: determining, based on a head and neck CTA image, vascular category segmentation data of the head and neck CTA image, plaque segmentation data, and vascular midline, a plaque region image and a plaque straightening region image corresponding to plaques included in the carotid artery in the head and neck CTA image; performing sparse convolution feature extraction operation and convolution feature extraction operation on the plaque region image and the plaque straightening region image respectively to obtain plaque stenosis rate prediction data; and determining the plaque stenosis category detection result corresponding to the plaque based on the plaque stenosis rate prediction data.
[0006] In one embodiment, sparse convolutional feature extraction and convolutional feature extraction operations are performed on the patch region image and the straightened patch region image, respectively, to obtain patch narrowing rate prediction data. This includes: processing the patch region image using the sparse convolutional feature extraction branch in the patch narrowing rate prediction model to obtain a first feature vector corresponding to the patch region image; processing the straightened patch region image using the convolutional feature extraction branch in parallel with the sparse convolutional feature extraction branch in the patch narrowing rate prediction model to obtain a second feature vector corresponding to the straightened patch region image; and merging the first and second feature vectors to obtain the patch narrowing rate prediction data.
[0007] In one embodiment, determining the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery in the head and neck CTA image based on the head and neck CTA image, the vessel category segmentation data of the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the vessel midline, includes: determining the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery vessel midline in the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the vessel midline; and determining the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery based on the carotid artery CTA image, the carotid artery plaque segmentation data, and the carotid artery vessel midline.
[0008] In one embodiment, determining the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from head and neck CTA images, plaque segmentation data of head and neck CTA images, and vascular midlines based on vascular category segmentation data includes: determining a first bounding box corresponding to the carotid artery based on vascular category segmentation data; determining a second bounding box corresponding to the carotid artery based on the first bounding box and a preset margin value; and using the second bounding box to extract the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from the head and neck CTA images, plaque segmentation data of head and neck CTA images, and vascular midlines.
[0009] In one embodiment, determining the plaque region image and the straightened plaque region image corresponding to the plaques included in the carotid artery based on carotid CTA images, carotid plaque segmentation data, and the carotid artery midline includes: straightening the carotid CTA images and carotid plaque segmentation data based on the carotid artery midline to obtain a straightened image; and resampling the straightened image and the carotid CTA images based on the location information of the connected regions of the plaques in the carotid plaque segmentation data to obtain the plaque region image and the straightened plaque region image.
[0010] In one embodiment, determining the plaque stenosis category detection result corresponding to a plaque based on plaque stenosis rate prediction data includes: performing a sliding window sampling operation on the plaque; determining the window plaque stenosis rate prediction value corresponding to each window in the sliding window sampling operation based on the plaque stenosis rate prediction data; determining the largest window plaque stenosis rate prediction value as the plaque stenosis rate prediction value corresponding to the plaque; and determining the plaque stenosis category detection result based on a preset relationship between the plaque stenosis rate prediction value and the plaque stenosis category.
[0011] In one embodiment, determining the predicted window plaque stenosis rate for each window in a sliding window sampling operation based on plaque stenosis rate prediction data includes: for each window, determining the coordinates of the center point of the window; and using the predicted plaque stenosis rate data located at the center point coordinates as the predicted window plaque stenosis rate for the window.
[0012] According to a second aspect of the embodiments of this application, a carotid artery plaque stenosis category detection device is characterized by comprising: a first determining module configured to determine, based on a head and neck CTA image, vascular category segmentation data of the head and neck CTA image, plaque segmentation data, and vascular midline, a plaque region image and a plaque straightening region image corresponding to plaques included in the carotid artery in the head and neck CTA image; a feature extraction module configured to perform sparse convolution feature extraction operation and convolution feature extraction operation on the plaque region image and the plaque straightening region image, respectively, to obtain plaque stenosis rate prediction data; and a second determining module configured to determine the plaque stenosis category detection result corresponding to the plaque based on the plaque stenosis rate prediction data.
[0013] According to a third aspect of the embodiments of this application, an electronic device is provided, including: a processor; and a memory storing computer program instructions, which, when executed by the processor, cause the processor to perform the carotid plaque stenosis category detection method as described in the first aspect above.
[0014] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, on which computer program instructions are stored, which, when executed by a processor, cause the processor to perform the carotid plaque stenosis category detection method as described in the first aspect above.
