Vascular embolism identification method based on deep learning and related equipment

Abstract

The invention relates to a vascular embolism recognition method based on deep learning, and the method comprises the steps: obtaining a seed point of a to-be-recognized abdominal medical image according to an abdominal aorta blood vessel image and a trained first model, carrying out the semi-automatic segmentation of a mesenteric artery region in the image through the seed point through a GrowCut algorithm, extracting an image of the mesenteric artery region, and carrying out the recognition of the mesenteric artery region. And delimiting an artery blood vessel center line for the image of the mesenteric artery area by using a VTK platform, inputting the image of the mesenteric artery area with the delimited artery blood vessel center line into a second model, and determining whether thrombus exists in the abdominal medical image to be identified and the position of the thrombus. According to the method, the influence factors of the acute superior mesenteric artery thrombus can be more comprehensively captured, the predictive factors with high correlation are screened out, the superior mesenteric artery thrombus on the CT image is accurately recognized, and the predictive capacity of the model is enhanced. The invention also relates to a device and a storage medium.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a vascular embolism recognition method based on deep learning and related equipment. Background Art

[0002] For existing CT images of thrombus with a single component, the grayscale characteristics are often used as the entry point, and the pixel value distribution law is found by traversing the area around the pixel to find the edge of the lesion, which is a binary classification task. At present, the classification method based on statistical learning in the field of image classification cannot fully capture the influencing factors of superior mesenteric artery thrombosis on CT images, and thus cannot accurately identify superior mesenteric artery thrombosis on CT images. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a vascular embolism identification method and related equipment based on deep learning.

[0004] In a first aspect, the present invention provides a method for identifying vascular embolism based on deep learning, the method comprising: Extracting a regional image from an abdominal medical image to be identified, wherein the regional image is an abdominal aorta blood vessel image; Obtaining a seed point of the to-be-identified abdominal medical image according to the abdominal aorta vascular image and the trained first model, wherein the first model is constructed based on a Unet network and is obtained by training the Unet network using an abdominal medical sample image set, and the seed point is a pixel point in a pixel of the abdominal aorta in the to-be-identified medical image; Using the seed points to semi-automatically segment the mesenteric artery region in the regional image using the GrowCut algorithm, and extracting the image of the mesenteric artery region; Delineating the centerline of the artery in the image of the mesenteric artery region using the VTK platform, wherein the VTK platform is an application for three-dimensional computer graphics; The image of the mesenteric artery area with the arterial centerline delineated is input into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal vascular image characteristics and the thrombus vascular image characteristics.

[0005] Based on the above embodiment, obtaining the seed point of the to-be-identified abdominal medical image according to the abdominal aorta vascular image and the trained first model specifically includes: Inputting the abdominal aorta blood vessel image into the trained first model to obtain the corresponding pixel points of the abdominal aorta in the abdominal aorta blood vessel image; Based on the corresponding pixel points of the abdominal aorta in the abdominal aorta vascular image, the seed points of the abdominal medical image to be identified are determined.

[0006] Based on the above embodiment, determining the seed point of the to-be-identified abdominal medical image based on the corresponding pixel point of the abdominal aorta in the abdominal aorta vascular image specifically includes: Determining a sliding window range according to the position of each pixel point in the abdominal aorta blood vessel image; Traversing the pixel points within the sliding window to obtain an average value of the pixel values ​​within the sliding window; Determine an original seed point in the abdominal aorta blood vessel image according to an average value of all pixel values ​​within the sliding window; Calculating the weighted distance between each pixel point in the abdominal aorta blood vessel image and the original seed point; The seed point of the abdominal medical image to be identified is determined according to the pixel value of each pixel point in the abdominal aorta blood vessel image, and the weighted distance, the maximum weighted distance and the minimum weighted distance between each pixel point and the original seed point.

[0007] Based on the above embodiment, using the seed point to semi-automatically segment the mesenteric artery region in the regional image using the GrowCut algorithm to extract the image of the mesenteric artery region specifically includes: S1, marking the seed point of the abdominal aorta blood vessel image as a foreground point, and marking other pixel points as background points; S2, obtaining the neighborhood points of each foreground point, and calculating the similarity value between the foreground point and each neighborhood point, wherein the similarity value includes calculating the color similarity, gray value similarity and spatial position similarity between the foreground point and each neighborhood point; S3, expanding the neighborhood points whose similarity values ​​are greater than a preset threshold as the foreground points, and returning to step S2; S4. In this round of iteration, no neighboring points are expanded to foreground points, and all the foreground points in the abdominal aorta vascular image are extracted to obtain a foreground image, and the remaining pixels are the background image, wherein the foreground image is an image of the mesenteric artery region.

[0008] Based on the above embodiment, the image of the mesenteric artery region is imaged using the VTK platform to delineate the arterial centerline of the image of the mesenteric artery region, wherein the VTK platform is an application for three-dimensional computer graphics, specifically including: Extracting a CSV file of the image of the mesenteric artery region, and obtaining the RAS coordinates of the end point of the superior mesenteric artery in the CSV file, where the RSA coordinates are right upper anterior coordinates used in the medical field; Convert the RAS coordinates of the end point of the superior mesenteric artery into a data object in the VTK platform; The centerline of the artery is calculated using the polygonal data object of the superior mesenteric artery in the VTK platform and the data object of the end point of the superior mesenteric artery in the VTK platform.

[0009] Based on the above embodiment, the method further includes: Acquire a sample image of the mesenteric artery region where the arterial blood vessel centerline is delineated in the abdominal medical sample image set; Extracting a local binary pattern feature vector of the sample image of the mesenteric artery region demarcating the arterial vessel centerline; Extracting gray-level co-occurrence matrix feature vectors of sample images of the mesenteric artery region demarcating the arterial vessel centerline; Combining the local binary pattern feature vector and the gray level co-occurrence matrix feature vector to obtain a comprehensive feature vector; The SVN classifier is trained using the comprehensive feature vector to obtain the second model.

