New coronary plaque prediction method and system based on coronary artery FAI

Abstract

The invention discloses a coronary artery FAI-based new coronary plaque prediction method and system. The new coronary plaque prediction method comprises the following steps: a scanning step: scanning an area from the lower part of tracheal carina to the upper part of diaphragm muscle of a testee through at least 64 layers of CT equipment to obtain a CTA scanning image; a PCAT range obtaining step: obtaining a PCAT range map of the epicardial coronary artery blood vessel through the CTA scanning image; a parameter acquisition step: calculating a PCAT volume value and an FAI value corresponding to the PCAT range map in the CTA scanning image; and a prediction step: obtaining a prediction result through the constructed new coronary plaque prediction model according to the PCAT volume value and the FAI value.

Description

Technical Field

[0001] The present invention relates to the technical field of novel coronary artery vascular risk marker detection, and more specifically, to a method and system for predicting new coronary artery plaques based on coronary artery FAI for predicting new coronary artery plaques. Background Art

[0002] Coronary atherosclerosis is a common cardiovascular disease. It is the manifestation of atherosclerosis in the coronary arteries and is the main cause of coronary artery disease. The development of this disease involves multiple factors, including genetics, metabolism, lifestyle, etc. The final result is obstruction of coronary blood flow, which may lead to serious consequences such as myocardial ischemia, angina pectoris, myocardial infarction, etc.

[0003] The formation of coronary atherosclerosis is a complex process, which is mainly characterized by lipid accumulation, inflammatory response, smooth muscle cell proliferation and fibrosis in the arterial wall. The pathophysiological process can be roughly divided into the following stages: Endothelial dysfunction: The function of coronary endothelial cells is impaired under the influence of damaging factors such as hypertension, hyperlipidemia, and tobacco smoke, resulting in increased permeability of the vascular wall, and the entry of pro-inflammatory cells and lipids into the vascular wall; Lipid deposition: Lipids such as low-density lipoprotein (LDL) accumulate and oxidize in the vascular wall, stimulating inflammatory cells (such as macrophages) to gather and phagocytize oxidized lipids to form foam cells; Inflammatory response: The formation of foam cells further promotes the aggregation of inflammatory cells and the release of inflammatory factors, exacerbating local inflammatory response; Fibrous plaque formation: Smooth muscle cells migrate to the plaque area, proliferate and produce matrix proteins such as collagen, forming a fibrous cap covering the surface of the plaque.

[0004] Early detection and intervention of coronary artery plaques can effectively reduce mortality and improve patient prognosis. Current conventional detection methods include electrocardiogram, exercise tolerance test, cardiac radionuclide scanning, intravascular ultrasound, coronary angiography, etc. However, these examination methods are based on morphological or functional examinations of coronary artery plaques that have already occurred, and cannot predict whether subjects who have not developed coronary artery plaques will develop coronary artery plaques in the future.

[0005] Therefore, after intensive research, the inventors have developed a method and system for predicting new coronary plaques based on coronary artery FAI that overcomes the above-mentioned defects. The aim is to construct a method and system for predicting new plaques in people at high risk of coronary heart disease, to conduct more effective and accurate assessments of such people, and to improve prognosis. Summary of the invention

[0006] In view of the above problems, the present invention provides a method for predicting new coronary plaque based on coronary artery FAI, which includes:

[0007] Scanning steps: Use a CT device with at least 64 slices to scan the area from below the tracheal carina to above the diaphragm of the subject to obtain a CTA scan image;

[0008] PCAT range acquisition step: obtaining a PCAT range map of the epicardial coronary artery through the CTA scan image;

[0009] Parameter acquisition step: calculating the PCAT volume value and FAI value corresponding to the PCAT range map in the CTA scan image;

[0010] Prediction step: obtaining a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

[0011] In the above-mentioned method for predicting newly-occurring coronary plaque, the scanning step comprises:

[0012] Retrospective ECG gating was used, with tube voltage setting range of 120kV-135kV, tube current setting range of 350mA-450mA, slice thickness setting range of 0.5mm-0.625mm, circle speed of 0.28s / circle or 0.35s / circle. A two-way high-pressure syringe was used to inject 50ml-70ml of non-ionic contrast agent at a flow rate of 4.0ml / s-5.0ml / s through the subject's antecubital vein indwelling needle, followed by injection of 30ml of normal saline at the same flow rate before scanning. The scanning range was 1.0cm below the tracheal carina to above the diaphragm.

