Methods and devices for determining the severity of portal vein thrombosis

Abstract

The application discloses a portal vein system thrombus severity determination method and device, and relates to the technical field of image processing. The method comprises the following steps: obtaining a blood vessel image and a thrombus image of a hepatic portal vein and a blood vessel skeleton line from a to-be-processed CTA image; inputting the skeleton line and the blood vessel image into a pre-trained bifurcation region segmentation model to segment a bifurcation region image from the blood vessel image; determining, according to the bifurcation region image, the blood vessel image and the thrombus image, to which thrombus the thrombus pixels contained in the bifurcation region image belong, separating the blood vessel branches in the image obtained by merging the blood vessel image and the thrombus image, determining the blood vessel branches with the same thrombus pixels in all the blood vessel branches, merging the blood vessel branches containing the same thrombus and the bifurcation region image, and determining the thrombus severity based on the merged image. The application can improve the accuracy of portal vein system thrombus severity evaluation.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and apparatus for determining the severity of portal vein thrombosis. Background Technology

[0002] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section.

[0003] Assessing the severity of portal venous thrombosis is of paramount importance. Current techniques primarily rely on imaging examinations such as ultrasound, CT (Computed Tomography), and MRI (Magnetic Resonance Imaging) to acquire images of the portal venous system, followed by manual analysis of the extent of thrombus involvement and the degree of vascular obstruction. This process has limitations, including subjectivity and low efficiency. Furthermore, the portal vein trunk exhibits a Y-shaped bifurcation geometry; thrombi located at this bifurcation point increase the difficulty of manual analysis and affect accuracy. Current research on thrombus severity assessment in bifurcation areas remains scarce, and the accuracy of portal venous thrombosis severity assessment needs improvement. Summary of the Invention

[0004] This invention provides a method for determining the severity of portal vein system thrombosis, thereby improving the accuracy of assessment of the severity of portal vein system thrombosis. The method includes:

[0005] Receive CTA (Computed Tomography Angiography) images to be processed;

[0006] The portal vein vascular image and thrombus image are obtained from the CTA image to be processed by image segmentation; the thrombus image includes one or more thrombus blocks, and the thrombus blocks are marked with thrombus block identifiers;

[0007] The vascular skeleton line is extracted from the vascular image;

[0008] The skeleton line and blood vessel image are input into a pre-trained bifurcation region segmentation model to segment the blood vessel image to obtain the first bifurcation region image; the bifurcation region segmentation model is trained on a 3D Attention Unet network using historical CTA images; the first bifurcation region image is a binary image.

[0009] Based on the position of the first bifurcation region image in the blood vessel image, the pixel values ​​of the merged image of the blood vessel image and the thrombus image are mapped onto the first bifurcation region image to obtain the second bifurcation region image with restored pixels, and it is determined whether the second bifurcation region image is mapped with pixels of the thrombus image.

[0010] Once it is determined that the second bifurcation region image maps to pixels of the thrombus image, the pixel coordinates of the thrombus image in the second bifurcation region image are extracted.

[0011] Based on the extracted thrombus image pixel coordinates of the second bifurcation region image, the thrombus block identifier within the second bifurcation region image is determined.

[0012] By setting the second bifurcation region image pixels as background on the merged image of the vascular image and thrombus image, multiple first vascular branches are separated from the merged image of the vascular image and thrombus image.

[0013] On each first blood vessel branch, detect whether it includes a pixel corresponding to the thrombus identifier, and determine the second blood vessel branch that includes the pixel corresponding to the thrombus identifier;

[0014] The images of the second blood vessel branch and the second bifurcation region are merged to obtain a merged image;

[0015] The severity of portal vein thrombosis is determined based on the proportion of thrombus pixels in the cross-section of blood vessels in the merged image.

[0016] This invention also provides a device for determining the severity of portal vein system thrombosis, to improve the accuracy of assessing the severity of portal vein system thrombosis. The device includes:

[0017] The data receiving module is used to receive CTA images to be processed;

[0018] The image preprocessing module is used to segment the CTA image to obtain the portal vein vascular image and thrombus image; and to extract the vascular skeleton line from the vascular image.

[0019] The bifurcation region segmentation module is used to input skeleton lines and blood vessel images into a pre-trained bifurcation region segmentation model to segment the blood vessel images to obtain a first bifurcation region image; the bifurcation region segmentation model is trained on a 3D Attention Unet network using historical CTA images; the first bifurcation region image is a binary image.