[0010] Based on the above embodiment, the method further includes: Acquire an original set of abdominal medical sample images, and screen out images with a slice thickness of 5 mm from the original set of abdominal medical sample images; The screened images were subjected to Gaussian filtering to reduce noise interference, and the screened images were adjusted by setting the window width to 3200 and the window level to -1400 to obtain the abdominal medical sample image set.

[0011] In a second aspect, the present invention further provides a device for identifying vascular embolism based on deep learning, the device comprising: A first processing module is used to extract a regional image from the abdominal medical image to be identified, wherein the regional image is an abdominal aorta blood vessel image; a second processing module, for obtaining a seed point of the abdominal medical image to be identified based on the abdominal aorta vascular image and the trained first model, wherein the first model is constructed based on a Unet network and is obtained by training the Unet network using a set of abdominal medical sample images, and the seed point is a pixel point in a pixel of the abdominal aorta in the medical image to be identified; A third processing module is used to use the GrowCut algorithm to semi-automatically segment the mesenteric artery region in the regional image using the seed point to extract the image of the mesenteric artery region; A fourth processing module is used to delineate the centerline of the artery in the image of the mesenteric artery region using a VTK platform, wherein the VTK platform is an application for three-dimensional computer graphics; The fifth processing module is used to input the image of the mesenteric artery area delineating the center line of the arterial blood vessel into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal blood vessel image characteristics and the thrombus blood vessel image characteristics.

[0012] Based on the above technical solution, further based on the above embodiment, the second processing module is specifically used to input the abdominal aorta blood vessel image into the trained first model to obtain the corresponding pixel points of the abdominal aorta in the abdominal aorta blood vessel image; Based on the corresponding pixel points of the abdominal aorta in the abdominal aorta vascular image, the seed points of the abdominal medical image to be identified are determined.

[0013] Based on the above embodiment, the second processing module is specifically used to determine the sliding window range according to the position of each pixel point in the abdominal aorta blood vessel image; Traversing the pixel points within the sliding window to obtain an average value of the pixel values ​​within the sliding window; Determine an original seed point in the abdominal aorta blood vessel image according to an average value of all pixel values ​​within the sliding window; Calculating the weighted distance between each pixel point in the abdominal aorta blood vessel image and the original seed point; The seed point of the abdominal medical image to be identified is determined according to the pixel value of each pixel point in the abdominal aorta blood vessel image, and the weighted distance, the maximum weighted distance and the minimum weighted distance between each pixel point and the original seed point.

[0014] Based on the above embodiment, the third processing module includes a first subunit, a second subunit, a third subunit, and a fourth subunit; A first subunit is used to mark the seed point of the abdominal aorta blood vessel image as a foreground point, and mark other pixel points as background points; A second subunit is used to obtain a neighborhood point of each foreground point, and calculate a similarity value between the foreground point and each neighborhood point, wherein the similarity value includes calculating a color similarity, a gray value similarity, and a spatial position similarity between the foreground point and each neighborhood point; A third subunit is used to expand the neighborhood points whose similarity values ​​are greater than a preset threshold into the foreground points, and return to the second subunit; The fourth subunit is used for extracting all the foreground points in the abdominal aorta vascular image when no neighboring points are expanded as foreground points in this iteration to obtain a foreground image, and the remaining pixels are the background image, wherein the foreground image is an image of the mesenteric artery area.

[0015] Based on the above embodiment, the fourth processing module is specifically used to extract the CSV file of the image of the mesenteric artery region, and the RAS coordinates of the end point of the superior mesenteric artery in the CSV file, where the RSA coordinates are the upper right anterior coordinates used in the medical field; Convert the RAS coordinates of the end point of the superior mesenteric artery into a data object in the VTK platform; The centerline of the artery is calculated using the polygonal data object of the superior mesenteric artery in the VTK platform and the data object of the end point of the superior mesenteric artery in the VTK platform.

[0016] Based on the above embodiment, a sixth processing module is further included, the sixth processing module is used to obtain a sample image of the mesenteric artery region where the arterial blood vessel centerline is delineated in the abdominal medical sample image set; Extracting a local binary pattern feature vector of the sample image of the mesenteric artery region demarcating the arterial vessel centerline; Extracting gray-level co-occurrence matrix feature vectors of sample images of the mesenteric artery region demarcating the arterial vessel centerline; Combining the local binary pattern feature vector and the gray level co-occurrence matrix feature vector to obtain a comprehensive feature vector; The SVN classifier is trained using the comprehensive feature vector to obtain the second model.

[0017] Based on the above embodiment, the sixth processing module is specifically used to obtain an original set of abdominal medical sample images, and screen out images with a layer thickness of 5 mm from the original set of abdominal medical sample images; The screened images were subjected to Gaussian filtering to reduce noise interference, and the screened images were adjusted by setting the window width to 3200 and the window level to -1400 to obtain the abdominal medical sample image set.

[0018] In a third aspect, the present invention further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, a deep learning-based vascular embolism identification method described in any one of the first aspects is implemented.

[0019] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for identifying vascular embolism based on deep learning as described in any one of the first aspects is implemented.