[0015] The carotid plaque stenosis category detection method provided in this application effectively utilizes the sparsity of plaques by performing sparse convolution feature extraction and convolution feature extraction operations on plaque region images and straightened plaque region images respectively. This effectively utilizes image information, improves the robustness of feature extraction, and ultimately enhances the accuracy of plaque stenosis category detection results. Furthermore, the method provided in this application, due to the use of sparse convolution feature extraction, reduces computational requirements, thereby reducing the demand on device memory and ultimately lowering hardware costs. Attached Figure Description
[0016] Figure 1 The diagram shown is a flowchart of an arterial plaque stenosis category detection method provided in an embodiment of this application.
[0017] Figure 2 The diagram shown is a flowchart illustrating the process of performing sparse convolution feature extraction and convolution feature extraction operations on patch region images and patch straightening region images respectively, in an embodiment of this application, to obtain patch narrowing rate prediction data.
[0018] Figure 3a The diagram shown is a flowchart of a model training method provided in an embodiment of this application.
[0019] Figure 3bThe diagram shown is a schematic diagram of the residual module of the SpConvNet model provided in an embodiment of this application.
[0020] Figure 4 The diagram shown is a flowchart illustrating the detection results of determining the plaque stenosis category corresponding to a plaque based on plaque stenosis rate prediction data according to an embodiment of this application.
[0021] Figure 5 The diagram shown is a structural schematic of an arterial plaque stenosis category detection device provided in an embodiment of this application.
[0022] Figure 6 The diagram shown is a structural schematic of the first determining module provided in an embodiment of this application.
[0023] Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application. Detailed Implementation
[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0025] Carotid artery plaque stenosis is classified according to the plaque stenosis rate. The higher the plaque stenosis rate, the more severe the head and neck obstruction. When the stenosis reaches a certain degree, normal blood circulation will be affected, potentially leading to cerebral hemorrhage. Therefore, the classification of carotid artery plaque stenosis is of great significance for clinical diagnosis. Based on the plaque stenosis rate, plaque stenosis is generally classified into five categories: no stenosis, mild stenosis, moderate stenosis, severe stenosis, and occlusion.
[0026] Currently, manual diagnosis of carotid artery plaque stenosis type is usually required in clinical practice. However, manual diagnosis is not only time-consuming and laborious, but also relies too much on the clinical experience of doctors. The test results of different doctors also vary greatly. In other words, there is a problem of inaccurate plaque stenosis type test results.
[0027] To address the aforementioned issues, this application provides a method for detecting carotid plaque stenosis categories. By performing sparse convolution feature extraction and convolution feature extraction operations on both the plaque region image and the straightened plaque region image, the method effectively utilizes the sparsity of the plaque, thereby effectively leveraging image information, improving the robustness of feature extraction, and ultimately enhancing the accuracy of plaque stenosis category detection results. Furthermore, the method provided in this application, due to the use of sparse convolution feature extraction, reduces computational requirements, thereby reducing the demand on device memory and ultimately lowering hardware costs.
[0028] The following is combined Figures 1 to 7 This application provides a detailed description of the arterial plaque stenosis category detection method, apparatus, electronic device, and computer-readable storage medium mentioned in the embodiments.
[0029] Exemplary method for detecting arterial plaque stenosis categories
[0030] Figure 1 The diagram shown is a flowchart illustrating an arterial plaque stenosis category detection method according to an embodiment of this application. Figure 1 As shown, the method for detecting the type of arterial plaque stenosis includes the following steps.
[0031] S101: Based on head and neck CTA images, vessel category segmentation data of head and neck CTA images, plaque segmentation data, and vessel midline, determine the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery in the head and neck CTA images.
[0032] Head and neck CTA images refer to computed tomography (CTA) images of the head and neck.
[0033] The vascular segmentation data from head and neck CTA images is used to segment different categories of blood vessels using different markers (e.g., different colors). For example, blood vessels in head and neck CTA images are classified into three main categories: carotid arches, carotid arteries, and intracranial arteries, and these three categories are classified using different colors.
[0034] For example, the specific implementation of obtaining the blood vessel category segmentation data of the head and neck CTA image is to input the neck CTA image into a pre-trained blood vessel category segmentation model to obtain the blood vessel category segmentation data of the head and neck CTA image.
