Method and device for identifying thrombi in the portal vein system

Abstract

The application discloses a kind of hepatic portal vein system thrombus identification method and device, it is related to medical image processing technical field, wherein the method comprises: the CTA image to be handled is input to rough segmentation model, and the output blood vessel cavity;Blood vessel cavity includes blood vessel and thrombus in blood vessel;Rough segmentation model is obtained by training nnunet network;The HU value distribution of pixel in the blood vessel cavity is counted, combined with human liver tissue standard HU value, determine the segmentation threshold for segmenting thrombus;Based on density clustering algorithm, determine the cluster center coordinates of thrombus in the blood vessel;With cluster center coordinates as seed point, according to the similarity of pixel HU value in blood vessel cavity and the segmentation threshold, region growth is carried out, until encountering blood vessel boundary or there is no pixel point to be grown, obtain fine segmentation result;The fine segmentation result includes each pixel where thrombus is located.The application can improve the accuracy of hepatic portal vein system thrombus identification.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method and device for identifying thrombosis in the portal vein system. 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] Portal vein system thrombosis (PVST) refers to thrombosis of the main portal vein and / or its left and right branches, with or without thrombosis of the mesenteric and splenic veins. Acute PVST can easily lead to mesenteric ischemia and even serious adverse outcomes such as intestinal necrosis; chronic PVST can lead to portal vein occlusion or cavernous degeneration of the portal vein, resulting in secondary portal hypertension. Therefore, early and accurate identification of thrombi is crucial.

[0004] Current technologies for identifying thrombi in the human body are mostly based on image processing, combined with other data such as pressure measurements. However, the abdominal organs are stacked on top of each other and lack stable bony structures. The relative positions of abdominal organs vary greatly among different people. The portal vein system's vascular network is intertwined with the hepatic artery and vein networks. Thrombi in the portal vein system are small structures attached to the walls of the blood vessels in the portal vein system, making their identification extremely difficult. Existing thrombus identification methods cannot accurately identify portal vein system thrombi, and currently, thrombus identification still requires visual inspection, with an accuracy of only 40-50%. Therefore, the accuracy of existing portal vein system thrombus identification needs to be improved. Summary of the Invention

[0005] This invention provides a method for identifying thrombi in the portal vein system to improve the accuracy of thrombus identification. The method includes:

[0006] The CTA (Computed Tomography Angiography) image to be processed is input into a pre-trained coarse segmentation model, which outputs the vascular lumen in the CTA image to be processed; the vascular lumen includes blood vessels and thrombi in the blood vessels; the coarse segmentation model is trained on an nnunet (Self-adapting Framework for U-Net-Based Medical Image Segmentation) network using historical images of the human portal vein system with labeled blood vessels and thrombi.

[0007] The distribution of HU values ​​of pixels within the blood vessel lumen is statistically analyzed, and combined with the standard HU values ​​of human liver tissue, a segmentation threshold for segmenting thrombi is determined.

[0008] Based on the density clustering algorithm, the coordinates of the cluster centers of the thrombi in the blood vessels are determined;

[0009] Using the cluster center coordinates as seed points, region growth is performed based on the similarity between the HU value of pixels in the blood vessel lumen and the segmentation threshold until the blood vessel boundary is encountered or there are no pixels to be grown, thus obtaining a fine segmentation result; the fine segmentation result includes each pixel where the thrombus is located.

[0010] This invention also provides a portal vein system thrombosis identification device to improve the accuracy of portal vein system thrombosis identification. The device includes:

[0011] The coarse segmentation processing module is used to input the CTA image to be processed into a pre-trained coarse segmentation model and output the vascular lumen in the CTA image to be processed; the vascular lumen includes blood vessels and thrombi in blood vessels; the coarse segmentation model is trained on the nnunet network using historical images of the human portal vein system with blood vessels and thrombi annotated.

[0012] The grayscale processing module is used to statistically analyze the HU value distribution of pixels within the blood vessel lumen and, in conjunction with the standard HU value of human liver tissue, determine the segmentation threshold for segmenting thrombi.