[0020] The present invention incorporates abdominal enhanced CT imaging data into the analysis, combines multiple laboratory test indicators, vital signs indicators and demographic information, and constructs a multi-dimensional machine learning model. This comprehensive data method can more comprehensively capture the influencing factors of acute superior mesenteric artery thrombosis and screen out predictive factors with strong correlation, thereby assisting clinicians in making more accurate diagnoses and providing a scientific basis for clinical prevention and treatment. A rapid location and identification method for acute superior mesenteric artery thrombosis based on a supervised machine learning classification algorithm was constructed, which can better locate and identify acute superior mesenteric artery thrombosis, help clinicians identify potential acute superior mesenteric artery thrombosis as early as possible, and ultimately achieve the purpose of early detection, early diagnosis, and early treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings: Figure 1 is a flowchart of a method for identifying vascular embolism based on deep learning provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of a vascular embolism identification device based on deep learning provided by another embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0023] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0024] It should be understood that superior mesenteric artery thrombosis is a disease caused by emboli entering the superior mesenteric artery and causing obstruction, which can cause vascular intestinal obstruction and then develop into acute ischemic necrosis of the intestine. In recent years, with the continuous improvement of diagnostic technology and the continuous increase in the incidence of vascular diseases, vascular acute abdomen has also increased. Mesenteric artery ischemia is the most common, and superior mesenteric artery thrombosis accounts for about 60%-70%. Emboli mostly come from valvular heart disease, atrial fibrillation, myocardial infarction, bacterial endocarditis, etc., and the most common one is superior mesenteric artery thrombosis. The mesenteric vessels are composed of three branches from the abdominal aorta and the collateral circulation formed by them, including the celiac trunk (CA), superior mesenteric artery (SMA), inferior mesenteric artery (IMA) and mesenteric vein (MV). The SMA supplies the duodenum, jejunum, ileum, ascending colon, and the proximal 1 / 3 of the transverse colon. The IMA supplies the remaining transverse colon, descending colon, sigmoid colon, and proximal rectum. In addition to being the main blood vessel supplying the intestine, the SMA also establishes three important collateral circulations: (1) The inferior pancreaticoduodenal artery, superior pancreaticoduodenal artery, gastroduodenal artery, and celiac artery form a collateral circulation. (2) The middle colic artery, colic artery, ileocolic artery, and inferior mesenteric artery form a collateral circulation. (3) The distal end of the mesenteric artery sends out branches to form a collateral circulation with the posterior peritoneum and greater omentum. When the mesenteric vessels are blocked by more than 70%, the collateral circulation opens in large numbers. In addition, the mesenteric venous system is different from other veins and has a corresponding nutrient supply function. It collects venous blood from the intestine and returns it to the portal vein system. 25% to 30% of the blood supply to the liver comes from the hepatic artery, and 70% to 75% comes from the portal vein. The blood supply of the mesentery accounts for 15% to 20% of the cardiac output when fasting, and 35% after eating. Therefore, it is precisely because of the duality and particularity of the intestinal blood supply that obvious ischemia will only occur when the blood supply to the intestine is reduced by 50% and the mesenteric artery pressure is less than 45mmHg (1mmHg=0.133kPa). When the blood supply to the small intestine is less than 75%, it should be able to maintain for 12 hours. Therefore, under normal circumstances, the human intestine is not prone to ischemia. Once acute ischemia occurs, the consequences will be very serious if it is not corrected in time.

[0025] The occurrence of mesenteric artery thrombosis is also related to the anatomical structure of the superior mesenteric artery: the superior mesenteric artery branches out from the abdominal aorta at an acute angle, runs parallel to the aorta, has a thick lumen, and is consistent with the direction of blood flow in the abdominal aorta. It is easy for detached emboli to enter, causing vascular thrombosis at the narrow part or bifurcation of the blood vessel. It is more common in the part where the middle colic artery originates or below. The intestinal mucosa is not easy to tolerate ischemia. If the ischemia time exceeds 15 minutes, the villus structure of the small intestinal mucosa will be destroyed and fall off. Then the blood in the intestinal wall stagnates, becomes congested and edematous, the intestinal tube loses tension, cyanotic edema occurs, a large amount of plasma seeps into the intestinal wall, and the intestinal wall presents hemorrhagic necrosis. A large amount of plasma seeps into the abdominal cavity and intestinal cavity, the circulating blood volume decreases sharply, the bacteria in the intestinal cavity multiply in large numbers, and the toxic metabolites that occur due to necrosis caused by ischemia and hypoxia of the intestinal tube are continuously absorbed, leading to hypovolemia and toxic shock. When the intestine is necrotized, the intestinal tube expands, peristalsis disappears, and vascular intestinal obstruction is manifested.

[0026] The current preferred method for diagnosing mesenteric artery thrombosis (MAE) is CT angiography (CTA), which has a sensitivity of 93%, a specificity of 100%, a positive rate of 94%, and an exclusion rate of 100%. Ultrasound examination is also helpful in the diagnosis of this disease. Although mesenteric angiography was once considered the gold standard, it cannot be used as a routine option due to various factors and is only used during surgical intervention.

[0027] Generally, for patients with acute abdominal pain of unknown cause, if the patient has severe abdominal pain that is difficult to relieve with general drugs, and the symptoms are not proportional to the signs, mesenteric vascular disease should be highly suspected. At this time, a detailed medical history should be inquired and a careful physical examination should be performed. A detailed medical history combined with special symptoms and signs often helps in the preliminary diagnosis of mesenteric vascular disease. If the patient has a history of heart disease, there are severe gastrointestinal emptying symptoms, and the symptoms and signs are inconsistent, it is a triad for early diagnosis of acute superior mesenteric artery thrombosis. Acute superior mesenteric artery thrombosis is often secondary to existing chronic arterial stenosis, so there is often evidence of other arterial ischemia or a history of arterial surgery. Once acute mesenteric vascular disease occurs, severe abdominal pain may occur, which is difficult to relieve with general drugs, but there are no obvious abdominal signs, or only mild abdominal tenderness and hyperactive bowel sounds. This is the early stage of intestinal ischemia, which is the best time for surgical treatment, but it is difficult to distinguish from other acute abdomens. If laparotomy is performed, only pale, contracted or dilated intestinal tubes are often found, so it is most likely to be misdiagnosed and mistreated. As the disease progresses rapidly, physical signs gradually worsen, tenderness, rebound pain, and muscle tension appear. Intestinal smooth muscle is aggravated by ischemia and hypoxia, resulting in intestinal paralysis, which manifests as abdominal distension, significantly weakened or absent bowel sounds, and frequent vomiting, with vomitus often bloody, often indicating intestinal necrosis. Once necrosis occurs, a large amount of body fluid will soon seep into the intestinal cavity and peritoneal cavity, leading to severe water and electrolyte disorders and acid-base imbalance. At this time, the patient may experience toxic shock symptoms such as rapid heart rate, hypotension, and acidosis.