[0035] Patchoid segmentation data from head and neck CTA images are used to separate patches from the background. For example, obtaining patch segmentation data from head and neck CTA images involves inputting the head and neck CTA image into a pre-trained patch segmentation model to obtain the patch segmentation data.
[0036] In some embodiments, determining the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery in the head and neck CTA image based on the blood vessel category segmentation data, plaque segmentation data, and blood vessel midline of the head and neck CTA image can be performed as follows: determining a first bounding box corresponding to the carotid artery based on the blood vessel category segmentation data; determining a second bounding box corresponding to the carotid artery based on the first bounding box and a preset margin value; using the second bounding box, cropping the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from the head and neck CTA image, the plaque segmentation data, and the blood vessel midline of the head and neck CTA image; straightening the carotid artery CTA image and carotid artery plaque segmentation data based on the carotid artery midline to obtain a straightened image; resampling the straightened image and carotid artery CTA image based on the location information of the plaque connected components in the carotid artery plaque segmentation data to obtain the plaque region image and plaque straightening region image.
[0037] Specifically, considering the need to determine the plaque stenosis category of the carotid artery, it is necessary to identify the carotid artery region. Based on the vessel category segmentation data from the head and neck CTA image, the minimum bounding box of the carotid artery is calculated and used as the first bounding box for the carotid artery. To avoid the influence of edges on subsequent processing, a margin value is preset to expand the first bounding box; for example, the first bounding box is expanded outward by 10 pixels in all directions to obtain the second bounding box. Using the second bounding box, the carotid artery CTA image, carotid plaque segmentation data, and carotid artery midline are extracted from the head and neck CTA image, the plaque segmentation data from the head and neck CTA image, and the vessel midline.
[0038] After obtaining the carotid CTA image, carotid plaque segmentation data from the carotid CTA image, and the carotid midline, the carotid CTA image and carotid plaque segmentation data are straightened using the carotid midline to obtain a straightened image. In the straightened image, the carotid artery is straightened, and each plaque on the carotid artery is also straightened (the carotid artery can have one or more plaques). The carotid plaque segmentation data from the carotid CTA image not only indicates the number of plaques in the carotid artery but also the specific location of each plaque. There is a correspondence between each plaque in the carotid CTA image and the straightened image. Based on the location information of the connected components of each plaque, the region corresponding to that connected component in the straightened image and the carotid CTA image can be determined. The region corresponding to that connected component in the straightened image and the carotid CTA image is resampled to a preset size to obtain the plaque region image of each plaque and the straightened plaque region image.
[0039] Specifically, the straightening operation mentioned above refers to curved planar reformation (CPR). CPR is often used for vascular analysis because the vascular structure is very tortuous, making it difficult to visually observe the overall state of the blood vessels on CT. CPR can straighten the tortuous blood vessels and display them on the same plane, making it easier to observe the condition of the inner wall of the blood vessels.
[0040] S102: Perform sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image respectively to obtain the patch narrowing rate prediction data.
[0041] Specifically, considering the natural characteristics of plaque growth along blood vessels and its inherent sparseness, sparse convolution feature extraction on plaque region images can effectively utilize this sparsity, thereby improving the robustness of feature extraction. Furthermore, combining this with convolution feature extraction on straightened plaque regions effectively utilizes image information, achieving the goal of improving the accuracy of plaque stenosis rate prediction data.
[0042] It should be noted that step S102 is performed on both the patch region image and the straightened region image for each patch.
[0043] For example, a specific implementation of sparse convolution feature extraction for patch region images can be achieved by using a sparse convolution network (SpConvNet) to perform sparse convolution feature extraction on the patch region images.
[0044] The specific implementation of convolution feature extraction for the patch straightening region image can be achieved by using a convolution network (ConvNet) to perform convolution feature extraction on the patch straightening region image.
[0045] S103: Based on the plaque stenosis rate prediction data, determine the plaque stenosis category detection result corresponding to the plaque.
[0046] Specifically, after obtaining the plaque stenosis rate prediction data, it is necessary to perform post-processing to obtain more accurate plaque stenosis category detection results.
[0047] In this embodiment, by performing sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image respectively, the sparsity characteristics of the patches are effectively utilized, thereby effectively utilizing image information, improving the robustness of feature extraction, and ultimately improving the accuracy of patch narrowing category detection results. Furthermore, the method provided in this embodiment, due to the use of sparse convolution feature extraction operations, can reduce computational power requirements, thereby reducing the requirements for device memory and ultimately reducing hardware costs.