[0020] The bifurcation region thrombosis detection module is used to map the pixel values ​​of the merged image of the blood vessel image and the thrombosis image onto the first bifurcation region image according to the position of the first bifurcation region image in the blood vessel image, to obtain the second bifurcation region image with restored pixels, and to determine whether the second bifurcation region image maps to pixels of the thrombosis image; when it is determined that the second bifurcation region image maps to pixels of the thrombosis image, the thrombosis image pixel coordinates of the second bifurcation region image are extracted; based on the extracted thrombosis image pixel coordinates of the second bifurcation region image, the thrombosis block identifier where the thrombosis is located in the second bifurcation region image is determined.

[0021] The vascular branch separation module is used to separate multiple first vascular branches from the merged image of the vascular image and the thrombus image by setting the second bifurcation region image pixels as the background on the merged image of the vascular image and the thrombus image;

[0022] The vessel branch removal module is used to detect whether the pixel corresponding to the thrombus identifier is included on each first vessel branch, determine the second vessel branch that includes the pixel corresponding to the thrombus identifier; and merge the images of the second vessel branch and the second bifurcation region to obtain a merged image.

[0023] The thrombosis grading calculation module is used to determine the severity of portal vein system thrombosis based on the proportion of thrombus pixels in the cross-section of blood vessels in the merged image.

[0024] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for determining the severity of portal vein thrombosis.

[0025] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for determining the severity of portal vein system thrombosis.

[0026] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for determining the severity of portal vein system thrombosis.

[0027] In this embodiment of the invention, portal vein vessel images and thrombus images are obtained from image segmentation of the CTA image to be processed. Then, the skeleton lines and vessel images are input into a pre-trained bifurcation region segmentation model to segment the first bifurcation region image from the vessel image. Based on the positional relationship between the first bifurcation region image and the vessel and thrombus images, it is determined whether the first bifurcation region image contains thrombus pixels. If the first bifurcation region image contains thrombus pixels, it is determined which thrombus block the thrombus pixel belongs to, and then it is determined which vascular branch has the thrombus block. Finally, vascular branches and bifurcation region images containing the same thrombus pixels or the same thrombus block are merged. Based on the proportion of thrombus pixels in the vascular cross-section of the merged image, the severity of portal vein thrombosis is determined. This embodiment of the invention achieves automated detection and precise removal of bifurcation branches in portal vein vessels based on image processing and deep learning, ultimately ensuring the structural integrity of the target vascular branch containing thrombus pixels and the triangular bifurcation region. By measuring the proportion of thrombus pixels in the vascular cross-section, the accuracy of portal vein thrombosis severity assessment is improved. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0029] Figure 1 This is a flowchart illustrating the method for determining the severity of portal vein system thrombosis in an embodiment of the present invention.

[0030] Figure 2 This is a specific embodiment of the method for determining the severity of portal vein system thrombosis in this invention.

[0031] Figure 3 This is a schematic diagram of thrombus detection in the bifurcation region in an embodiment of the present invention;

[0032] Figure 4 This is a schematic diagram of blood vessel branch separation in an embodiment of the present invention;

[0033] Figure 5 This is a schematic diagram of image merging in an embodiment of the present invention;

[0034] Figure 6 This is a schematic diagram of the device for determining the severity of portal vein thrombosis in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0036] To facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order.

[0037] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0038] Assessing the severity of portal venous thrombosis is of paramount importance. However, the portal venous system has a complex tree-like bifurcation structure, and existing methods for calculating thrombosis severity face technical challenges when thrombosis involves the bifurcation areas: the vascular geometry at the bifurcation points exhibits multi-branching fusion characteristics, and the measurement of thrombus morphology in these areas lacks clear scientific basis, leading to deviations in thrombus burden quantification results. This problem is particularly prominent in the main portal vein region, where the bifurcation points exhibit a unique "Y-shaped three-way intersection" geometry, further reducing the accuracy of portal venous system thrombosis severity assessment.

[0039] To address the aforementioned issues, this invention proposes a method for automatic removal of portal vein bifurcation vessels and determination of thrombus severity based on deep learning and image processing. This method achieves automated detection and precise removal of bifurcation branches in portal vein bifurcation areas, and determines the severity of thrombi. It solves the problem of calculating thrombus grading at vessel bifurcation points, provides a standardized preprocessing workflow for thrombus burden quantification analysis based on vessel cross-sections, and significantly improves the accuracy and efficiency of thrombus grading.