[0028] Laboratory tests for acute mesenteric vascular disease lack specificity and are not very sensitive. They are often manifested as increased white blood cells and neutrophils in the blood, and increased red blood cell counts, hematocrit, and blood amylase when the blood is concentrated. As the disease progresses, metabolic acidosis, electrolyte disorders, and enzymatic changes may also occur, but these changes are easily misdiagnosed as intra-abdominal infectious diseases in clinical practice. Among them, increased serum lactate levels and metabolic acidosis often indicate that the patient has intestinal obstruction, while increased plasma glutathione S-transferase isoenzymes often indicate intestinal ischemia. Plasma D-dimer is non-specific for acute intravascular thrombosis or thrombosis, but it is highly sensitive. If the D-dimer test result increases by more than 5 times, mesenteric vascular thrombosis or thrombotic disease should be highly suspected.

[0029] Early selection of appropriate imaging examinations often helps to quickly diagnose the disease. In the early stage of the disease, abdominal X-ray examinations may show no abnormalities, and occasionally small intestine and colon dilatation, rigid folds, and thickened, unchanged intestinal loops may be seen. As the disease progresses, gas accumulation in the intestinal cavity, liquid-gas levels, and free gas in the abdominal cavity may be shown. Although most patients have abnormal changes in abdominal X-rays, they lack specificity. Doppler ultrasound examination is a simple, effective, and inexpensive screening method that can not only detect the thickness of the intestinal wall and emboli in the arteries and veins, but also comprehensively observe the degree of vascular stenosis, the location and range of obstruction, and changes in hemodynamics, thereby determining the obstruction of the blood vessels and the degree of intestinal ischemia. However, some patients have intestinal flatulence that affects the results of ultrasound examinations. Therefore, if ultrasound exploration is unclear, CT examinations should be performed further. CT helps to differentiate from other acute abdomens, and can also show focal or segmental thickening of the intestinal wall, submucosal edema and bleeding, free gas in the abdominal cavity, gas in the mesenteric veins, and ascites. The diagnostic rate of enhanced CT and CT angiography (CTA) for mesenteric vascular diseases can be as high as over 90%. Digital subtraction angiography (DSA), as a diagnostic measure in vascular surgery, is currently the gold standard for diagnosing mesenteric vascular obstruction. However, due to its professionalism and invasiveness, and the fact that the entire process of mesenteric DSA from preparation to surgery may take several hours, which often delays the condition, many hospitals find it difficult to implement it in emergency situations. In addition, magnetic resonance angiography (MRA) can also be selected for those who are allergic to contrast agents. The sensitivity and specificity of MRA in diagnosing mesenteric vascular diseases are similar to those of CTA, but its disadvantage is that it cannot identify non-obstructive mesenteric ischemia and thrombosis of distal arteries.

[0030] For thrombus CT images with a single component, grayscale characteristics are often used as the entry point. By traversing the field around the pixel, the pixel value distribution law is found to find the edge of the lesion, which belongs to a binary classification task. At present, there are mainly two major methods in the field of image classification, one is the classification method based on statistical learning, and the other is the classification method based on deep learning. The classification method based on statistical learning often requires three steps of preprocessing, feature extraction and classification. The classification method based on deep learning is one of the research hotspots in the field of image classification in recent years, and the architecture of convolutional neural network (CNN) is mainly used.

[0031] The present invention is conducted to solve the above-mentioned problems, and aims to provide a method for rapid positioning and identification of acute superior mesenteric artery thrombosis based on deep learning, which specifically includes: Data collection and organization: According to the inclusion and exclusion indicators, clinical medical records of acute superior mesenteric artery thrombosis between January 1, 2015 and December 31, 2022 were found, relevant medical record data were downloaded and data preprocessing was completed. Inclusion criteria: ① Patients with pathologically confirmed acute superior mesenteric artery thrombosis; ② Patients with pre-treatment abdominal enhanced CT data. Exclusion criteria: ① Patients with poor image quality and no preoperative imaging data and unable to perform postoperative imaging evaluation. The processed data were divided into training set and test set.

[0032] The whole abdomen enhanced CT examination equipment used was Siemens Somatom Sensation 64 (Siemens, Forchheim); GE Discovery CT 750m (GE Healthcare, Princeton, NJ) or GE Light Speed ​​VCT (GE Healthcare, Princeton, NJ). The parameters of spiral CT were: tube voltage 120 kV, automatic milliampere-second technology (effective milliampere-second 210-250 mAs), pitch 1, rotation time 0.5 s, field of view size 350 mm × 350 mm, matrix 512 × 512, layer thickness 5 mm, spacing 5 mm. Patients were fasting for at least 3 hours for CT examination, and 500 ml of water was taken orally 5-10 minutes before scanning. Iodine non-ionic contrast agent (300-370 mg I / ml) 1-1.5 ml / kg was injected intravenously for enhanced scanning, and the flow rate was 3 ml / s. The arterial phase scan used bolus tracking technology, the monitoring plane was the level of the celiac trunk opening of the abdominal aorta, and the scan was performed 12 seconds after the trigger threshold (100Hu). The venous phase and delayed phase were delayed by 28 seconds and 50 seconds, respectively.