[0013] In the above-mentioned method for predicting newly-occurring coronary plaque, the PCAT range acquisition step comprises:

[0014] Intercepting the CTA scan image according to a preset range to obtain the PCAT range map;

[0015] The preset range is the proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded;

[0016] The PCAT range map includes: a right coronary artery PCAT range map, an anterior descending branch PCAT range map and a circumflex branch PCAT range map.

[0017] In the above-mentioned method for predicting newly-occurring coronary plaque, the parameter acquisition step comprises:

[0018] Volume value acquisition step: calculating and obtaining the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map;

[0019] FAI value acquisition step: calculating the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map;

[0020] FAI value processing step: obtaining the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

[0021] In the above-mentioned method for predicting newly-developed coronary plaque, the volume value acquisition step and the FAI value acquisition step both include:

[0022] The PCAT volume value and the FAI value are calculated by acquiring the CT value of each pixel of the right coronary artery PCAT range map, the anterior descending branch PCAT range map, and the circumflex branch PCAT range map.

[0023] In the above-mentioned method for predicting newly-occurring coronary plaque, the prediction step comprises:

[0024] Information acquisition step: acquiring status information of multiple patients and processing the status information to obtain a training set and a test set;

[0025] Model construction step: construct multiple initial prediction models based on the logistics regression model, decision tree model, random forest model and XGBoost model, and respectively train and test the multiple initial prediction models through the training set and the test set to obtain test results;

[0026] Model selection step: calculating the test results of each of the initial prediction models to obtain the corresponding test accuracy, and selecting the initial prediction model corresponding to the highest test accuracy as the new coronary plaque prediction model;

[0027] Prediction result acquisition step: obtaining the prediction result through the new coronary plaque prediction model according to the PCAT volume value and the FAI value of the subject.

[0028] In the above-mentioned method for predicting new coronary plaques, the status information includes: age, gender, BMI, interval between two coronary artery CTA examinations, diabetes, hypertension, smoking history, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, and FAI value.

[0029] In the above-mentioned method for predicting newly-occurring coronary plaque, the model selection step comprises:

[0030] Obtaining the prediction result of whether each patient has coronary artery plaque by using each of the initial prediction models according to the test set;

[0031] Calculating the prediction results of the plurality of patients obtained by each of the initial prediction models to obtain the test accuracy rate corresponding to the initial prediction model;

[0032] Sorting the multiple test accuracy rates;

[0033] The initial prediction model corresponding to the highest test accuracy is selected as the new coronary plaque prediction model.

[0034] The present invention also provides a new coronary plaque prediction system based on coronary artery FAI, wherein the new coronary plaque prediction method described above is applied, and the new coronary plaque prediction system comprises:

[0035] A scanning unit, which scans the area from below the tracheal carina to above the diaphragm of the subject by means of a CT device with at least 64 slices to obtain a CTA scanning image;

[0036] A PCAT range acquisition unit, which obtains a PCAT range map of the epicardial coronary artery through the CTA scan image;

[0037] A parameter acquisition unit, which calculates a PCAT volume value and a FAI value corresponding to the PCAT range map in the CTA scan image;

[0038] A prediction unit obtains a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

[0039] In the above-mentioned new coronary plaque prediction system, the PCAT range acquisition unit includes: intercepting the CTA scan image according to a preset range to obtain the PCAT range map; wherein the preset range is the proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded; the PCAT range map includes: the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map;

[0040] The parameter acquisition unit comprises:

[0041] A volume value acquisition module, which calculates and obtains the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map;

[0042] An FAI value acquisition module, which calculates and obtains the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map;

[0043] The FAI value processing module obtains the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

[0044] Compared with the prior art, the effectiveness of the present invention lies in that, in view of the current lack of evidence-based medical basis, the present invention can predict whether the current high-risk population for coronary heart disease will develop coronary atherosclerosis, and conduct real-time arteriosclerosis monitoring on the high-risk population for coronary heart disease, thereby conducting a more effective and accurate assessment of such population to improve prognosis.