[0013] The fine segmentation processing module is used to determine the cluster center coordinates of the thrombus in the blood vessel based on the density clustering algorithm; using the cluster center coordinates as seed points, it performs region growth based on the similarity between the HU value of the pixel in the blood vessel lumen and the segmentation threshold until the blood vessel boundary is encountered or there are no pixels to be grown, thus obtaining the fine segmentation result; the fine segmentation result includes each pixel where the thrombus is located.

[0014] 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 identifying portal vein thrombosis.

[0015] 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 identifying thrombosis in the portal vein system.

[0016] 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 identifying portal vein thrombosis.

[0017] In this embodiment of the invention, the CTA image to be processed is input into a pre-trained coarse segmentation model, which outputs the vascular lumen in the CTA image. The vascular lumen includes blood vessels and thrombi within them. The coarse segmentation model is trained on an NNUNET network using historically labeled images of the human portal vein system with blood vessels and thrombi. The distribution of HU values ​​of pixels within the vascular lumen is statistically analyzed, and combined with the standard HU values ​​of human liver tissue, a segmentation threshold for segmenting thrombi is determined. Based on a density clustering algorithm, the coordinates of the cluster centers of the thrombi in the blood vessels are determined. Using the coordinates of the cluster centers as seed points, region growing is performed based on the similarity between the HU values ​​of pixels in the vascular lumen and the segmentation threshold until a blood vessel boundary is encountered or there are no pixels to be grown, resulting in a fine segmentation result. The fine segmentation result includes each pixel containing the thrombus. In this embodiment of the invention, the NNUNET network is first used to perform automatic coarse segmentation of blood vessels and thrombi to obtain complete vascular cavities. The segmentation threshold is determined based on the HU value, and the region growth criterion is constructed using the segmentation threshold to perform secondary precise segmentation and identification of thrombi. Compared with existing methods, this method can identify thrombi in the portal vein system based on the characteristics of thrombi in the portal vein system, which greatly improves the accuracy of thrombus identification in the portal vein system. Attached Figure Description

[0018] 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:

[0019] Figure 1 This is a flowchart illustrating the method for identifying portal vein thrombosis in an embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of vascular thrombus segmentation in an embodiment of the present invention;

[0021] Figure 3 This is a histogram showing the distribution of HU values ​​of pixels within the blood vessel lumen in an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram of the fine segmentation results of thrombi in an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram of the portal vein system thrombosis identification device in an embodiment of the present invention. Detailed Implementation

[0024] 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.

[0025] First, the technical terms involved in the embodiments of the present invention will be explained.

[0026] The Dice metric, also known as the Dice coefficient or Sørensen-Dice coefficient, is a statistic used in machine learning and image processing, particularly in binary segmentation tasks, to measure the similarity between two sample sets. It is commonly used to evaluate the similarity between model predictions and the true labels.

[0027] HU value: The Hounsfield Unit is a standardized unit used to quantify tissue density in CT images, named after the British scientist Godfrey Hounsfield (who was also a co-inventor of CT). Essentially, it converts the X-ray attenuation coefficients of different tissues into a unified numerical scale. A higher HU value results in a brighter (whiter) CT image; a lower HU value results in a darker (blacker) image.

[0028] nnU-Net: nnU-Net is a fully automated medical image segmentation framework based on the classic U-Net architecture, which can automatically adjust hyperparameters according to dataset attributes.

[0029] Existing technologies have poor accuracy in automatically identifying portal vein thrombosis. Typically, thin-slice CTA images of the abdominal venous phase are used for visual thrombosis identification, but the accuracy is only 40-50%. Therefore, this invention proposes a method for identifying portal vein thrombosis to improve the accuracy of automatic identification.