[0033] The preprocessing of image data automatically screens out image data with a layer thickness of exactly 5 mm. Image processing techniques such as Gaussian filtering are used to reduce noise interference, and the image is adjusted by setting the window width to 3200 (ranging from -3000 to 200) and the window level to -1400, so that the thrombus part is more conspicuous on the image. Deep learning models such as convolutional neural networks (CNNs) are deployed to perform automatic detection and segmentation of thrombi. The model can accurately identify the area of ​​thrombus, that is, the region of interest (ROI), and automatically create the corresponding mask file, providing a solid data foundation for further analysis. The system automatically marks the location of thrombus and gives a probability score to assist doctors in interpreting images more accurately. In order to ensure the accuracy and stability of the deep learning model, a large amount of annotated medical image data is used to train the model, and methods such as cross-validation are used to evaluate the performance of the model to ensure its effectiveness in practical applications. For the identified mesenteric artery thrombus site, the system automatically takes a screenshot and saves it in 512*512 pixel png format for more in-depth analysis and recording.

[0034] The AI ​​model building includes: 3D image reconstruction of abdominal artery tree: Find seed points, select the abdominal aorta as the seed point, construct a data set, use the Unet network for learning, and obtain a model (the existing method requires manual selection of seeds). The model accuracy can reach 95% (acc = the value of the pixel 1 in the test set × the training result at the corresponding coordinate / the value of the pixel 1 in the test set). That is, the training result can extract almost all the corresponding pixels of the abdominal aorta. After obtaining the mask value of the abdominal aorta, select the coordinates of the middle pixel as the seed point.

[0035] The growcut algorithm is used for 3D reconstruction. The growcut algorithm: each pixel is marked as foreground and background through growth iteration. The steps include: 1) Initialization: the seed point is marked as foreground and other pixels are marked as background; 2) Growth iteration: the algorithm considers the neighborhood of the pixel (26 neighborhoods) and expands the area based on the similarity between pixels. Similarity is usually calculated based on pixel value (color, grayscale, etc.) and spatial position; 3) Iteration convergence: no pixel label changes in a certain iteration, and the algorithm stops; 4) Result output: according to the final label, the image is segmented into foreground and background.

[0036] The growcut algorithm is a graph-based segmentation method in the field of 3D reconstruction, which is particularly suitable for processing images with complex boundaries and textures. The growcut algorithm shows significant advantages in the localization of thrombi in the superior mesenteric artery and the 3D reconstruction of the abdominal arterial tree. Its unsupervised characteristics effectively reduce manual intervention and significantly improve the degree of automation, which is particularly suitable for rapid diagnosis in emergency situations. Through the growth iteration method, the algorithm can effectively handle complex vascular structures, gradually iterate and improve the segmentation results, accurately depict the vascular network and locate thrombi. In addition, the growcut algorithm takes into account the noise and inhomogeneity in the image when designing, showing good robustness and enhancing the accuracy of segmentation, especially when the contrast between blood vessels and surrounding tissues is not high. Compared with other algorithms that require a lot of computing resources, growcut has high computational efficiency while maintaining high segmentation accuracy. It is suitable for clinical application scenarios that require rapid reconstruction and real-time feedback, such as real-time 3D reconstruction during surgery, so that doctors can make decisions quickly. It is worth mentioning that the growcut algorithm has good improvement and scalability, and can be adjusted and optimized according to specific application scenarios to improve the ability to identify thrombus areas. These technical advantages together promote the accuracy and efficiency of diagnosis, enabling the growcut algorithm to play an important role in clinical practice.

[0037] Calculate the centerline of the blood vessel and locate the end point of the superior mesenteric artery. Use the centerlinecomputation library of vtk to obtain the RAS coordinates of the end point from the table.

[0038] A model was trained to learn the LBP, Haralick and other features of normal blood vessels and thrombus blood vessels. The accuracy of the svm classifier was 0.766, and the accuracy of the random forest was 0.83. The subsequent image augmentation and network replacement were required to increase the training accuracy.

[0039] Starting from the current coordinate (x, y), the above model is used to determine whether the blood vessels in a series of nearby slices are thrombi. The judgment range is [[x-10, x+10], [y-10, y+10]].

[0040] The present invention provides a method for rapid positioning and identification of acute superior mesenteric artery vascular thrombosis based on deep learning, which has the following characteristics: training the medical record data based on a deep convolutional neural network algorithm according to medical record data to obtain rapid positioning of acute superior mesenteric artery vascular thrombosis, further training to obtain a fusion model, and saving the qualitative model and fusion model of acute superior mesenteric artery thrombosis as a callable automatic discrimination artificial intelligence model; calling the automatic discrimination artificial intelligence model, inputting the medical record data to be predicted for prediction, and obtaining the identification of acute superior mesenteric artery thrombosis.

[0041] The present invention incorporates abdominal enhanced CT imaging data into the analysis, combines multiple laboratory examination indicators, vital signs indicators and demographic information, and constructs a multi-dimensional machine learning model. This comprehensive data method can more comprehensively capture the influencing factors of acute superior mesenteric artery thrombosis, screen out predictive factors with strong correlation, thereby assisting clinicians to make more accurate diagnoses and provide a scientific basis for clinical prevention and treatment. A rapid positioning and identification method for acute superior mesenteric artery thrombosis based on a supervised machine learning classification algorithm was constructed, which can better locate and identify acute superior mesenteric artery thrombosis, help clinicians identify potential acute superior mesenteric artery thrombosis as early as possible, and ultimately achieve the purpose of early detection, early diagnosis, and early treatment. The indicator mIoU, which is not commonly used in diagnosis, was introduced as training data, and its technical effect was proved, which not only enhanced the predictive ability of the model, but also provided a new perspective for clinical diagnosis.

[0042] The following will be combined with the attached Figure 1 , a method for identifying vascular embolism based on deep learning provided by an embodiment of the present invention is described, comprising the following steps: 110. Extract a regional image from the abdominal medical image to be identified, where the regional image is an abdominal aorta blood vessel image.