[0040] Figure 1 This is a flowchart illustrating the method for determining the severity of portal vein system thrombosis in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0041] Step 101: Receive the CTA image to be processed;

[0042] Step 102: Obtain the portal vein vessel image and thrombus image from the CTA image to be processed by image segmentation; the thrombus image includes one or more thrombus blocks, and the thrombus blocks are marked with thrombus block identifiers;

[0043] Step 103: Extract the vascular skeleton line from the vascular image;

[0044] Step 104: Input the skeleton line and blood vessel image into the pre-trained bifurcation region segmentation model to segment the first bifurcation region image from the blood vessel image; the bifurcation region segmentation model is trained on the 3DA attention Unet network using historical CTA images; the first bifurcation region image is a binary image.

[0045] Step 105: Based on the position of the first bifurcation region image in the blood vessel image, map the pixel values ​​of the merged image of the blood vessel image and the thrombus image onto the first bifurcation region image to obtain the second bifurcation region image with restored pixels, and determine whether the second bifurcation region image maps to the pixels of the thrombus image.

[0046] Step 106: Once it is determined that the second bifurcation region image maps to pixels of the thrombus image, extract the pixel coordinates of the thrombus image in the second bifurcation region image.

[0047] Step 107: Based on the thrombus image pixel coordinates of the extracted second bifurcation region image, determine the thrombus block identifier where the thrombus is located in the second bifurcation region image;

[0048] Step 108: By setting the second bifurcation region image pixels as the background on the merged image of the blood vessel image and the thrombus image, multiple first blood vessel branches are separated from the merged image of the blood vessel image and the thrombus image.

[0049] Step 109: Detect whether each first blood vessel branch includes a pixel corresponding to the thrombus identifier, and determine the second blood vessel branch that includes the pixel corresponding to the thrombus identifier;

[0050] Step 1010: Merge the images of the second blood vessel branch and the second bifurcation region to obtain a merged image;

[0051] Step 1011: Determine the severity of portal vein thrombosis based on the proportion of thrombus pixels in the cross-section of blood vessels in the merged image.

[0052] The following describes a method for determining the severity of portal vein system thrombosis in an embodiment of the present invention.

[0053] In this embodiment, a bifurcation region segmentation model is pre-trained. The bifurcation region segmentation model is trained on a 3D Attention Unet network using historical CTA images.

[0054] During implementation, venous phase CTA image data is obtained based on CT scans. Then, the anatomical structure of the portal vein and thrombus, i.e., labeled images, is obtained using methods such as deep learning, image processing, and manual annotation. Labeled images of the triangular regions at vessel bifurcation points are obtained based on machine annotation, manual review, or manual annotation. Sample data is formed based on the labeled images of vessels, thrombi, and triangular regions at vessel bifurcation points.

[0055] When processing the blood vessel bifurcation region segmentation model, the sample data is first preprocessed. Based on the contour range of the blood vessel, the background pixels of the blood vessel image are cropped to reduce background noise and memory usage. The same cropping method is used to crop the triangular region of the blood vessel bifurcation, ensuring that the cropped blood vessel image and the triangular region label image are aligned.

[0056] Next, skeleton lines are extracted from the blood vessel images. The skeleton lines of the blood vessels are extracted using the skeletonize function in the skimage library (an open-source library in Python specifically for image processing). These skeleton lines can intuitively and effectively represent the central path and overall structural direction of the blood vessels, containing key structural information of the blood vessels. The location of the bifurcation point of the blood vessel skeleton line is used as the center of the triangular region at the bifurcation point to guide the model's ability to locate the bifurcation region of the blood vessels.

[0057] In one embodiment, the bifurcation region segmentation model includes an encoder, a decoder, and an attention module; wherein the encoder includes a 3D convolutional layer and a downsampling layer, the decoder includes an upsampling layer and a 3D deconvolutional layer, and the attention module is embedded between the layers of the encoder and the decoder, and the attention module assigns higher weights to pixels associated with the bifurcation region than to pixels not associated with the bifurcation region.

[0058] In this embodiment of the invention, the input to the bifurcation region segmentation model is to merge the vascular skeleton line as prior information with the vascular image in the channel dimension to assist the model in segmenting the vascular bifurcation region.