[0045] Other features and advantages of the present invention will be described in the following description, and partly become obvious from the description, or be understood by implementing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 is a flow chart of the method for predicting new coronary plaque of the present invention;

[0048] Figure 2 for Figure 1 A step-by-step flow chart of step S3;

[0049] Figure 3 for Figure 1 The step-by-step flow chart of step S4;

[0050] Figure 4 It is a schematic diagram of the structure of the new coronary plaque prediction system of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0052] The exemplary embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention. In addition, the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0053] The terms “first”, “second”, “S1”, “S2”, etc. used in this document do not specifically refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0054] The directional terms used in this document, such as up, down, left, right, front or back, etc., are only used to refer to the directions of the drawings. Therefore, the directional terms used are used to illustrate and not to limit the present invention.

[0055] The words “include,” “including,” “have,” “contain,” etc. used in this document are open-ended terms, meaning including but not limited to.

[0056] As used herein, "and / or" includes any or all combinations of the items described.

[0057] As used herein, “plurality” includes “two” and “more than two”; as used herein, “plurality groups” includes “two groups” and “more than two groups”.

[0058] The terms "substantially" and "approximately" used herein are used to modify any quantity or error that may vary slightly, but these slight changes or errors do not change their essence. Generally speaking, the range of slight changes or errors modified by such terms may be 20% in some embodiments, 10% in some embodiments, 5% in some embodiments, or other values. Those skilled in the art should understand that the aforementioned values ​​can be adjusted according to actual needs and are not limited thereto.

[0059] Certain terms used to describe the present application are discussed below or elsewhere in this specification to provide additional guidance to those skilled in the art regarding the description of the present application.

[0060] Please refer to Figure 1 , Figure 1 FIG. 4 is a flow chart of the method for predicting new coronary plaque of the present invention. Figure 1 As shown, a method for predicting new coronary plaque based on coronary artery FAI of the present invention includes:

[0061] Scanning step S1: scanning the area from below the tracheal carina to above the diaphragm of the subject by means of a CT device having at least 64 slices to obtain a CTA scan image;

[0062] PCAT range acquisition step S2: obtaining a PCAT range map of the epicardial coronary artery through the CTA scan image;

[0063] Parameter acquisition step S3: calculating the PCAT volume value and FAI value corresponding to the PCAT range map in the CTA scan image;

[0064] Prediction step S4: obtaining a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

[0065] Clinically, patients at high risk of coronary heart disease with or without typical symptoms usually need a systemic assessment, including coronary CT angiography (CTA), to determine whether they are at risk of atherosclerosis. However, there is a lack of evidence-based medical research on whether people at high risk of coronary heart disease will develop coronary atherosclerosis, and when and how to monitor coronary atherosclerosis. Coronary CTA has significant advantages in plaque detection and plaque composition analysis, but the location and morphology of new coronary plaques and the characteristics of the pericoronary fat attenuation index (FAI) during the formation of new plaques are still unclear.

[0066] The present invention aims to derive the characteristics of coronary CTA based on the coronary fat attenuation index (FAI) of patients with new plaques, analyze the relationship between pericoronary fat (PCAT) based on coronary CTA and new coronary atherosclerosis, and construct a prediction model, so as to predict new coronary plaques in high-risk patients with coronary heart disease with or without typical symptoms based on pericoronary fat (PCAT) and FAI.