[0043] 120. According to the abdominal aorta vascular image and the trained first model, the seed point of the abdominal medical image to be identified is obtained. The first model is constructed based on the Unet network and the Unet network is trained using the abdominal medical sample image set. The seed point is the pixel point in the abdominal aorta pixel in the medical image to be identified.

[0044] 130. The seed points are used to semi-automatically segment the mesenteric artery region in the regional image using the GrowCut algorithm to extract the image of the mesenteric artery region.

[0045] 140. The image of the mesenteric artery region is imaged using the VTK platform to delineate the centerline of the artery. The VTK platform is an application for three-dimensional computer graphics.

[0046] 150. The image of the mesenteric artery region demarcating the center line of the arterial vessel is input into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal vascular image characteristics and the thrombus vascular image characteristics.

[0047] Based on the above embodiment, step 120 specifically includes: 121. Input the abdominal aorta blood vessel image into the trained first model to obtain the corresponding pixel points of the abdominal aorta in the abdominal aorta blood vessel image.

[0048] 122. Determine the seed point of the abdominal medical image to be identified based on the corresponding pixel points of the abdominal aorta in the abdominal aorta vascular image.

[0049] Based on the above embodiment, step 122 specifically includes: The sliding window range is determined according to the position of each pixel point in the abdominal aorta vascular image.

[0050] Traverse the pixel points within the sliding window range and get the average value of the pixel values ​​within the sliding window range.

[0051] The original seed point in the abdominal aorta vascular image is determined according to the average value of the pixel values ​​within the range of all sliding windows.

[0052] Calculate the weighted distance between each pixel in the abdominal aorta vascular image and the original seed point.

[0053] The seed point of the abdominal medical image to be identified is determined according to the pixel value of each pixel point in the abdominal aorta blood vessel image, and the weighted distance, the maximum weighted distance and the minimum weighted distance between each pixel point and the original seed point.

[0054] Specifically, for example, when the pixel value of each pixel point in the abdominal aorta blood vessel image is greater than a preset pixel value, and the weighted distance between the pixel point and the original seed point is between the maximum weighted distance and the minimum weighted distance, the pixel point is determined to be a seed point.

[0055] Based on the above embodiment, step 130 specifically includes: 131. The seed points of the abdominal aorta blood vessel image are marked as foreground points, and the other pixel points are marked as background points.

[0056] 132. Obtain the neighborhood points of each foreground point, and calculate the similarity value between the foreground point and each neighborhood point, wherein the similarity value includes calculating the color similarity, gray value similarity and spatial position similarity between the foreground point and each neighborhood point.

[0057] 133. Expand the neighborhood points whose similarity values ​​are greater than a preset threshold into foreground points, and return to step 132.

[0058] 134. In this round of iteration, no neighboring points are expanded as foreground points. All foreground points in the abdominal aorta vascular image are extracted to obtain a foreground image, and the remaining pixels are the background image, where the foreground image is the image of the mesenteric artery area.

[0059] Based on the above embodiment, step 140 specifically includes: A CSV file of the image of the mesenteric artery region is extracted, and the RAS coordinates of the end points of the superior mesenteric artery in the CSV file are obtained. The RSA coordinates are right superior anterior coordinates used in the medical field.

[0060] Convert the RAS coordinates of the end point of the superior mesenteric artery into a data object in the VTK platform.

[0061] The center line of the superior mesenteric artery is calculated by using the polygonal data object of the superior mesenteric artery in the VTK platform and the data object of the end point of the superior mesenteric artery in the VTK platform.

[0062] Based on the above embodiment, it also includes: A sample image of a mesenteric artery region in which a centerline of an arterial vessel is delineated in an abdominal medical sample image set is acquired.

[0063] The local binary pattern feature vector of the sample image of the mesenteric artery region demarcating the arterial vessel centerline is extracted.

[0064] The gray-level co-occurrence matrix eigenvectors of the sample images of the mesenteric artery region that delineates the arterial vessel centerline are extracted.

[0065] The local binary pattern eigenvector and the gray-level co-occurrence matrix eigenvector are combined to obtain a comprehensive eigenvector.

[0066] The SVN classifier is trained using the comprehensive feature vector to obtain the second model.

[0067] Based on the above embodiment, it also includes: An original set of abdominal medical sample images is obtained, and images with a slice thickness of 5 mm are screened out from the original set of abdominal medical sample images.

[0068] The screened images were subjected to Gaussian filtering to reduce noise interference, and the screened images were adjusted by setting the window width to 3200 and the window level to -1400 to obtain an abdominal medical sample image set.

[0069] The present invention incorporates abdominal enhanced CT imaging data into the analysis, combines multiple laboratory examination indicators, vital signs indicators and demographic information, and constructs a multi-dimensional machine learning model. This comprehensive data method can more comprehensively capture the influencing factors of acute superior mesenteric artery thrombosis, screen out predictive factors with strong correlation, thereby assisting clinicians to make more accurate diagnoses and provide a scientific basis for clinical prevention and treatment. A rapid positioning and identification method for acute superior mesenteric artery thrombosis based on a supervised machine learning classification algorithm was constructed, which can better locate and identify acute superior mesenteric artery thrombosis, help clinicians identify potential acute superior mesenteric artery thrombosis as early as possible, and ultimately achieve the purpose of early detection, early diagnosis, and early treatment. The indicator mIoU, which is not commonly used in diagnosis, was introduced as training data, and its technical effect was proved, which not only enhanced the predictive ability of the model, but also provided a new perspective for clinical diagnosis.

[0070] like Figure 2 As shown, the present invention also provides a vascular embolism identification device based on deep learning, the device comprising: The first processing module is used to extract a regional image from the abdominal medical image to be identified, where the regional image is an abdominal aorta blood vessel image.