[0059] In this embodiment, the segmentation of the triangular region of blood vessel bifurcation uses a 3D Attention UNet network, whose overall structure consists of an encoder, a decoder, and an attention module. The encoder comprises multiple 3D convolutional layers and downsampling layers, progressively reducing the spatial resolution of the feature map while increasing the number of channels to extract deep semantic features from the blood vessel image. The decoder progressively restores the spatial resolution of the feature map through upsampling layers and 3D deconvolutional layers, while simultaneously making skip connections with the corresponding layers of the feature map in the encoder, fusing shallow detail features with deep semantic features to ensure the accuracy and completeness of the segmentation results. An attention module is embedded between the layers of the encoder and decoder. This module calculates the attention weights for each region in the feature map, automatically focusing on key features related to the blood vessel bifurcation region and suppressing irrelevant background and noise features, thereby further improving the model's ability to identify and segment bifurcation regions.

[0060] The loss function L, which combines Dice loss and cross-entropy loss, is shown in the following formula:

[0061] ;

[0062] In the formula, The Dice loss function, Let cross-entropy be the loss function. and Here, is the weighting coefficient, and i is the index of the i-th pixel. Let be the probability predicted by the model that the i-th pixel belongs to the bifurcation region. Let N be the true label of the pixel, K be the total number of pixels, and c be the index of the c-th category. Let be the probability predicted by the model that the i-th pixel belongs to the c-th class. Let be the true label of the i-th pixel belonging to class c. By using a joint loss function, the model's accuracy in recognizing small targets is improved. At the same time, it ensures that the model learns the global probability distribution and forces the model to focus on the accuracy of overlapping regions, thus avoiding the one-sidedness of a single loss function.

[0063] After training the bifurcation region segmentation model, it is put into use.

[0064] Figure 2 This is a specific embodiment of the method for determining the severity of portal vein system thrombosis in this invention, as shown in the diagram. Figure 2 As shown, the method includes:

[0065] 1) Beginning:

[0066] 2) Construct labeled images of blood vessels, thrombi, and triangular regions of blood vessel bifurcation;

[0067] 3) Detect blood vessel bifurcation regions based on the 3D Attention Unet network;

[0068] 4) Detect whether the bifurcation area includes thrombus pixels;

[0069] 5) If not, end. If yes, proceed with vascular branch dissection;

[0070] 6) Remove vascular branches.

[0071] Figure 2 The diagram illustrates the automated removal process for portal vein bifurcation vessels. After removing branches that do not contain thrombi in the bifurcation area, the remaining branches and thrombi in the bifurcation area are graded and treated.

[0072] In step 101, the CTA image to be processed is received. In step 102, the portal vein vascular image and thrombus image are obtained from the image segmentation of the CTA image to be processed.

[0073] In one embodiment, obtaining the portal vein vascular image and thrombus image from the CTA image to be processed can include: obtaining the portal vein vascular image and thrombus image from the CTA image to be processed by image processing or convolutional neural network.

[0074] During implementation, blood vessels and thrombi are pre-identified. Based on the contour range of the blood vessels, the background pixels of the blood vessel and thrombus images are cropped to obtain the blood vessel and thrombus images of the portal vein.

[0075] During implementation, to facilitate subsequent image matching, merging, and other processing, the cropped vascular image and thrombus image are placed in the same three-dimensional coordinate system.

[0076] By analyzing connected components, one or more thrombus blocks in a thrombus image are labeled, and each thrombus block has a unique thrombus block identifier.

[0077] Next, in step 103, the skeleton lines of the blood vessels are extracted using the skeletonize function based on the blood vessel image.

[0078] In step 104, the extracted vascular skeleton line is used as prior information, and the vascular image is input into a pre-trained bifurcation region segmentation model to segment the first bifurcation region image from the vascular image. At this time, the first bifurcation region image is a binary image.

[0079] In step 105, according to the position of the first bifurcation region image in the blood vessel image, the pixel values ​​of the merged image of the blood vessel image and the thrombus image are mapped onto the first bifurcation region image to obtain the second bifurcation region image with restored pixels, and it is determined whether the second bifurcation region image is mapped with pixels of the thrombus image.

[0080] In step 106, once it is determined that the second bifurcation region image maps to pixels of the thrombus image, the thrombus image pixel coordinates of the second bifurcation region image are extracted.