[0067] Wherein, the scanning step includes:

[0068] Retrospective ECG gating was adopted, with the tube voltage setting range of 120kV-135kV, the tube current setting range of 350mA-450mA, the slice thickness setting range of 0.5mm-0.625mm, the circle speed of 0.28s / circle or 0.35s / circle, and a two-way high-pressure injector was used to inject 50ml-70ml of non-ionic contrast agent at a flow rate of 4.0ml / s-5.0ml / s through the subject's antecubital vein indwelling needle, followed by injection of 30ml of normal saline at the same flow rate before scanning. The scanning range was from 1.0cm below the tracheal carina to above the diaphragm. In the present embodiment, the non-ionic contrast agent contained 370mg / ml of iodine.

[0069] Further, the PCAT range acquisition step S2 includes:

[0070] Intercepting the CTA scan image according to a preset range to obtain the PCAT range map;

[0071] The preset range is the proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded;

[0072] The PCAT range map includes: a right coronary artery PCAT range map, an anterior descending branch PCAT range map and a circumflex branch PCAT range map.

[0073] Further, please refer to Figure 2 , Figure 2 for Figure 1 The step-by-step flow chart of step S3 in FIG. Figure 2 As shown, the parameter acquisition step S3 includes:

[0074] Volume value acquisition step S31: calculating and obtaining the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map;

[0075] FAI value acquisition step S32: calculating and obtaining the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map;

[0076] FAI value processing step S33: obtaining the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

[0077] Wherein, the volume value acquisition step S31 and the FAI value acquisition step S32 both include:

[0078] The PCAT volume value and the FAI value are calculated by obtaining the CT value of each pixel of the right coronary artery PCAT range map, the anterior descending branch PCAT range map, and the circumflex branch PCAT range map.

[0079] Specifically, to measure perivascular FAI, the present invention focuses on the proximal 40 mm segments of all three major epicardial coronary artery vessels, namely the right coronary artery (RCA), the anterior descending artery (LAD), and the circumflex artery (LCX). The corresponding perivascular fat is defined as the adipose tissue within the radial distance equal to the vessel diameter from the external vessel wall. To avoid the influence of the aortic wall, the most proximal 10 mm of the RCA was excluded. The proximal 10-50 mm of the vessel was measured. The fat density was calculated in the range of -190HU to -30HU. In LAD and LCX, the proximal 40 mm of each vessel was measured, excluding the left main coronary artery. The perivascular FAI value was determined after correcting the attenuation histogram of weighted perivascular fat by adjusting the technical parameters within the range of -190HU to -30HU. FAI values ​​were measured in three coronary arteries (RCA, LAD, and LCX) of each patient and those with normal coronary arteries. For each measurement, the artery with the highest FAI value was selected for comparison. Based on the measured pericoronary fat tissue, we semi-automatically extracted the value of each pixel, selected the CT value of each pixel for measurement and recording, and calculated the values ​​of all pixel points for statistical selection and analysis. The pixel value of each point was more accurate.

[0080] In this embodiment, since the three different blood vessels, RCA, LAD and LCX, may have different predictive abilities for coronary plaques, we weighted the FAI values ​​of different blood vessels to obtain the overall FAI index.

[0081] For further information, please refer to Figure 3 , Figure 3 for Figure 1 The step-by-step flow chart of step S4 in FIG. Figure 3 As shown, the prediction step S4 includes:

[0082] Information acquisition step S41: acquiring status information of multiple patients and processing the status information to obtain a training set and a test set;

[0083] Model construction step S42: constructing multiple initial prediction models based on the logistics regression model, decision tree model, random forest model and XGBoost model, and respectively training and testing the multiple initial prediction models through the training set and the test set to obtain test results;

[0084] Model selection step S43: calculating the test results of each of the initial prediction models to obtain the corresponding test accuracy, and selecting the initial prediction model corresponding to the highest test accuracy as the new coronary plaque prediction model;

[0085] Prediction result acquisition step S44: obtaining the prediction result through the new coronary plaque prediction model according to the PCAT volume value and the FAI value of the subject.