[0048] The following is combined Figure 2 This document details the specific implementation methods for performing sparse convolution feature extraction and convolution feature extraction operations on patch region images and patch straightening region images, respectively, to obtain patch narrowing rate prediction data.
[0049] like Figure 2 As shown, the steps for performing sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image to obtain patch narrowing rate prediction data include the following steps.
[0050] S201: Utilize the sparse convolution feature extraction branch in the patch stenosis rate prediction model to process the patch region image and obtain the first feature vector corresponding to the patch region image.
[0051] For example, the patch region image is input into the sparse convolution feature extraction branch of the patch stenosis rate prediction model, and SpConvNet is used to perform sparse convolution feature extraction on the patch region image to obtain the first feature vector corresponding to the patch region image.
[0052] S202: Using the convolutional feature extraction branch that runs parallel to the sparse convolutional feature extraction branch in the patch stenosis rate prediction model, the patch straightening region image is processed to obtain the second feature vector corresponding to the patch straightening region image.
[0053] For example, the image of the straightened patch region is input into the convolutional feature extraction branch, which runs parallel to the sparse convolutional feature extraction branch, in the patch narrowing rate prediction model. ConvNet is used to perform convolutional feature extraction on the image of the straightened patch region to obtain the second feature vector corresponding to the image of the straightened patch region.
[0054] 0S203: Merge the first and second feature vectors to obtain the plaque stenosis rate prediction data.
[0055] Specifically, merging includes operations such as integration.
[0056] In some embodiments, steps S201 to S203 described above can be performed using a pre-trained plaque stenosis rate prediction model.
[0057] Specifically, the patch narrowing rate prediction model includes parallel sparse convolutional feature extraction branches and convolutional feature extraction branches. The sparse convolutional feature extraction branch uses a SpConvNet model, while the convolutional feature extraction branch uses a ConvNet model. The patch region image is input into the sparse convolutional feature extraction branch for sparse convolutional feature extraction, and the patch straightening region image is input into the convolutional feature extraction branch for convolutional feature extraction, resulting in the first feature vector corresponding to the patch region image and the patch straightening region.
[0058] The second feature vector corresponding to the image is used by the plaque stenosis prediction model to fuse and combine the first and second feature vectors, and then combine them with either the diameter method or the area method (the diameter method or the area method is used to calculate the plaque stenosis).
[0059] Narrowing rate is a common method for obtaining plaque stenosis rate, and it is used to obtain plaque stenosis rate prediction data.
[0060] The embodiments of this application can accurately obtain plaque stenosis rate prediction data by means of a plaque stenosis rate prediction model. Compared with related technologies that require manual identification, the process is not only simpler but also more complex.
[0061] It is less demanding, saves time and effort, and effectively utilizes image information to improve the accuracy of plaque stenosis rate prediction data. Furthermore, it avoids excessive reliance on personal experience, reducing human-induced variability in plaque stenosis rate prediction data and further improving its accuracy.
[0062] For example, the plaque stenosis rate prediction model mentioned above is obtained by training an initial neural network model.
[0063] The following is combined Figure 3a and 3b This application provides a detailed description of the model training methods mentioned in its embodiments. For example... Figure 3aAs shown in the embodiments of this application, the model training method includes the following steps.
[0064] S301: Based on head and neck CTA image samples, blood vessel category segmentation data of head and neck CTA image samples, plaque segmentation data and blood vessel midline, determine the plaque region image samples and plaque straightening region image samples corresponding to the plaques included in the carotid artery in the head and neck CTA image samples, and determine the plaque stenosis rate labeling data corresponding to the plaques.
[0065] Specifically, the plaque stenosis rate marker data corresponding to a plaque is used to characterize the plaque stenosis rate of that plaque. For example, the plaque stenosis rate marker data can be determined using either the diameter method or the area method.
[0066] Based on head and neck CTA image samples, blood vessel category segmentation data of head and neck CTA image samples, plaque segmentation data, and blood vessel midline, the specific implementation method for determining the plaque region image samples and plaque straightening region image samples corresponding to the plaques included in the carotid artery in the head and neck CTA image samples is similar to the above step S101, and will not be repeated here.