[0030] Figure 1 This is a flowchart illustrating the method for identifying portal vein thrombosis in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0031] Step 101: Input the CTA image to be processed into a pre-trained coarse segmentation model and output the vascular lumen in the CTA image to be processed; the vascular lumen includes blood vessels and thrombi in the blood vessels; the coarse segmentation model is trained on the nnunet network using historical images of the human portal vein system with labeled blood vessels and thrombi.

[0032] Step 102: Statistically analyze the HU value distribution of pixels within the blood vessel lumen, and combine it with the standard HU value of human liver tissue to determine the segmentation threshold for segmenting thrombi.

[0033] Step 103: Determine the coordinates of the cluster centers of the thrombi in the blood vessels based on the density clustering algorithm;

[0034] Step 104: Using the cluster center coordinates as seed points, perform region growth based on the similarity between the pixel HU value in the blood vessel lumen and the segmentation threshold until the blood vessel boundary is encountered or there are no pixels to be grown, thus obtaining a fine segmentation result; the fine segmentation result includes each pixel where the thrombus is located.

[0035] The method for identifying portal vein thrombosis in the present invention will be explained in detail below.

[0036] Part 1: Blood Vessels and Thrombi.

[0037] In practice, a coarse segmentation model is pre-trained and used to process the CTA image to be processed, thereby obtaining the vascular lumen in the CTA image.

[0038] In this embodiment, the coarse segmentation model is obtained by training a convolutional network, the UET network, for biomedical image segmentation using historical images of the human portal vein system with labeled blood vessels and thrombi.

[0039] In one embodiment, the coarse segmentation model can be trained as follows:

[0040] A training set was constructed using historically labeled images of the human portal venous system with blood vessels and thrombi.

[0041] Using the training set, a coarse segmentation model is trained by employing 5-fold cross-validation and Adam (Adaptive Moment Estimation) gradient optimization methods.

[0042] For example, a large number of abdominal venous phase CTA images are collected. First, the images are automatically labeled by machine, then the labels are reviewed and corrected by professionals, and then sample data is obtained through morphological opening and closing operations. The sample data includes positive and negative data. The training set and validation set are split in an 8:2 ratio, and the training is performed on a V100 GPU with 500 epochs, deep supervision mode, learning rate of 0.001, and Adam gradient optimization.

[0043] Specifically, there were 84 positive cases and 7 negative cases. The test set was divided into two datasets: our own hospital's dataset (22 cases) and datasets from two multi-center sites (25 cases each from site A and site B). The DICE index for our own hospital's dataset was 0.71, for site A's dataset it was 0.682, and for site B's dataset it was 0.643.

[0044] In existing technologies, collecting imaging data of portal vein thrombosis is difficult, and the abdominal cavity background is complex. Accurate annotation of thrombi layer by layer on thin-slice data requires a very high level of expertise from the annotators. To address this challenge, in one embodiment of this invention, a coarse segmentation model is obtained through federated learning training. In implementation, a federated learning architecture is constructed, including a server and various clients. The server collects model update parameters from each client, and each client trains the model using its local data. Federated learning training can improve the model's accuracy and generalization ability.

[0045] Step 101 involves inputting the CTA image to be processed into a pre-trained coarse segmentation model and outputting the vascular lumen in the CTA image. This may include: inputting the CTA image to be processed into the pre-trained coarse segmentation model and outputting the blood vessel and the thrombus within the blood vessel; merging the blood vessel and the thrombus within the blood vessel using a morphological closing operation to obtain the vascular lumen in the CTA image. Filling gaps and smoothing boundaries through dilation and erosion operations in the morphological closing operation helps improve the quality of subsequent thrombus segmentation and reduce noise.

[0046] Figure 2 This is a schematic diagram of vascular thrombus segmentation in an embodiment of the present invention, as shown below. Figure 2 As shown, the left image displays the CTA image to be processed, and the right image shows the identified vascular lumen. Green represents blood vessels, and red represents thrombi within the vessels. It should be noted that during actual processing, the identified thrombi in the blood vessels may include multiple thrombi. For example, small thrombi may be present in multiple locations within the blood vessel, which are represented as multiple small thrombus areas in the image.