[0086] The status information includes: age, gender, BMI, interval between two coronary CTA examinations, diabetes, hypertension, smoking history, total cholesterol, high-density, low-density lipoprotein cholesterol, triglycerides, and FAI value. Specifically, the general information of the patient is collected on the same day as the first coronary CTA; diabetes is defined as blood glucose level ≥7.0mmol / L in the fasting state, ≥11.1mmol / L 2 hours after a meal or at a random time, or glycosylated hemoglobin A1C level ≥6.5%; hypertension is defined as systolic blood pressure exceeding 140mmHg, diastolic blood pressure exceeding 90mmHg, or taking antihypertensive drugs. ; Smokers are defined as smoking one or more cigarettes a day for at least 6 months; BMI exceeding 25kg / m2 is considered overweight, and BMI exceeding 30kg / m2 is considered obese. Blood parameters collected from patients included: blood samples were tested within 30 days before and after the first coronary CTA. All blood samples were obtained in the morning after an overnight fast of more than 12 hours. The hematological and biochemical parameters analyzed included: total cholesterol, high-density and low-density lipoprotein cholesterol, and triglycerides.

[0087] The model selection step S43 includes:

[0088] Obtaining the prediction result of whether each patient has coronary artery plaque by using each of the initial prediction models according to the test set;

[0089] Calculating the prediction results of the plurality of patients obtained by each of the initial prediction models to obtain the test accuracy rate corresponding to the initial prediction model;

[0090] Sorting the multiple test accuracy rates;

[0091] The initial prediction model corresponding to the highest test accuracy is selected as the new coronary plaque prediction model.

[0092] In this embodiment, specifically, based on the above method, the baseline general information, blood indexes and two coronary CTAs of 160 patients with negative baseline coronary CTA and high risk of cardiovascular disease were collected, and the following process was performed:

[0093] 1. Data cleaning

[0094] Missing value processing: For missing values, statistics such as the median\(Median\), mean\(\bar{X}\) or mode\(Mode\) of the variable can be used to fill them.

[0095] [X_{\text{filled}}=\text{fill\_missing}(X,\text{strategy})\]

[0096] Outlier handling: Use the IQR (interquartile range) method to identify outliers. If a value is less than \(Q1-1.5\times IQR\) or greater than \(Q3+1.5\times IQR\), it may be considered an outlier.

[0097] [IQR = Q3-Q1\]

[0098] Data standardization: Perform Z-score standardization or minimum-maximum normalization on numerical variables.

[0099] [Z=\frac{X-\bar{X}}{S}\]

[0100] [X_{\text{norm}}=\frac{X-X_{\text{min}}}{X_{\text{max}}-X_{\text{min}}}\]

[0101] 2. Feature selection and data splitting

[0102] Correlation analysis: Analyze the correlation between each variable and the target variable, i.e., new coronary plaque, and select variables with higher correlation as features.

[0103] Multicollinearity check: Use VIF (variance inflation factor) to check for multicollinearity issues between features. If there are highly collinear features, consider removing some.

[0104] Split the data randomly into training and test sets, for example, use 70% of the data for training and 30% for testing.

[0105] 3. Model Construction

[0106] 3.1 Based on the logistics regression model, the initial prediction model was constructed.

[0107] The formula of the Logistic regression model can be expressed as:

[0108]

[0109] in:

[0110] (P(y=1|FAI)) represents the probability that the dependent variable (y) takes the value of 1 when the input variable (FAI) is (x);

[0111] (\beta_0) represents a constant term;

[0112] (\beta_1) represents the regression coefficient, which indicates the impact of (FAI) on (y).

[0113] At the same time, after incorporating multidimensional variables, a more accurate model was constructed using machine learning methods. The variables included:

[0114] a: General information: age, gender, BMI, interval between two coronary CTA examinations, diabetes, hypertension, smoking history

[0115] b: Blood test information: total cholesterol, high-density lipoprotein cholesterol and low-density lipoprotein cholesterol, triglycerides

[0116] c: Imaging information: FAI index

[0117] The output variable is whether coronary artery plaque occurs.