[0081] In step 107, the thrombus block identifier within the second bifurcation region image is determined based on the thrombus image pixel coordinates extracted from the second bifurcation region image.

[0082] Specifically, the identification of thrombus involves performing connected component detection on the thrombus image. After obtaining the connected component detection results, the coordinates are used to determine which thrombus it belongs to. Then, the complete pixels of the thrombus are obtained for subsequent detection of whether the thrombus exists in the blood vessel branch.

[0083] Figure 3 This is a schematic diagram of thrombus detection in the bifurcation region in an embodiment of the present invention, as shown below. Figure 3 As shown, the upper left half is the merged image of the blood vessel and the thrombus (e.g., the blood vessel pixel value is 1, the thrombus pixel value is 2, and the background pixel value is 0, clearly distinguishing the blood vessel, thrombus, and background). The lower left half is the image of the first bifurcation region segmented from the blood vessel image. These two images are multiplied to extract the blood vessel and thrombus pixels in the triangular bifurcation region, achieving pixel restoration of the binary bifurcation region image. Next, each triangular bifurcation region is traversed, detecting whether the region contains thrombus pixels, and determining which thrombus block the thrombus pixels in the bifurcation region belong to, for example, determining the thrombus block identifier to which the thrombus pixels in the second bifurcation region belong.

[0084] When a thrombus is detected in the triangular bifurcation region, in step 108, the second bifurcation region image pixels are set as background on the image after merging the vascular image and the thrombus image, and multiple first vascular branches are separated from the image after merging the vascular image and the thrombus image.

[0085] In one embodiment, by setting the second bifurcation region image pixels as background on the merged image of the vascular image and the thrombus image, multiple first vascular branches are separated from the merged image of the vascular image and the thrombus image, which may include:

[0086] A region masking operation is performed on the merged image of the blood vessel image and the thrombus image to set the pixel value of the second bifurcation region image to the background pixel value, thus obtaining the processed blood vessel image.

[0087] Perform connected component analysis on the processed vascular image to calculate the number of connected components in the current processed vascular image;

[0088] If the number of connected components is less than 3, the following operations are performed on the currently processed vascular image: based on the contour of the second bifurcation region, morphological dilation is performed, the pixel values ​​of the second bifurcation region image are set to the background pixel values, and the number of connected components is analyzed, until 3 independent vascular connected components are formed in the currently processed vascular image, that is, 3 independent first vascular branches.

[0089] Figure 4 This is a schematic diagram of vascular branch separation in an embodiment of the present invention, as shown below. Figure 4 As shown, to ensure that the Y-shaped bifurcation region can be separated into three vascular branches, a region masking operation is performed on the merged image of the vascular image and the thrombus image. The pixel values ​​of the triangular bifurcation region are set to background pixel values. Subsequently, connected component analysis is performed on the processed merged image of the vascular image and the thrombus image to calculate the number of connected components in the current image. If the number of connected components is less than 3, a morphological dilation operation is performed: based on the contour of the triangular bifurcation region, a dilation operation with a radius of 1 pixel is performed each time. After that, the pixel values ​​of the triangular bifurcation region are set to background pixel values, and connected component analysis is immediately repeated until three independent vascular connected components are formed in the image.

[0090] In step 109, it is detected whether each first blood vessel branch includes a pixel corresponding to the thrombus identifier, and a second blood vessel branch including the pixel corresponding to the thrombus identifier is determined.

[0091] This step determines which specific blood vessel branch contains the same thrombus pixel as the bifurcation region.

[0092] Step 1010: Merge the images of the second blood vessel branch and the second bifurcation region to obtain a merged image.

[0093] right Figure 4 For each separated blood vessel connected region, pixel inclusion detection is performed: traversing all pixels within each blood vessel connected region, it is determined whether it contains pixels corresponding to the thrombus block identifier. Blood vessel connected regions that do not contain pixels corresponding to the thrombus block identifier are backgrounded; that is, through a pixel-by-pixel assignment operation, the pixel values ​​of such connected regions are uniformly set to background values, while only preserving the structural integrity of the target blood vessel branches and triangular bifurcation regions containing pixels corresponding to the thrombus block identifier. Figure 5 This is a schematic diagram of image merging in an embodiment of the present invention, such as... Figure 5 As shown, the final image, after removing other branches, is a merged image containing thrombi and bifurcation regions.

