Construction method of CABG postoperative bridge vessel reocclusion condition prediction model

Abstract

The invention discloses a construction method of a CABG postoperative bridge vessel reocclusion condition prediction model, which effectively achieves the purpose of assisting medical personnel in performing risk stratification and prognosis prediction on a CABG postoperative patient, and comprises the following steps: S1, obtaining a coronary image to be marked; s2, segmenting the coronary artery image by using a VB-net network; s3, using a point cloud segmentation and segmentation technology of a graph convolutional neural network to obtain a coronary artery segmentation model; s4, screening related variables of bridge vessel reocclusion after CABG operation by using regression analysis; step S5, constructing a CABG postoperative bridge vessel reocclusion prediction model; and step S6, constructing a column diagram to display the performance of each prediction model in the step S5, the structure is novel, the concept is ingenious, the operation is simple and convenient, high-risk patients can be found as early as possible, intervention measures can be taken in time, and the accuracy and individualization level of diagnosis and treatment effects on the patients are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of coronary image processing and relates to a method for constructing a prediction model for reocclusion of a bridge vessel after CABG surgery. Background Art

[0002] Coronary artery disease is one of the leading causes of death worldwide. As my country's aging population continues to deepen, the incidence of cardiovascular diseases such as coronary artery disease in the elderly has also increased year by year. According to statistics, the number of patients with cardiovascular diseases accounts for more than one-third of all patients. Coronary artery bypass grafting (CABG) is an effective method for treating coronary heart disease and is also the preferred revascularization strategy for patients with severe multivessel disease. In 2022, more than 1.3 million patients nationwide underwent CABG surgery, and patients in Henan Province accounted for about 10%. Although surgical technology has made progress, due to the complexity of coronary artery disease, the long-term prognosis of patients after CABG surgery is still poor. Previous studies have shown that postoperative graft occlusion of CABG is the main factor affecting the long-term prognosis of patients. Therefore, for patients undergoing CABG, it is crucial to pay attention to the condition of graft lesions and identify reliable influencing factors.

[0003] In recent years, deep learning has been widely used in biomedical research. Compared with other methods, deep learning is more flexible, more general, and can be applied to discrete or continuous data. In the cardiovascular field, the accuracy of high-risk plaque diagnosis can be significantly improved by combining the extraction of radiomic features with machine learning. At present, the accuracy of machine learning in identifying the napkin ring sign in high-risk plaques is very high. In addition, the use of deep learning for the segmentation and reconstruction of chronic coronary artery occlusions can not only help distinguish chronic total occlusions from subtotal occlusions, but also save a lot of manual reconstruction time. However, the prediction of long-term reocclusion of graft vessels in patients after CABG based on CCTA images is still a research gap at home and abroad.

[0004] The publication number CN111369542B was retrieved, and the patent name is invention patent of blood vessel labeling method, image processing system and storage medium. It discloses a blood vessel labeling method, but its research on the prediction of graft vessel reocclusion is still insufficient. Therefore, a method for constructing a prediction model of graft vessel reocclusion after CABG surgery is needed to solve the above problem. Summary of the invention

[0005] In view of the above problems, the present invention proposes a method for constructing a prediction model for the reocclusion of the graft vessel after CABG surgery, which effectively solves the problems in the prior art.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] A method for constructing a prediction model for reocclusion of a graft vessel after CABG surgery comprises the following steps:

[0008] Step S1, obtaining a coronary artery image to be marked;

[0009] Step S2, using the VB-net network to segment the vascular tree and each chamber of the heart in the coronary image, and removing false positive segmentation results in non-vascular areas;

[0010] Step S3, using the point cloud segmentation and segmentation technology of the graph convolutional neural network, integrating the spatial position information of the coronary heart and the semantic information of the blood vessels in the original image, obtaining a coronary segmentation model, and realizing the accurate segmentation of the 18 segments of the coronary vascular tree and the segmentation of the bridge vessels;

[0011] Step S4, using univariate logistics regression analysis and multivariate logistics regression analysis to screen variables related to graft reocclusion after CABG surgery;

[0012] Step S5, constructing a prediction model for graft reocclusion after CABG surgery, using the variables obtained in step S4 to construct a simple clinical indicator model, a clinical indicator + graft anatomical parameter model, and a clinical indicator + graft anatomical parameter + hemodynamic parameter model, and using ROC curves and Delong analysis to compare the performance differences between different models;

[0013] Step S6: construct a nomogram to show the performance of each prediction model in step S5.