[0067] For example, in order to improve the generalization ability of the plaque stenosis rate prediction model, data augmentation operations need to be performed on the neck CTA image samples before step S301. The data augmentation operations include at least one of the following operations: random cropping, random rotation, and random brightness.
[0068] S302: Input the image samples of the patch area and the image samples of the patch straightening area into the initial neural network model to obtain the patch stenosis rate prediction data.
[0069] Specifically, the initial neural network model includes two parallel branches: a sparse convolutional feature extraction branch and a convolutional feature extraction branch. The sparse convolutional feature extraction branch uses the SpConvNet model, while the convolutional feature extraction branch uses the ConvNet model. Image samples from the patch region and the straightened patch region are input into the initial neural network model. The ConvNet model performs sparse convolutional feature extraction on the patch region image samples to obtain the third feature vector corresponding to the patch region image samples. The ConvNet model then performs convolutional feature extraction on the straightened patch region image samples to obtain the fourth feature vector corresponding to the straightened patch region image samples. The initial neural network model fuses and combines the third and fourth feature vectors, and uses either the diameter method or the area method to obtain the patch narrowing rate prediction data.
[0070] In a further embodiment, Figure 3b The diagram shown is a schematic representation of the residual module of the SpConvNet5 model provided in an embodiment of this application. Combined with... Figure 3bAs shown, to prevent or mitigate model overfitting, a dropblock is added to the skip and concat layers of each residual module (res module) in the SpConvNet model. This dropblock uses a drop probability of 0.7 and a kernel size of 7. The convolutional layer randomly drops a kernel-7 block from that feature layer with a probability of 0.7.
[0071] Compared to the original residual block, the residual block with the added dropblock can alleviate the model overfitting problem. Optionally, the dropblock can be selectively used when the sample size is small.
[0072] S303: Based on plaque stenosis rate prediction data and plaque stenosis rate labeling data, determine the loss function value to adjust the parameters of the initial neural network model and obtain the plaque stenosis rate prediction model.
[0073] Specifically, the parameters of the initial neural network model are adjusted using the obtained loss function value until the obtained loss function value meets the preset conditions, thus obtaining the action recognition model.
[0074] 5. Further, in order to ensure the balance between positive and negative samples, the patch region image samples are determined as negative samples, and the patch stenosis rate label data corresponding to the patch is determined as positive samples. The relative entropy loss function (Kullback-Leiblerdivergence, KLDivLoss) is used to train the initial neural network model.
[0075] In this embodiment, two branches, SpConvNet and ConvNet, are used for feature extraction to improve the robustness of feature extraction, thereby optimizing the model towards more accurate predictions and less memory usage.
[0076] Specifically, considering that a plaque is a region, in order to obtain the most accurate plaque stenosis category detection results, post-processing is required on the plaque stenosis rate prediction data corresponding to each plaque. The following section combines...
[0077] Figure 4 This document details the specific implementation method for determining the plaque stenosis category corresponding to a plaque based on plaque stenosis rate prediction data (i.e., the post-processing procedure).
[0078] like Figure 4 As shown, the steps for determining the plaque stenosis category of a plaque based on plaque stenosis rate prediction data include the following steps.
[0079] S401: Perform sliding window sampling on the patch.
[0080] S402: Based on the plaque stenosis rate prediction data, determine the predicted value of the plaque stenosis rate for each window in the sliding window sampling operation.
[0081] For example, for each window, the coordinates of the center point of the window are determined, and the plaque stenosis prediction data located at the center point coordinates is used as the window plaque stenosis prediction value corresponding to the window.
[0082] Specifically, when sliding the window, the predicted plaque stenosis rate data located at the center point coordinates is used as the predicted window plaque stenosis rate value corresponding to the window.
[0083] S403: The largest window plaque stenosis rate prediction value is determined as the plaque stenosis rate prediction value corresponding to the plaque.
[0084] Specifically, considering that a plaque is a region, the predicted stenosis rate of the corresponding window is different during the sliding window process. The largest predicted stenosis rate of the window is selected as the predicted stenosis rate of the corresponding plaque. In other words, the most crowded area of the current plaque is used as the predicted stenosis rate of the corresponding plaque. This helps doctors to understand the worst-case situation of the plaque and to intervene and treat it earlier.