[0094] In step 1011, the severity of portal vein thrombosis is determined based on the proportion of thrombus pixels in the cross-section of blood vessels in the merged image.

[0095] In one embodiment, determining the severity of portal vein system thrombosis based on the proportion of thrombus pixels in the cross-section of blood vessels in the merged image may include:

[0096] Obtain multiple blood vessel cross sections from the merged image;

[0097] Calculate the percentage of thrombus pixels in each blood vessel cross section, and determine the severity of thrombosis in the portal vein system based on the percentage of the largest thrombus pixels.

[0098] During implementation, if the thrombus in the triangular bifurcation region spans three branches, then one branch is removed in turn, and the thrombus grading is calculated in the merged image after connecting the two branches and the triangular region; if the thrombus in the triangular bifurcation region only involves one branch, then one redundant branch is arbitrarily removed.

[0099] After removing excess vascular branches, the vascular region where the thrombus is located is approximately cylindrical in shape. Multiple vascular cross-sections of this region are obtained, and the proportion of thrombus pixels and vascular pixels in the cross-sections is calculated. When the maximum proportion is greater than 90%, the severity level of the thrombus is determined to be level 3. When the maximum proportion is between 50% and 90%, it is level 2. When the maximum proportion is greater than 0 and less than 50%, it is level 1.

[0100] In summary, the embodiments of the present invention can achieve automatic removal of portal vein bifurcation vessels and regional thrombosis classification, thereby improving the accuracy and efficiency of thrombosis classification calculation.

[0101] This invention also provides a device for determining the severity of portal vein system thrombosis, as described in the following embodiments. Since the principle by which this device solves the problem is similar to the method for determining the severity of portal vein system thrombosis, the implementation of this device can refer to the implementation of the method for determining the severity of portal vein system thrombosis; repeated details will not be elaborated further.

[0102] Figure 6 This is a schematic diagram of the portal vein system thrombosis severity determination device in an embodiment of the present invention, as shown below. Figure 6 As shown, the device 600 includes:

[0103] Data receiving module 601 is used to receive CTA images to be processed;

[0104] Image preprocessing module 602 is used to segment the image of the portal vein and the thrombus image from the CTA image to be processed; extract the vascular skeleton line from the vascular image; the thrombus image includes one or more thrombus blocks, and the thrombus blocks are marked with thrombus block identifiers;

[0105] The bifurcation region segmentation module 603 is used to input skeleton lines and blood vessel images into a pre-trained bifurcation region segmentation model to segment a first bifurcation region image from the blood vessel image; the bifurcation region segmentation model is trained on a 3D Attention Unet network using historical CTA images; the first bifurcation region image is a binary image.

[0106] The bifurcation region thrombosis detection module 604 is used to map the pixel values ​​of the merged image of the blood vessel image and the thrombosis image onto the first bifurcation region image according to the position of the first bifurcation region image in the blood vessel image, to obtain the second bifurcation region image with restored pixels, and to determine whether the second bifurcation region image maps to pixels of the thrombosis image; when it is determined that the second bifurcation region image maps to pixels of the thrombosis image, the thrombosis image pixel coordinates of the second bifurcation region image are extracted; based on the extracted thrombosis image pixel coordinates of the second bifurcation region image, the thrombosis block identifier where the thrombosis is located in the second bifurcation region image is determined.

[0107] The vascular branch separation module 605 is used to separate multiple first vascular branches from the image after merging the vascular image and the thrombus image by setting the second bifurcation region image pixels as the background on the image after merging the vascular image and the thrombus image;

[0108] The vessel branch removal module 606 is used to detect whether the first vessel branch includes a pixel corresponding to the thrombus identifier, determine the second vessel branch including the pixel corresponding to the thrombus identifier, and merge the images of the second vessel branch and the second bifurcation region to obtain a merged image.

[0109] The thrombosis grading calculation module 607 is used to determine the severity of portal vein system thrombosis based on the proportion of thrombosis pixels in the cross-section of blood vessels in the merged image.

[0110] In one embodiment, the image preprocessing module 602 is specifically used for:

[0111] Image processing methods or convolutional neural networks are used to segment images of the portal vein and thrombus from the CTA images to be processed.