[0047] The second part is the processing of pixel distribution within the cavity.

[0048] In step 102, the distribution of HU values ​​of pixels within the blood vessel lumen is statistically analyzed, and combined with the standard HU value of human liver tissue, a segmentation threshold for segmenting thrombi is determined.

[0049] In this embodiment of the invention, each pixel of the CTA image to be processed has a HU value. Considering the standard HU value of human liver tissue, a segmentation threshold is designed to segment thrombi, and the vascular cavity after coarse segmentation model processing is finely segmented.

[0050] In one embodiment, statistically analyzing the HU value distribution of pixels within the blood vessel lumen and combining it with the standard HU value of human liver tissue to determine a segmentation threshold for segmenting thrombi may include:

[0051] A histogram is constructed using the HU values ​​of the pixels in the blood vessel cavity; the horizontal axis of the histogram represents the HU values, and the vertical axis represents the number of pixels with each HU value.

[0052] If the histogram shows two peaks, the HU value with the smaller horizontal axis value in the two peaks is recorded as the segmentation threshold;

[0053] If the histogram does not show a bimodal pattern, it indicates that there is no thrombus in the CTA image to be processed.

[0054] In this embodiment of the invention, the data used for segmenting the portal venous system vessels is venous-phase CTA imagery, acquired during the period after contrast agent injection and when the veins are visualized. Vessels are characterized by high-density shadows, while thrombi, being emboli attached to the vessel wall, are characterized by low-density shadows. If the vessel segmentation is inaccurate, such as expanding the vessel boundary several times, it may include other non-vascular tissues. Incorrectly segmented non-vascular portions may be liver tissue. When calculating the pixels within the lumen, the vessel labels are subjected to three-dimensional erosion. Since thrombi are located inside the vessel lumen, this operation will not significantly affect the thrombus. The grayscale values ​​within the lumen are then calculated. Figure 3 This is a histogram showing the distribution of HU values ​​of pixels within the blood vessel lumen in an embodiment of the present invention, such as... Figure 3 As shown, the horizontal axis represents the HU value, and the vertical axis represents the number of pixels. If the grayscale distribution exhibits a clear bimodal distribution, with one peak corresponding to a grayscale value (horizontal axis: unit HU) higher than that of liver tissue and the other lower, and the HU value of liver tissue in CTA images is 70-80 HU, then a thrombus is present in the lumen; otherwise, it is not present. The HU values ​​corresponding to the two peaks are obtained, and the lower HU value is used as the thrombus segmentation threshold.

[0055] Part Three: Fine Segmentation of Thrombi.

[0056] In step 103, the cluster center coordinates of the thrombi in the blood vessels are determined based on a density clustering algorithm, which includes the DBSCAN (density-based spatial application clustering with noise) algorithm.

[0057] In practice, the identified cluster center coordinates may include the cluster center coordinates of multiple thrombi.

[0058] In step 104, using the cluster center coordinates as seed points, region growth is performed based on the similarity between the pixel HU value in the blood vessel lumen and the segmentation threshold until the blood vessel boundary is encountered or there are no pixels to be grown, thus obtaining a fine segmentation result; the fine segmentation result includes each pixel where the thrombus is located.

[0059] For example, using the cluster center coordinates as the seed point, and taking the difference between the pixel HU value in the blood vessel lumen and the segmentation threshold as less than 10 HU values ​​as the growth criterion, the system searches for pixels in 26 neighborhood directions until it encounters the blood vessel boundary or there are no pixels to be grown, thus obtaining a fine segmentation result.

[0060] The three key elements of a region growing algorithm are: seed point, growth criterion, and stopping condition. In this example, the DBSCAN algorithm is used to obtain the coordinates of the cluster centers of thrombi. Thrombi can be single or multiple, and the cluster centers correspond to one or more clusters, which are set as seed points. The corresponding seed points are also one or more. The growth criterion is the similarity of pixel grayscale values. A difference of less than 10 HU between the grayscale value of a pixel within the lumen and the segmentation threshold is used as the growth criterion, and similar pixels are searched in 26 neighboring directions. The stopping condition is when a blood vessel boundary is encountered or no seed point to be grown is found.