[0085] S404: Determine the plaque stenosis category detection result based on the preset relationship between the predicted plaque stenosis rate and the plaque stenosis category.
[0086] Specifically, the mapping relationship between the predicted plaque stenosis rate and five plaque stenosis categories is pre-set, such as no stenosis (0%), mild stenosis (1-30%), moderate stenosis (31-70%), severe stenosis (71-90%), and occlusion (91-100%). Based on the pre-set relationship between the predicted plaque stenosis rate and the plaque stenosis category, the detection result of the plaque stenosis category is determined.
[0087] In this embodiment of the application, the plaque stenosis rate prediction data corresponding to each plaque is post-processed in the manner described above, thereby achieving the goal of obtaining the most accurate plaque stenosis category detection result.
[0088] The above text combined Figures 2 to 4 The present disclosure describes in detail the method embodiments, which are then combined with the following. Figure 5 and Figure 6 The present disclosure provides a detailed description of the apparatus embodiments. Furthermore, it should be understood that the descriptions of the method embodiments correspond to the descriptions of the apparatus embodiments; therefore, any parts not described in detail can be found in the foregoing method embodiments.
[0089] Exemplary arterial plaque stenosis category detection device
[0090] Figure 5The diagram shown is a structural schematic of a carotid artery plaque stenosis category detection device provided in an embodiment of this disclosure. Figure 5 As shown, the carotid plaque stenosis category detection device 500 provided in this embodiment includes a first determination module 510, a feature extraction module 520, and a second determination module 530.
[0091] In this embodiment of the disclosure, the first determining module 510 is configured to determine, based on the head and neck CTA image, the vessel category segmentation data of the head and neck CTA image, plaque segmentation data, and the vessel midline, the plaque region image and the straightened plaque region image corresponding to the plaque in the carotid artery of the head and neck CTA image. The feature extraction module 520 is configured to perform sparse convolution feature extraction and convolution feature extraction operations on the plaque region image and the straightened plaque region image, respectively, to obtain plaque stenosis rate prediction data. The second determining module 530 is configured to determine the plaque stenosis category detection result corresponding to the plaque based on the plaque stenosis rate prediction data.
[0092] In this embodiment, by performing sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image respectively, the sparsity characteristics of the patches are effectively utilized, thereby effectively utilizing image information, improving the robustness of feature extraction, and ultimately improving the accuracy of patch narrowing category detection results. Furthermore, the method provided in this embodiment, due to the use of sparse convolution feature extraction operations, can reduce computational power requirements, thereby reducing the requirements for device memory and ultimately reducing hardware costs.
[0093] In some embodiments, the feature extraction module 520 is further configured to perform sparse convolutional feature extraction and convolutional feature extraction operations on the patch region image and the patch straightening region image, respectively, to obtain patch narrowing rate prediction data, including: processing the patch region image using the sparse convolutional feature extraction branch in the patch narrowing rate prediction model to obtain a first feature vector corresponding to the patch region image; processing the patch straightening region image using the convolutional feature extraction branch in the patch narrowing rate prediction model that runs parallel to the sparse convolutional feature extraction branch to obtain a second feature vector corresponding to the patch straightening region image; and merging the first feature vector and the second feature vector to obtain the patch narrowing rate prediction data.
[0094] Figure 6 The diagram shown is a structural schematic of a first determining module provided in an embodiment of this disclosure. Figure 5 As shown, the first determining module 510 further includes a first determining unit 610 and a second determining unit 620.
[0095] In one embodiment, the first determining unit 610 is configured to determine the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from the head and neck CTA image, plaque segmentation data of the head and neck CTA image, and the midline of the blood vessel based on the blood vessel category segmentation data; the second determining unit 620 is configured to determine the plaque region image and plaque straightening region image corresponding to the plaque included in the carotid artery based on the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline.
[0096] In one embodiment, the first determining unit 610 is further configured to: determine a first bounding box corresponding to the carotid artery based on blood vessel category segmentation data; determine a second bounding box corresponding to the carotid artery based on the first bounding box and a preset margin value; and use the second bounding box to extract the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the blood vessel midline.
[0097] In one embodiment, the second determining unit 620 is further configured to: perform a straightening operation on the carotid CTA image and carotid plaque segmentation data based on the carotid artery midline to obtain a straightened image; and perform a resampling operation on the straightened image and the carotid CTA image based on the location information of the plaque connected regions in the carotid plaque segmentation data to obtain a plaque region image and a plaque straightened region image.