[0014] Preferably, in step S4, the relevant variables before screening include image-related parameters and patient information.

[0015] Preferably, the image-related parameters are obtained based on the coronary segmentation model measurement in step S3, and the image-related parameters include: contrast attenuation gradient within the lumen of the bridge vessel, pericoronary fat attenuation index, bridge vessel length, anastomotic diameter, proximal anastomotic diameter, and distal anastomotic diameter.

[0016] Preferably, the patient information is obtained by retrospectively reviewing electronic medical records, and the patient information includes: age, gender, previous medical history, and laboratory tests.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. The present invention screened the variables related to the reocclusion of the graft vessel after CABG surgery, obtained the predictive factors related to the anastomotic occlusion, and further obtained the factors closely related to the reocclusion of the graft vessel after CABG surgery.

[0019] 2. The present invention completes the construction of a prediction model through regression analysis, which is helpful for risk stratification and prognosis prediction of patients after CABG, and helps to identify high-risk patients early and take timely intervention measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the process of coronary artery segmentation and segmentation of the present invention.

[0021] Figure 2 The figure shows the ROC results of the three models in step S5 of the present invention.

[0022] Figure 3 It is a display diagram of the Delong results of the three models in step S5 of the present invention.

[0023] Figure 4 This is a display diagram of the result of the nomogram in step S6 of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] The following is combined with Figure 1-Figure 4 The specific implementation modes of the present invention are described in further detail.

[0026] The present invention comprises the following steps:

[0027] Step S1, obtaining a coronary artery image to be marked;

[0028] It should be noted that, in step S1, the coronary image to be marked is obtained by scanning the subject with a medical imaging device, and the medical image should contain a coronary vascular area, such as a CT angiography image (CTA);

[0029] Step S2, using the VB-net network to segment the vascular tree and each chamber of the heart in the coronary image, and removing false positive segmentation results in non-vascular areas;

[0030] It should be noted that while using VB-net developed based on the U-Net network to achieve segmentation of the coronary CTA vascular tree, the above-mentioned segmentation model is used to segment each chamber of the heart, and based on the anatomical distribution characteristics of the coronary arteries, a reliable annular coronary artery area is provided to remove false positive segmentation results of non-vascular areas; further, the training method for segmentation of the VB-net network may include: obtaining a certain number of training sample images, and having experienced doctors annotate them to obtain a segmentation standard; it should be noted that the standard is determined by the doctor; then the training sample images are input into the initial vascular segmentation network to obtain a predicted vascular segmentation result; the loss between the predicted vascular segmentation result and the segmentation standard is calculated, and the network parameters of the initial vascular segmentation network are adjusted using the loss, thereby iteratively training until the network converges;

[0031] Step S3, using the point cloud segmentation and segmentation technology of the graph convolutional neural network, integrating the spatial position information of the coronary heart and the semantic information of the blood vessels in the original image, obtaining a coronary segmentation model, and realizing the accurate segmentation of the 18 segments of the coronary vascular tree and the segmentation of the bridge vessels;

[0032] It should be noted that the point cloud is a set of points that express the target space under the same spatial reference system. The multi-label segmentation task of the vascular segmentation result is converted into a classification task based on a point set. After obtaining the vascular segmentation result, the computer device can perform point clouding on the vascular segmentation result, that is, perform point clouding on the above-mentioned binary mask image, and input the point clouded vascular segmentation result into a preset point cloud network to obtain a coronary artery segmentation model.