[0061] Figure 4 This is a schematic diagram of the fine segmentation results of thrombi in an embodiment of the present invention, for reference. Figure 4 There are 4 cluster centers. Based on the 4 cluster centers, 4 thrombi were identified. For example, the fine segmentation result can be: to indicate whether it is a thrombus by marking the pixel as 1 or 0, where 1 indicates a thrombus area and 0 indicates a non-thrombus area.

[0062] In one embodiment, after obtaining the fine segmentation result, the process may further include: performing a closing operation on the fine segmentation result to obtain the final thrombus segmentation result.

[0063] For example, after initial segmentation using NNUNET and two steps of region growing based on lumen grayscale statistics, the accurate segmentation result of the thrombus is obtained. Combined with post-processing of a one-scale closing operation in three-dimensional morphology to compensate for pixel gaps caused by white noise interference in region growing, the final thrombus segmentation result can be obtained.

[0064] Subsequently, based on the final thrombus segmentation results, further calculations can be performed to obtain thrombus quantitative parameters such as vessel volume, vessel surface area, vessel curvature, thrombus volume, thrombus surface area, vessel-to-thrombus volume ratio, maximum thrombus cross-section ratio, and average thrombus cross-section ratio.

[0065] In summary, this invention, based on the open-source framework NNUNET, automatically segments blood vessels and thrombi. Utilizing the physiological characteristics of thrombi attaching to blood vessels, a morphological closure operation is employed to merge thrombi and blood vessels, forming a complete vascular lumen. Within the vascular lumen, histogram statistics are used to determine the grayscale value of the thrombus. This grayscale value serves as a growth criterion. The results of the initial DBSCAN clustering of the segmented thrombi are used as seed points. A region growing method is then used for secondary refinement of the thrombus segmentation, resulting in accurate thrombus segmentation. This achieves precise identification of thrombi in the portal vein system.

[0066] This invention also provides a portal vein system thrombosis identification device, as described in the following embodiments. Since the principle by which this device solves the problem is similar to the portal vein system thrombosis identification method, the implementation of this device can refer to the implementation of the portal vein system thrombosis identification method, and repeated details will not be elaborated further.

[0067] Figure 5 This is a schematic diagram of the portal vein system thrombosis identification device in an embodiment of the present invention, as shown below. Figure 5 As shown, the device 500 includes:

[0068] The coarse segmentation processing module 501 is used to input the CTA image to be processed into a pre-trained coarse segmentation model and output the vascular cavity in the CTA image to be processed; the vascular cavity includes blood vessels and thrombi in blood vessels; the coarse segmentation model is trained on the nnunet network using historical images of the human portal vein system with blood vessels and thrombi annotated.

[0069] The grayscale processing module 502 is used to statistically analyze the HU value distribution of pixels within the blood vessel lumen and, in conjunction with the standard HU value of human liver tissue, determine the segmentation threshold for segmenting thrombi.

[0070] The fine segmentation processing module 503 is used to determine the cluster center coordinates of the thrombus in the blood vessel based on the density clustering algorithm; using the cluster center coordinates as seed points, it performs region growth based on the similarity between the pixel HU value in the blood vessel lumen and the segmentation threshold until the blood vessel boundary is encountered or there are no pixels to be grown, thus obtaining a fine segmentation result; the fine segmentation result includes each pixel where the thrombus is located.

[0071] In one embodiment, the coarse segmentation model is trained as follows:

[0072] A training set was constructed using historically labeled images of the human portal venous system with blood vessels and thrombi.

[0073] Using the training set, a coarse segmentation model was trained by employing 5-fold cross-validation and Adam gradient optimization.

[0074] In one embodiment, the coarse segmentation processing module 501 is specifically used for:

[0075] The CTA image to be processed is input into a pre-trained coarse segmentation model, which outputs blood vessels and thrombi in the blood vessels.