[0112] In one embodiment, the bifurcation region segmentation model includes an encoder, a decoder, and an attention module; wherein the encoder includes a 3D convolutional layer and a downsampling layer, the decoder includes an upsampling layer and a 3D deconvolutional layer, and the attention module is embedded between the layers of the encoder and the decoder, and the attention module assigns higher weights to pixels associated with the bifurcation region than to pixels not associated with the bifurcation region.

[0113] In one embodiment, the blood vessel branch separation module 605 is specifically used for:

[0114] A region masking operation is performed on the merged image of the blood vessel image and the thrombus image to set the pixel value of the second bifurcation region image to the background pixel value, thus obtaining the processed blood vessel image.

[0115] Perform connected component analysis on the processed vascular image to calculate the number of connected components in the current processed vascular image;

[0116] If the number of connected components is less than 3, iterate through the current processed vascular image: using the contour of the second bifurcation region as a reference, perform morphological dilation operation, set the pixel value of the second bifurcation region image to the background pixel value, and analyze the number of connected components until 3 independent vascular connected components are formed in the current processed vascular image.

[0117] In one embodiment, the thrombosis grading calculation module 607 is specifically used for:

[0118] Obtain multiple blood vessel cross sections from the merged image;

[0119] Calculate the percentage of thrombus pixels in each blood vessel cross section, and determine the severity of thrombosis in the portal vein system based on the percentage of the largest thrombus pixels.

[0120] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for determining the severity of portal vein thrombosis.

[0121] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for determining the severity of portal vein system thrombosis.

[0122] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for determining the severity of portal vein system thrombosis.

[0123] In this embodiment of the invention, portal vein vessel images and thrombus images are obtained from image segmentation of the CTA image to be processed. Then, the skeleton lines and vessel images are input into a pre-trained bifurcation region segmentation model to segment the first bifurcation region image from the vessel image. Based on the positional relationship between the first bifurcation region image, the vessel image, and the thrombus image, it is determined whether the first bifurcation region image contains thrombus pixels. If the first bifurcation region image contains thrombus pixels, it is determined which thrombus block the thrombus pixel belongs to, and then it is determined which vascular branch has the thrombus block pixel. Finally, vascular branches and bifurcation region images containing the same thrombus pixels or the same thrombus block are merged. Based on the proportion of thrombus pixels in the vascular cross-section of the merged image, the severity of portal vein thrombosis is determined. This embodiment of the invention achieves automated detection and precise removal of bifurcation branches in portal vein vessel bifurcation regions based on image processing and deep learning, ultimately ensuring the structural integrity of the target vascular branch containing thrombus pixels and the triangular bifurcation region. By calculating the proportion of thrombus pixels in the vascular cross-section, the accuracy of portal vein thrombosis severity assessment is improved.

[0124] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0126] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0127] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0128] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of determining the severity of a thrombus in the portal venous system, characterized in that, The method comprises the following steps: receiving a computer tomography angiography (CTA) image to be processed; obtaining a blood vessel image and a thrombus image of a hepatic portal vein from the CTA image to be processed by image segmentation; the thrombus image comprises one or more thrombus blocks, and each thrombus block has a thrombus block identifier; extracting a blood vessel skeleton from the blood vessel image; inputting the blood vessel skeleton and the blood vessel image into a pre-trained bifurcation region segmentation model to segment a first bifurcation region image from the blood vessel image; the bifurcation region segmentation model is obtained by training a 3D Attention Unet network using historical CTA images; the first bifurcation region image is a binary image; mapping pixel values of an image obtained by merging the blood vessel image and the thrombus image to the first bifurcation region image according to a position of the first bifurcation region image in the blood vessel image, to obtain a second bifurcation region image with restored pixels, and determining whether the second bifurcation region image is mapped with pixels of the thrombus image; when it is determined that the second bifurcation region image is mapped with pixels of the thrombus image, extracting thrombus image pixel coordinates of the second bifurcation region image; determining a thrombus block identifier of a thrombus in the second bifurcation region image according to the extracted thrombus image pixel coordinates of the second bifurcation region image; separating a plurality of first blood vessel branches from the image obtained by merging the blood vessel image and the thrombus image by setting pixels of the second bifurcation region image as background in the image obtained by merging the blood vessel image and the thrombus image; detecting whether each first blood vessel branch comprises pixels of a thrombus block corresponding to the thrombus block identifier, and determining a second blood vessel branch comprising the pixels of the thrombus block corresponding to the thrombus block identifier; merging the second blood vessel branch and the second bifurcation region image to obtain a merged image; determining a severity of a portal vein system thrombus based on a proportion of thrombus pixels in a blood vessel cross section in the merged image.