[0118] Decision tree model, random forest model and XGBoost model

[0119] 3.2 Building an initial prediction model based on the decision tree model

[0120] Decision tree: Information gain, select the feature that maximizes the information gain for segmentation.

[0121] [IG(D_p,f)=I(D_p)-\sum_{j=1}^{m}\frac{N_j}{N_p}I(D_j)\],

[0122] Gini Impurity: You can also choose features that reduce the Gini impurity the most.

[0123] [Gini(D)=1-\sum_{i=1}^{n}p_i^2\],

[0124] Build a decision tree until each leaf node is pure or the minimum number of node samples is reached.

[0125] 3.3 Building an initial prediction model based on the random forest model

[0126] Random forests improve prediction accuracy by building multiple decision trees and averaging or majority voting their predictions.

[0127] During the training of each decision tree, a subset of features is randomly selected to increase the diversity of the model.

[0128] [RF=\frac{1}{N}\sum_{i=1}^{N}DT_i(X)\].

[0129] 3.4 Building the initial prediction model XGBoost based on the XGBoost model

[0130] XGBoost (Extreme Gradient Boosting) is an ensemble learning algorithm based on the decision tree algorithm. It builds multiple weak classifiers (usually decision trees) and combines them into a strong classifier to achieve higher prediction accuracy.

[0131] XGBoost adds trees sequentially during training, and only optimizes the tree model in the current step in each iteration. Each new tree is built based on the residuals of all previous tree predictions and the true values ​​of the training samples.

[0132] F m (x i )=F m-1 (x i )+f m (x i )

[0133] Among them, f m (x i ) is the sub-model of the current m-th step, F m-1 (x i ) are the first m-1 sub-models that have been fixed after training.

[0134] 3.4.1 XGBoost Objective Function

[0135] XGBoost reduces a defined objective function at each step during training, which consists of two parts: training loss and regularization term.

[0136] [\text{Obj}=\sum_{i}I(y_i,\hat{y}_i^{(t)})+\sum_{k}\Omega(f_k)\]

[0137]

[0138] Among them, F m (xi ) is the i-th sample x i The predicted value, y i is the i-th sample x i The true value of the regularization term Ω(f j ) represents the sub-model f j complexity.

[0139] XGBoost uses the second-order expansion to approximate the loss function, that is

[0140]

[0141] To prevent overfitting, XGBoost sets tree-based complexity as a regularization term:

[0142]

[0143] Among them, T is the number of leaf nodes of tree f, ω is the vector composed of the output regression values ​​of all leaf nodes, ||ω|| 2 is the square of the L2 norm (modulus) of the vector, and γ and λ are hyperparameters. As a regression tree, the more leaf nodes there are, the larger the output regression value is, and the higher the complexity of the tree is.

[0144] The training loss part is about f m (x i ) are all constants and do not affect f m (x i ), the final objective function is as follows:

[0145]

[0146] In order to combine the regularization term and the empirical risk term, we convert the empirical risk term summed at the sample level into a summed at the leaf node level. Define the sample set at node j as I(j) = {x i |q(x i )=j}, where |q(x i ) is the index function that maps samples to leaf nodes, and the regression value on leaf node j is ω j =f m (x i ), i∈I(j).

[0147]

[0148] Let ∑ i∈I(j) g i =G j ,∑ i∈I(j) h i =H j ,have to

[0149]

[0150] If the structure of a tree is determined, G j , H j , T has also been determined, and the regression value output by each leaf node should minimize the above formula. It can be seen that the extreme point of the above quadratic function is:

[0151]

[0152] The objective function value (the score of the tree) is as follows:

[0153]

[0154] 3.4.2 Node Splitting Criteria

[0155] When a node splits, XGBoost will sort the eigenvalues ​​and traverse the partition points, taking the optimal split benefit as the split benefit of the feature, selecting the feature with the optimal split benefit as the partition feature of the current node, and performing binary partitioning according to its optimal partition point to obtain the left and right subtrees.