[0071] The second processing module is used to obtain the seed points of the abdominal medical image to be identified based on the abdominal aorta vascular image and the trained first model. The first model is constructed based on the Unet network and the Unet network is trained using the abdominal medical sample image set. The seed points are the pixels in the abdominal aorta pixels in the medical image to be identified.

[0072] The third processing module is used to use the GrowCut algorithm to semi-automatically segment the mesenteric artery region in the regional image using the seed points, and extract the image of the mesenteric artery region.

[0073] A fourth processing module is used for delineating the center line of the artery in the image of the mesenteric artery region using the VTK platform, wherein the VTK platform is an application for three-dimensional computer graphics; The fifth processing module is used to input the image of the mesenteric artery area that delineates the center line of the arterial blood vessel into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal blood vessel image characteristics and the thrombus blood vessel image characteristics.

[0074] Based on the above embodiment, the second processing module is specifically used to input the abdominal aorta blood vessel image into the trained first model to obtain the corresponding pixel points of the abdominal aorta in the abdominal aorta blood vessel image.

[0075] Based on the corresponding pixel points of the abdominal aorta in the abdominal aorta vascular image, the seed points of the abdominal medical image to be identified are determined.

[0076] Based on the above embodiment, the second processing module is specifically configured to determine the sliding window range according to the position of each pixel point in the abdominal aorta blood vessel image.

[0077] Traverse the pixel points within the sliding window range and get the average value of the pixel values ​​within the sliding window range.

[0078] The original seed point in the abdominal aorta vascular image is determined according to the average value of the pixel values ​​within the range of all sliding windows.

[0079] Calculate the weighted distance between each pixel in the abdominal aorta vascular image and the original seed point.

[0080] The seed point of the abdominal medical image to be identified is determined according to the pixel value of each pixel point in the abdominal aorta blood vessel image, and the weighted distance, the maximum weighted distance and the minimum weighted distance between each pixel point and the original seed point.

[0081] Based on the above embodiment, the third processing module includes a first subunit, a second subunit, a third subunit, and a fourth subunit; The first subunit is used to mark the seed point of the abdominal aorta blood vessel image as a foreground point, and mark other pixel points as background points.

[0082] The second subunit is used to obtain the neighborhood points of each foreground point and calculate the similarity value between the foreground point and each neighborhood point, wherein the similarity value includes calculating the color similarity, gray value similarity and spatial position similarity between the foreground point and each neighborhood point.

[0083] The third subunit is used to expand the neighborhood points whose similarity values ​​are greater than a preset threshold into foreground points and return to the second subunit.

[0084] The fourth subunit is used for extracting all foreground points in the abdominal aorta vascular image when no neighboring points are expanded as foreground points in this round of iteration to obtain a foreground image, and the remaining pixels are the background image, wherein the foreground image is the image of the mesenteric artery area.

[0085] Based on the above embodiment, the fourth processing module is specifically used to extract the CSV file of the image of the mesenteric artery region, and the RAS coordinates of the end point of the superior mesenteric artery in the CSV file. The RSA coordinates are the upper right anterior coordinates used in the medical field.

[0086] Convert the RAS coordinates of the end point of the superior mesenteric artery into a data object in the VTK platform.

[0087] The center line of the superior mesenteric artery is calculated by using the polygonal data object of the superior mesenteric artery in the VTK platform and the data object of the end point of the superior mesenteric artery in the VTK platform.

[0088] Based on the above embodiment, a sixth processing module is also included, and the sixth processing module is used to obtain a sample image of the mesenteric artery area in which the center line of the arterial blood vessel is delineated in the abdominal medical sample image set.

[0089] The local binary pattern feature vector of the sample image of the mesenteric artery region demarcating the arterial vessel centerline is extracted.

[0090] The gray-level co-occurrence matrix eigenvectors of the sample images of the mesenteric artery region that delineates the arterial vessel centerline are extracted.

[0091] The local binary pattern eigenvector and the gray-level co-occurrence matrix eigenvector are combined to obtain a comprehensive eigenvector.

[0092] The SVN classifier is trained using the comprehensive feature vector to obtain the second model.

[0093] Based on the above embodiment, the sixth processing module is specifically used to obtain an original set of abdominal medical sample images, and screen out images with a layer thickness of 5 mm from the original set of abdominal medical sample images.

[0094] The screened images were subjected to Gaussian filtering to reduce noise interference, and the screened images were adjusted by setting the window width to 3200 and the window level to -1400 to obtain an abdominal medical sample image set.

[0095] The present invention incorporates abdominal enhanced CT imaging data into the analysis, combines multiple laboratory examination indicators, vital signs indicators and demographic information, and constructs a multi-dimensional machine learning model. This comprehensive data method can more comprehensively capture the influencing factors of acute superior mesenteric artery thrombosis, screen out predictive factors with strong correlation, thereby assisting clinicians to make more accurate diagnoses and provide a scientific basis for clinical prevention and treatment. A rapid positioning and identification method for acute superior mesenteric artery thrombosis based on a supervised machine learning classification algorithm was constructed, which can better locate and identify acute superior mesenteric artery thrombosis, help clinicians identify potential acute superior mesenteric artery thrombosis as early as possible, and ultimately achieve the purpose of early detection, early diagnosis, and early treatment. The indicator mIoU, which is not commonly used in diagnosis, was introduced as training data, and its technical effect was proved, which not only enhanced the predictive ability of the model, but also provided a new perspective for clinical diagnosis.

[0096] In addition, an embodiment of the present invention includes a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, a deep learning-based vascular embolism identification method described in any one of the above technical solutions is implemented.

[0097] An embodiment of the present invention also includes a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for identifying vascular embolism based on deep learning described in any one of the above technical solutions is implemented.