[0098] In one embodiment, the second determining module 530 is further configured to perform a sliding window sampling operation on the plaque; determine the predicted value of the plaque stenosis rate corresponding to each window in the sliding window sampling operation based on the plaque stenosis rate prediction data; determine the largest predicted value of the plaque stenosis rate as the predicted value of the plaque stenosis rate corresponding to the plaque; and determine the plaque stenosis category detection result based on the preset relationship between the predicted value of the plaque stenosis rate and the plaque stenosis category.
[0099] In one embodiment, the second determining module 530 is further configured to determine the center point coordinates of each window; and use the plaque stenosis prediction data located at the center point coordinates as the window plaque stenosis prediction value corresponding to the window.
[0100] Exemplary electronic devices and computer-readable storage media
[0101] Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this disclosure. Figure 7 The electronic device 700 shown (which may specifically be a computer device) includes a memory 701, a processor 702, a communication interface 703, and a bus 704. The memory 701, processor 702, and communication interface 703 are interconnected via the bus 704.
[0102] The memory 701 may be a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 701 may store a program, and when the program stored in the memory 701 is executed by the processor 702, the processor 702 and the communication interface 703 are used to execute the various steps of the carotid plaque stenosis category detection method of the present disclosure embodiments.
[0103] The processor 702 may be a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), graphics processing unit (GPU), or one or more integrated circuits, for executing relevant programs to achieve the functions required by the units in the carotid plaque stenosis category detection method apparatus of this disclosure.
[0104] The processor 702 can also be an integrated circuit chip with signal processing capabilities. In implementation, each step of the carotid artery plaque stenosis category detection method of this disclosure can be completed by the integrated logic circuitry in the hardware of the processor 702 or by instructions in software form. The processor 702 can also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this disclosure. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this disclosure can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory 701. The processor 702 reads the information in the memory 701 and, in conjunction with its hardware, performs the functions required by the units included in the carotid plaque stenosis category detection method apparatus of this disclosure, or executes the carotid plaque stenosis category detection method of this disclosure.
[0105] The communication interface 703 uses transceiver devices, such as, but not limited to, transceivers, to enable communication between the electronic device 700 and other devices or communication networks.
[0106] Bus 704 may include a pathway for transmitting information between various components of electronic device 700 (e.g., memory 701, processor 702, communication interface 703).
[0107] It should be noted that, although Figure 7 The illustrated electronic device 700 only shows the memory, processor, and communication interface. However, those skilled in the art should understand that in specific implementations, the electronic device 700 may also include other devices necessary for normal operation. Furthermore, depending on specific needs, those skilled in the art should understand that the electronic device 700 may also include hardware devices for implementing other additional functions. Moreover, those skilled in the art should understand that the electronic device 700 may only include the devices necessary for implementing the embodiments of this disclosure, and may not necessarily include... Figure 7 All the devices shown.
[0108] In addition to the methods, apparatus, and devices described above, embodiments of this disclosure may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the various steps of the carotid plaque stenosis category detection method provided in the various embodiments of this disclosure.
[0109] The computer program product can be written in any combination of one or more programming languages to perform the operations of the embodiments of this disclosure. The programming languages include object-oriented programming languages such as Java and C++, as well as conventional procedural programming languages such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.
[0110] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform various steps of the carotid plaque stenosis category detection method provided in the various embodiments of this disclosure.
[0111] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.
[0112] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.
[0113] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0114] In the several embodiments provided in this disclosure, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.
[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0116] In addition, the functional units in the various embodiments of this disclosure can be integrated into a similar region segmentation unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0117] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.
[0118] The above description is merely a specific embodiment of this disclosure, but the scope of protection of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this disclosure should be included within the scope of protection of this disclosure. Therefore, the scope of protection of this disclosure should be determined by the scope of the claims.
Claims
1. A method for detecting carotid artery plaque stenosis category, characterized in that, include: Based on head and neck CTA images, vessel category segmentation data, plaque segmentation data, and vessel midline, the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery in the head and neck CTA images are determined. Sparse convolution feature extraction and convolution feature extraction operations are performed on the patch region image and the patch straightening region image, respectively, to obtain a first feature vector corresponding to the patch region image and a second feature vector corresponding to the patch straightening region image. Based on the first feature vector and the second feature vector, patch stenosis rate prediction data is obtained. Based on the plaque stenosis rate prediction data, the detection result of the plaque stenosis category corresponding to the plaque is determined.