[0156] The gain from a split is the score of the tree before the split minus the score of the tree after the split, because the scores of the non-split nodes are offset, leaving only the nodes after the split to compare with the nodes before the split, so the expression for the gain from a split is:

[0157]

[0158] Among them, L is the left node after the split, and R is the right node after the split

[0159] 3.4.3 Greedy Algorithm

[0160] In the process of generating new trees by XGBoost, the most basic operation is node splitting. The most important step in node splitting is to find the optimal features and the optimal splitting points, and then split the leaf nodes according to the optimal features and the optimal splitting points.

[0161] The idea of ​​selecting the optimal features and optimal splitting points based on the greedy algorithm is as follows:

[0162] First, all candidate features and all candidate split points are found, and the split benefits are calculated in turn. Then, the feature with the maximum split benefit and its split point are selected as the optimal feature and the optimal split point, so it is called the exact greedy algorithm. This algorithm is a heuristic algorithm because when a node splits, it only selects the current optimal split strategy, not the global optimal split strategy.

[0163] 3.4.4XGBoost algorithm training process

[0164] 1). Initialize the predicted value of each sample,

[0165] 2). Calculate the derivative of the loss function for each sample prediction value,

[0166] 3). Build a new decision tree based on derivative information,

[0167] 4). Add the new decision tree prediction sample value to the original prediction value.

[0168] 5). Repeat steps 2, 3, and 4 to build a decision tree until the stopping condition is met.

[0169] 4. Model Training

[0170] The model is trained using the training set data, that is, the optimal model parameters are found by minimizing the loss function.

[0171] 5. Model validation and tuning

[0172] Cross-validation: K-fold cross-validation is used to evaluate the generalization ability of the model.

[0173] Performance evaluation: Use metrics such as accuracy, recall, and F1 score to evaluate model performance.

[0174] Parameter Tuning: By adjusting the parameters of the model, such as the maximum depth of the decision tree, the number of trees in the random forest, the learning rate of XGBoost, etc., cross-validation is used to find the optimal parameter combination.

[0175] 6. Model Application

[0176] Use the optimal model and parameters to make predictions on the test set to evaluate the performance of the model on unseen data.

[0177] Please refer to Figure 4 , Figure 4 FIG. 1 is a schematic diagram of the structure of the new coronary plaque prediction system of the present invention. Figure 4 As shown, a new coronary plaque prediction system based on coronary artery FAI of the present invention, wherein the new coronary plaque prediction method mentioned above is applied, and the new coronary plaque prediction system comprises:

[0178] The scanning unit 11 scans the area from below the tracheal carina to above the diaphragm of the subject by means of a CT device with at least 64 slices to obtain a CTA scan image;

[0179] A PCAT range acquisition unit 12, which obtains a PCAT range map of the epicardial coronary artery through the CTA scan image;

[0180] A parameter acquisition unit 13 calculates a PCAT volume value and a FAI value corresponding to the PCAT range map in the CTA scan image;

[0181] The prediction unit 14 obtains a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

[0182] Further, the PCAT range acquisition unit 12 includes: intercepting the CTA scan image according to a preset range to obtain the PCAT range map; wherein the preset range is a proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded; the PCAT range map includes: a right coronary artery PCAT range map, anterior descending branch PCAT range map and circumflex branch PCAT range map;

[0183] Furthermore, the parameter acquisition unit package 13 includes:

[0184] The volume value acquisition module 131 calculates and obtains the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map;

[0185] The FAI value acquisition module 132 calculates the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map;

[0186] The FAI value processing module 133 obtains the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

[0187] Wherein, the volume value acquisition module 131 and the FAI value acquisition module 132 both include:

[0188] The PCAT volume value and the FAI value are calculated by acquiring the CT value of each pixel of the right coronary artery PCAT range map, the anterior descending branch PCAT range map, and the circumflex branch PCAT range map.