[0076] The morphological closure operation method is used to merge blood vessels and thrombi in blood vessels to obtain the vascular lumen in the CTA image to be processed.

[0077] In one embodiment, the grayscale processing module 502 is specifically used for:

[0078] A histogram is constructed using the HU values ​​of the pixels within the blood vessel lumen; the horizontal axis of the histogram represents the HU values, and the vertical axis represents the number of pixels with each HU value.

[0079] If the histogram shows two peaks, the HU value with the smaller horizontal axis value in the two peaks is recorded as the segmentation threshold;

[0080] If the histogram does not show a bimodal pattern, it indicates that there is no thrombus in the CTA image to be processed.

[0081] In one embodiment, the fine segmentation processing module 503 is specifically used for:

[0082] Using the cluster center coordinates as the seed point, and taking the difference between the pixel HU value in the blood vessel lumen and the segmentation threshold as less than 10 HU values ​​as the growth criterion, the system searches for pixels in 26 neighborhood directions until it encounters the blood vessel boundary or there are no pixels to be grown, thus obtaining a fine segmentation result.

[0083] In one embodiment, the device 500 further includes:

[0084] The final result determination module is used to perform a closing operation on the fine segmentation result after the fine segmentation processing module 503 obtains the fine segmentation result, so as to obtain the final thrombus segmentation result.

[0085] 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 identifying portal vein thrombosis.

[0086] 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 identifying thrombosis in the portal vein system.

[0087] 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 identifying portal vein thrombosis.

[0088] In this embodiment of the invention, the CTA image to be processed is input into a pre-trained coarse segmentation model, which outputs the vascular lumen in the CTA image. The vascular lumen includes blood vessels and thrombi within them. The coarse segmentation model is trained on an NNUNET network using historically labeled images of the human portal vein system with blood vessels and thrombi. The distribution of HU values ​​of pixels within the vascular lumen is statistically analyzed, and combined with the standard HU values ​​of human liver tissue, a segmentation threshold for segmenting thrombi is determined. Based on a density clustering algorithm, the coordinates of the cluster centers of the thrombi in the blood vessels are determined. Using the coordinates of the cluster centers as seed points, region growing is performed based on the similarity between the HU values ​​of pixels in the vascular lumen and the segmentation threshold until a blood vessel boundary is encountered or there are no pixels to be grown, resulting in a fine segmentation result. The fine segmentation result includes each pixel containing the thrombus. In this embodiment of the invention, the NNUNET network is first used to perform automatic coarse segmentation of blood vessels and thrombi to obtain complete vascular cavities. The segmentation threshold is determined based on the HU value, and the region growth criterion is constructed using the segmentation threshold to perform secondary precise segmentation and identification of thrombi. Compared with existing methods, this method can identify thrombi in the portal vein system based on the characteristics of thrombi in the portal vein system, which greatly improves the accuracy of thrombus identification in the portal vein system.

[0089] 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.

[0090] 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.

[0091] 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 1 The function specified in one or more boxes.

[0092] 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.

[0093] 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 identifying a thrombus in a hepatic portal vein system, characterized by, The method comprises the following steps: inputting a to-be-processed computed tomography angiography (CTA) image into a pre-trained rough segmentation model to output a blood vessel cavity in the to-be-processed CTA image; the blood vessel cavity comprises a blood vessel and a thrombus in the blood vessel; the rough segmentation model is trained by using historical human liver portal vein system images that have been labeled with blood vessels and thrombi based on a medical image segmentation self-adaptive framework nnunet network based on a U-shaped network structure; statistically counting a Hounsfield unit (HU) value distribution of pixels in the blood vessel cavity, combining a standard HU value of human liver tissue to determine a segmentation threshold value for segmenting the thrombus; determining a clustering center coordinate of the thrombus in the blood vessel based on a density clustering algorithm; taking the clustering center coordinate as a seed point, performing region growing according to a similarity between a HU value of a pixel in the blood vessel cavity and the segmentation threshold value until a blood vessel boundary is encountered or there is no pixel point to be grown, to obtain a fine segmentation result; the fine segmentation result comprises each pixel where the thrombus is located; wherein the step of statistically counting the HU value distribution of the pixels in the blood vessel cavity, combining the standard HU value of human liver tissue to determine the segmentation threshold value for segmenting the thrombus comprises: constructing a histogram by using the HU values of the pixels in the blood vessel cavity; the horizontal axis of the histogram is the HU value, and the vertical axis is the number of pixels of each HU value; if the histogram shows double peaks, recording the HU value with a smaller horizontal axis value in the double peaks as the segmentation threshold value; if the histogram does not show double peaks, determining that there is no thrombus in the to-be-processed CTA image.