[0033] Step S4, using univariate logistics regression analysis and multivariate logistics regression analysis to screen variables related to graft reocclusion after CABG surgery;

[0034] The relevant variables before screening include imaging-related parameters and patient information;

[0035] The image-related parameters are obtained based on the coronary segmentation model measurement in step S3, and the image-related parameters include: contrast attenuation gradient in the lumen of the graft vessel, fat attenuation index around the coronary artery, length of the graft vessel, diameter of the anastomosis, proximal diameter of the anastomosis, and distal diameter of the anastomosis;

[0036] It should be noted that logistic regression is a binary or multi-classification probabilistic nonlinear regression model used to study the relationship between dependent variables and influencing factors. Unlike linear regression, the purpose of logistics regression is to find the best fitting parameters so as to assign different weights to different features. In order to evaluate the predictive factors related to anastomotic occlusion, SPSS Statistics 25.0 software (SPSS Inc., Chicago, IL) was used in this embodiment, and "anastomotic occlusion" was used as the dependent variable, and the patient's basic information (age, gender, past medical history), laboratory tests (blood biochemistry, blood routine), and anatomical characteristics of the graft (graft length, anastomotic diameter, proximal anastomotic diameter, and distal anastomotic diameter) were used as independent variables to establish a model and observe the influencing factors of the model;

[0037] The patient information was obtained by retrospectively reviewing the electronic medical records, and the patient information included: age, gender, previous medical history, and laboratory tests;

[0038] In one embodiment, after screening by univariate logistics regression analysis and multivariate logistics regression analysis, it was found that the anatomical characteristics of the graft vessel (graft vessel length, distal diameter of the anastomosis), hemodynamic parameters (TAG), and laboratory test parameters (triglycerides, low-density lipoprotein cholesterol, and glycosylated hemoglobin) were closely related to the reocclusion of the graft vessel after CABG surgery;

[0039] Step S5: Figure 2 , Figure 3 It is given that, in order to construct a prediction model for graft reocclusion after CABG surgery, the variables obtained in step S4 are used to construct a simple clinical indicator model (Model 1), a clinical indicator + graft anatomical parameter model (Model 2), and a clinical indicator + graft anatomical parameter + hemodynamic parameter model (Model 3), and the ROC curve and Delong analysis are used to compare the performance differences between different models;

[0040] Step S6: Figure 4 Given, a nomogram is constructed to show the performance of each prediction model in step S5.

[0041] The present invention has a novel structure, ingenious conception, and simple and convenient operation. Through this design, it can effectively assist medical staff in risk stratification and prognosis prediction for patients after CABG surgery, help to detect high-risk patients early and take intervention measures in time, and improve the accuracy and individualization of diagnosis and treatment effects on patients.

[0042] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a prediction model for reocclusion of graft vessels after CABG surgery, characterized in that: The following steps are involved: Step S1, obtaining a coronary artery image to be marked; Step S2, using the VB-net network to segment the vascular tree and each chamber of the heart in the coronary image, and removing false positive segmentation results in non-vascular areas; Step S3, using the point cloud segmentation and segmentation technology of the graph convolutional neural network, integrating the spatial position information of the coronary heart and the semantic information of the blood vessels in the original image, obtaining a coronary segmentation model, and realizing the accurate segmentation of the 18 segments of the coronary vascular tree and the segmentation of the bridge vessels; Step S4, using univariate logistics regression analysis and multivariate logistics regression analysis to screen variables related to graft reocclusion after CABG surgery; Step S5, constructing a prediction model for graft reocclusion after CABG surgery, using the variables obtained in step S4 to construct a simple clinical indicator model, a clinical indicator + graft anatomical parameter model, and a clinical indicator + graft anatomical parameter + hemodynamic parameter model, and using ROC curves and Delong analysis to compare the performance differences between different models; Step S6: construct a nomogram to show the performance of each prediction model in step S5.

2. The method for constructing a prediction model for reocclusion of graft vessels after CABG surgery according to claim 1, characterized in that: In step S4, the relevant variables before screening include image-related parameters and patient information.

3. The method for constructing a prediction model for reocclusion of graft vessels after CABG surgery according to claim 2, characterized in that: The image-related parameters are measured based on the coronary segmentation model in step S3, and the image-related parameters include: contrast attenuation gradient within the lumen of the bridge vessel, pericoronary fat attenuation index, bridge vessel length, anastomotic diameter, proximal anastomotic diameter, and distal anastomotic diameter.

4. The method for constructing a prediction model for reocclusion of graft vessels after CABG surgery according to claim 2, characterized in that: The patient information was obtained by retrospectively reviewing electronic medical records, and the patient information included: age, gender, previous medical history, and laboratory tests.

Images