2. The method of claim 1, wherein, The method for obtaining the blood vessel image and the thrombus image from the CTA image to be processed comprises the following steps: obtaining the blood vessel image and the thrombus image from the CTA image to be processed by image segmentation or a convolutional neural network.

3. The method of claim 1, wherein, The bifurcation region segmentation model comprises an encoder, a decoder and an attention module; the encoder comprises a 3D convolution layer and a down-sampling layer, the decoder comprises an up-sampling layer and a 3D deconvolution layer, and the attention module is embedded between layers of the encoder and the decoder; the attention module gives a pixel associated with a bifurcation region a higher weight than a pixel not associated with the bifurcation region.

4. The method of claim 1, wherein, The method for separating the plurality of first blood vessel branches from the image obtained by merging the blood vessel image and the thrombus image by setting pixels of the second bifurcation region image as background in the image obtained by merging the blood vessel image and the thrombus image comprises the following steps: performing a region mask operation on the image obtained by merging the blood vessel image and the thrombus image to set pixel values of the second bifurcation region image as background pixel values, to obtain a processed blood vessel image; performing connected component analysis on the processed blood vessel image to calculate a number of connected components in the processed blood vessel image; and performing connected component analysis on the processed blood vessel image to calculate a number of connected components in the processed blood vessel image. If the number of connected domains is less than 3, the current processed blood vessel image is looped: based on the outline of the second bifurcation region, morphological dilation operation is performed, the pixel value of the second bifurcation region image is set as the background pixel value, and the number of connected domains is analyzed, until 3 independent blood vessel connected domains are formed in the current processed blood vessel image.

5. The method of claim 1, wherein, Based on the proportion of thrombus pixels in the blood vessel cross section in the merged image, the severity of the portal vein system thrombus is determined, including: Obtaining a plurality of blood vessel cross sections in the merged image; Calculating the proportion of thrombus pixels in each blood vessel cross section, and determining the severity of the portal vein system thrombus according to the maximum proportion of thrombus pixels.

6. A portal vein system thrombus severity determination apparatus characterized by, It includes: A data receiving module for receiving a to-be-processed CTA image; An image preprocessing module for obtaining a blood vessel image and a thrombus image of a hepatic portal vein from image segmentation of the to-be-processed CTA image; extracting a blood vessel skeleton line from the blood vessel image; the thrombus image includes one or more thrombus blocks, and the thrombus block has a thrombus block identifier; A bifurcation region segmentation module for inputting the skeleton line and the blood vessel image into a pre-trained bifurcation region segmentation model to segment a first bifurcation region image from the blood vessel image; the bifurcation region segmentation model is trained by using historical CTA images on a 3D Attention Unet network; the first bifurcation region image is a binary image; A bifurcation region thrombus detection module for mapping pixel values of an image obtained by merging the blood vessel image and the thrombus image to the first bifurcation region image according to the position of the first bifurcation region image in the blood vessel image, obtaining a second bifurcation region image with restored pixels, and determining whether the second bifurcation region image is mapped with pixels of the thrombus image; when it is determined that the second bifurcation region image is mapped with pixels of the thrombus image, the pixel coordinates of the thrombus image of the second bifurcation region image are extracted; according to the extracted pixel coordinates of the thrombus image of the second bifurcation region image, the thrombus block identifier in which the thrombus is located in the second bifurcation region image is determined; A blood vessel branch separation module for separating a plurality of first blood vessel branches from the image obtained by merging the blood vessel image and the thrombus image by setting the pixel values of the second bifurcation region image as background in the image obtained by merging the blood vessel image and the thrombus image; A blood vessel branch removal module for detecting whether each first blood vessel branch includes pixels of a thrombus block corresponding to the thrombus block identifier, determining a second blood vessel branch including the pixels of the thrombus block corresponding to the thrombus block identifier, and merging the second blood vessel branch and the second bifurcation region image to obtain a merged image; A thrombus grading calculation module for determining the severity of the portal vein system thrombus based on the proportion of thrombus pixels in the blood vessel cross section in the merged image.

7. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1-5.

9. A computer program product, characterised in that, The computer program product includes a computer program, and the computer program is executed by the processor to implement the method of any one of claims 1-5.

Images