2. The method of claim 1, wherein, The rough segmentation model is trained in the following manner: constructing a training set by using historical human liver portal vein system images that have been labeled with blood vessels and thrombi; training the training set by using 5-fold cross-validation and an adaptive moment estimation (adam) gradient optimization method to obtain the trained rough segmentation model.

3. The method of claim 2, wherein, inputting the to-be-processed CTA image into the pre-trained rough segmentation model to output the blood vessel cavity in the to-be-processed CTA image, comprising: inputting the to-be-processed CTA image into the pre-trained rough segmentation model to output the blood vessel and the thrombus in the blood vessel; merging the blood vessel and the thrombus in the blood vessel by using a morphological closing operation method to obtain the blood vessel cavity in the to-be-processed CTA image.

4. The method of claim 1, wherein, taking the clustering center coordinate as a seed point, performing region growing according to a similarity between a HU value of a pixel in the blood vessel cavity and the segmentation threshold value until a blood vessel boundary is encountered or there is no pixel point to be grown, to obtain a fine segmentation result; the fine segmentation result comprises each pixel where the thrombus is located; taking the clustering center coordinate as a seed point, taking a difference between the HU value of the pixel in the blood vessel cavity and the segmentation threshold value being less than 10 HU values as a growth criterion, searching for a pixel point in 26 neighborhood directions until a blood vessel boundary is encountered or there is no pixel point to be grown, to obtain the fine segmentation result.

5. The method of claim 4, wherein, After obtaining the fine segmentation result, further comprising: performing a closing operation on the fine segmentation result to obtain a final thrombus segmentation result.

6. A hepatic portal vein system thrombus recognition apparatus characterized by comprising: The method comprises the following steps: a rough segmentation processing module is configured to input a to-be-processed computed tomography angiography (CTA) image into a pre-trained rough segmentation model to output a blood vessel cavity in the to-be-processed CTA image; the blood vessel cavity comprises a blood vessel and a thrombus in the blood vessel; the rough segmentation model is trained by using historical human liver portal vein system images that have been labeled with blood vessels and thrombi based on a nnunet network; The gray scale processing module is configured to count HU value distribution of the pixels in the blood vessel lumen, determine a segmentation threshold for segmenting the thrombus in combination with standard HU value of human liver tissue, and determine the segmentation threshold for segmenting the thrombus. The fine segmentation processing module is configured to determine a cluster center coordinate of the thrombus in the blood vessel based on a density clustering algorithm, take the cluster center coordinate as a seed point, and perform region growing according to similarity of HU value of the pixels in the blood vessel lumen and the segmentation threshold until a blood vessel boundary is encountered or there is no pixel point to be grown, so as to obtain a fine segmentation result. The fine segmentation result includes each pixel where the thrombus is located. The gray scale processing module is specifically configured to: construct a histogram using the HU value of the pixels in the blood vessel lumen, wherein the horizontal axis of the histogram is the HU value and the vertical axis is the number of pixels of each HU value; if the histogram shows double peaks, record the HU value with a smaller horizontal axis value in the double peaks as the segmentation threshold; if the histogram does not show double peaks, determine that the thrombus does not exist in the CTA image to be processed.

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 to 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 to 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 to 5.

Images