[0098] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for identifying vascular embolism based on deep learning, characterized in that: The method comprises: Extracting a regional image from an abdominal medical image to be identified, wherein the regional image is an abdominal aorta blood vessel image; Obtaining a seed point of the to-be-identified abdominal medical image according to the abdominal aorta vascular image and the trained first model, wherein the first model is constructed based on a Unet network and is obtained by training the Unet network using an abdominal medical sample image set, and the seed point is a pixel point in a pixel of the abdominal aorta in the to-be-identified abdominal medical image; Using the seed points to semi-automatically segment the mesenteric artery region in the regional image using the GrowCut algorithm, and extracting the image of the mesenteric artery region; Delineating the centerline of the artery in the image of the mesenteric artery region using the VTK platform, wherein the VTK platform is an application for three-dimensional computer graphics; The image of the mesenteric artery area with the arterial centerline delineated is input into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal vascular image characteristics and the thrombus vascular image characteristics.

2. The method according to claim 1, characterized in that The step of obtaining the seed point of the to-be-identified abdominal medical image according to the abdominal aorta vascular image and the trained first model specifically includes: Inputting the abdominal aorta blood vessel image into the trained first model to obtain the corresponding pixel points of the abdominal aorta in the abdominal aorta blood vessel image; Based on the corresponding pixel points of the abdominal aorta in the abdominal aorta vascular image, the seed points of the abdominal medical image to be identified are determined.

3. The method according to claim 2, characterized in that The step of determining the seed point of the abdominal medical image to be identified based on the pixel point corresponding to the abdominal aorta in the abdominal aorta vascular image specifically includes: Determining a sliding window range according to the position of each pixel point in the abdominal aorta blood vessel image; Traversing the pixel points within the sliding window to obtain an average value of the pixel values ​​within the sliding window; Determine an original seed point in the abdominal aorta blood vessel image according to an average value of all pixel values ​​within the sliding window; Calculating the weighted distance between each pixel point in the abdominal aorta blood vessel image and the original seed point; The seed point of the abdominal medical image to be identified is determined according to the pixel value of each pixel point in the abdominal aorta blood vessel image, and the weighted distance, the maximum weighted distance and the minimum weighted distance between each pixel point and the original seed point.

4. The method according to claim 1, characterized in that The step of using the seed points to semi-automatically segment the mesenteric artery region in the regional image using the GrowCut algorithm to extract the image of the mesenteric artery region specifically includes: S1, marking the seed point of the abdominal aorta blood vessel image as a foreground point, and marking other pixel points as background points; S2, obtaining the neighborhood points of each foreground point, and calculating the similarity value between the foreground point and each neighborhood point, wherein the similarity value includes calculating the color similarity, gray value similarity and spatial position similarity between the foreground point and each of the neighborhood points; S3, expanding the neighborhood points whose similarity values ​​are greater than a preset threshold as the foreground points, and returning to step S2; S4. In this round of iteration, no neighboring points are expanded to foreground points, and all the foreground points in the abdominal aorta vascular image are extracted to obtain a foreground image, and the remaining pixels are the background image, wherein the foreground image is an image of the mesenteric artery region.

5. The method according to claim 1, characterized in that The image of the mesenteric artery region is imaged using the VTK platform to delineate the arterial blood vessel centerline, wherein the VTK platform is an application for three-dimensional computer graphics, specifically including: Extracting a CSV file of the image of the mesenteric artery region, and obtaining the RAS coordinates of the end point of the superior mesenteric artery in the CSV file, where the RSA coordinates are right upper anterior coordinates used in the medical field; Convert the RAS coordinates of the end point of the superior mesenteric artery into a data object in the VTK platform; The centerline of the artery is calculated using the polygonal data object of the superior mesenteric artery in the VTK platform and the data object of the end point of the superior mesenteric artery in the VTK platform.

6. The method according to claim 1, characterized in that The method further comprises: Acquire a sample image of the mesenteric artery region where the arterial blood vessel centerline is delineated in the abdominal medical sample image set; Extracting a local binary pattern feature vector of the sample image of the mesenteric artery region demarcating the arterial vessel centerline; Extracting gray-level co-occurrence matrix feature vectors of sample images of the mesenteric artery region demarcating the arterial vessel centerline; Combining the local binary pattern feature vector and the gray level co-occurrence matrix feature vector to obtain a comprehensive feature vector; The SVN classifier is trained using the comprehensive feature vector to obtain the second model.

7. The method according to claim 1, characterized in that The method further comprises: Acquire an original set of abdominal medical sample images, and screen out images with a slice thickness of 5 mm from the original set of abdominal medical sample images; The screened images were subjected to Gaussian filtering to reduce noise interference, and the screened images were adjusted by setting the window width to 3200 and the window level to -1400 to obtain the abdominal medical sample image set.

8. A deep learning-based vascular embolism identification device, characterized in that: The device comprises: A first processing module is used to extract a regional image from the abdominal medical image to be identified, wherein the regional image is an abdominal aorta blood vessel image; A second processing module is used to obtain the seed point of the abdominal medical image to be identified according to the abdominal aorta vascular image and the trained first model, wherein the first model is constructed based on a Unet network and the Unet network is trained using an abdominal medical sample image set, and the seed point is a pixel point in the abdominal aorta pixel in the abdominal medical image to be identified; A third processing module is used to use the GrowCut algorithm to semi-automatically segment the mesenteric artery region in the regional image using the seed point to extract the image of the mesenteric artery region; A fourth processing module is used to delineate the centerline of the artery in the image of the mesenteric artery region using a VTK platform, wherein the VTK platform is an application for three-dimensional computer graphics; The fifth processing module is used to input the image of the mesenteric artery area delineating the center line of the arterial blood vessel into the second model to determine whether there is a thrombus and the location of the thrombus in the abdominal medical image to be identified. The second model is established based on the SVN classifier and is used to determine whether there is a thrombus based on the normal blood vessel image characteristics and the thrombus blood vessel image characteristics.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the deep learning-based vascular embolism identification method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the deep learning-based vascular embolism identification method according to any one of claims 1 to 7 is implemented.

Images