2. The method according to claim 1, characterized in that, The step of performing sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image respectively to obtain patch narrowing rate prediction data includes: The patch region image is processed using the sparse convolution feature extraction branch in the patch stenosis rate prediction model to obtain the first feature vector corresponding to the patch region image. The convolutional feature extraction branch, which runs parallel to the sparse convolutional feature extraction branch in the patch stenosis rate prediction model, is used to process the patch straightening region image to obtain the second feature vector corresponding to the patch straightening region image. The first feature vector and the second feature vector are merged, and the plaque stenosis rate prediction data is obtained based on the merged feature vector.
3. The method according to claim 1 or 2, characterized in that, The determination of the plaque region image and plaque straightening region image corresponding to the plaques in the carotid artery of the head and neck CTA image based on the head and neck CTA image, the vessel category segmentation data of the head and neck CTA image, plaque segmentation data, and vessel midline includes: Based on the vessel category segmentation data, the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline are determined from the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the vessel midline. Based on the carotid CTA image, the carotid plaque segmentation data, and the carotid midline, the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery are determined.
4. The method according to claim 3, characterized in that, The step of determining the carotid artery CTA image, carotid artery plaque segmentation data, and carotid artery midline from the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the midline of the blood vessels based on the blood vessel category segmentation data includes: Based on the segmented data of the blood vessel categories, the first bounding box corresponding to the carotid artery is determined; Based on the first bounding box and the preset margin value, the second bounding box corresponding to the carotid artery is determined; Using the second bounding box, the carotid artery CTA image, the carotid artery plaque segmentation data, and the carotid artery midline are extracted from the head and neck CTA image, the plaque segmentation data of the head and neck CTA image, and the midline of the blood vessel.
5. The method according to claim 3, characterized in that, The step of determining the plaque region image and plaque straightening region image corresponding to the plaques included in the carotid artery based on the carotid CTA image, the carotid plaque segmentation data, and the carotid artery midline includes: Based on the midline of the carotid artery, the carotid CTA image and the carotid plaque segmentation data are straightened to obtain a straightened image. Based on the location information of the connected regions of the plaque in the carotid plaque segmentation data, the regions corresponding to the connected regions in the straightened image and the carotid CTA image are resampled to obtain the plaque region image and the straightened plaque region image.
6. The method according to claim 1 or 2, characterized in that, The step of determining the plaque stenosis category detection result corresponding to the plaque based on the plaque stenosis rate prediction data includes: Perform a sliding window sampling operation on the patch; Based on the plaque stenosis rate prediction data, determine the window plaque stenosis rate prediction value corresponding to each window in the sliding window sampling operation; The largest predicted window plaque stenosis rate is determined as the predicted plaque stenosis rate for the plaque. Based on the preset relationship between the predicted plaque stenosis rate and the plaque stenosis category, the detection result of the plaque stenosis category is determined.
7. The method according to claim 6, wherein determining the predicted window plaque stenosis rate value corresponding to each window in the sliding window sampling operation based on the plaque stenosis rate prediction data includes: For each window, determine the coordinates of the center point of the window; The predicted plaque stenosis rate data located at the coordinates of the center point is used as the predicted window plaque stenosis rate value corresponding to the window.
8. A device for detecting carotid artery plaque stenosis category, characterized in that, include: The first determining module is configured to determine, based on the head and neck CTA image, the vessel category segmentation data of the head and neck CTA image, the plaque segmentation data, and the vessel midline, the plaque region image and the plaque straightening region image corresponding to the plaque in the carotid artery of the head and neck CTA image. The feature extraction module is configured to perform sparse convolution feature extraction and convolution feature extraction operations on the patch region image and the patch straightening region image respectively, to obtain a first feature vector corresponding to the patch region image and a second feature vector corresponding to the patch straightening region image, and to obtain patch stenosis rate prediction data based on the first feature vector and the second feature vector. The second determining module is configured to determine the plaque stenosis category detection result corresponding to the plaque based on the plaque stenosis rate prediction data.
9. An electronic device, characterized in that, include: processor; as well as A memory storing computer program instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1 to 7.