[0189] Furthermore, the prediction unit 14 includes:

[0190] An information acquisition module 141 acquires status information of multiple patients and processes the status information to obtain a training set and a test set;

[0191] A model building module 142 builds a plurality of initial prediction models based on a logistics regression model, a decision tree model, a random forest model, and an XGBoost model, and respectively trains and tests the plurality of initial prediction models through the training set and the test set to obtain test results;

[0192] The model selection module 143 calculates the test results of each of the initial prediction models to obtain the corresponding test accuracy, and selects the initial prediction model corresponding to the highest test accuracy as the new coronary plaque prediction model;

[0193] The prediction result acquisition module 144 obtains the prediction result through the new coronary plaque prediction model according to the PCAT volume value and the FAI value of the subject.

[0194] The status information includes: age, gender, BMI, interval between two coronary artery CTA examinations, diabetes, hypertension, smoking history, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, and FAI value.

[0195] Furthermore, the model selection module 143 includes:

[0196] According to the test set, the prediction result of whether each patient has coronary artery plaque is obtained through each initial prediction model; the prediction results of multiple patients obtained by each initial prediction model are calculated to obtain the test accuracy rate corresponding to the initial prediction model; the multiple test accuracy rates are sorted; and the initial prediction model corresponding to the highest test accuracy rate is selected as the new coronary artery plaque prediction model.

[0197] In summary, in view of the current lack of evidence-based medical evidence, the present invention can predict whether the current high-risk population for coronary heart disease will develop coronary atherosclerosis, and conduct real-time arteriosclerosis monitoring for the high-risk population for coronary heart disease, thereby conducting a more effective and accurate assessment of such population to improve prognosis.

[0198] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting new coronary plaque based on coronary artery FAI, characterized in that: include: Scanning steps: Use a CT device with at least 64 slices to scan the area from below the tracheal carina to above the diaphragm of the subject to obtain a CTA scan image; PCAT range acquisition step: obtaining a PCAT range map of the epicardial coronary artery through the CTA scan image; Parameter acquisition step: calculating the PCAT volume value and FAI value corresponding to the PCAT range map in the CTA scan image; Prediction step: obtaining a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

2. The method for predicting newly-occurring coronary plaque according to claim 1, characterized in that: The scanning step comprises: Retrospective ECG gating was used, with tube voltage setting range of 120kV-135kV, tube current setting range of 350mA-450mA, slice thickness setting range of 0.5mm-0.625mm, circle speed of 0.28s / circle or 0.35s / circle, and a two-way high-pressure syringe was used to inject 50ml-70ml of non-ionic contrast agent at a flow rate of 4.0ml / s-5.0ml / s through the subject's antecubital vein indwelling needle, followed by injection of 30ml of normal saline at the same flow rate before scanning. The scanning range was 1.0cm below the tracheal carina to above the diaphragm.

3. The method for predicting newly-occurring coronary plaque according to claim 1, characterized in that: The PCAT range acquisition step includes: Intercepting the CTA scan image according to a preset range to obtain the PCAT range map; The preset range is the proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded; The PCAT range map includes: a right coronary artery PCAT range map, an anterior descending branch PCAT range map and a circumflex branch PCAT range map.

4. The method for predicting newly-occurring coronary plaque according to claim 3, characterized in that: The parameter acquisition step comprises: Volume value acquisition step: calculating and obtaining the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map; FAI value acquisition step: calculating the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map; FAI value processing step: obtaining the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

5. The method for predicting newly-occurring coronary plaque according to claim 4, characterized in that: The volume value acquisition step and the FAI value acquisition step both include: The PCAT volume value and the FAI value are calculated by acquiring the CT value of each pixel of the right coronary artery PCAT range map, the anterior descending branch PCAT range map, and the circumflex branch PCAT range map.

6. The method for predicting newly-occurring coronary plaque according to claim 5, characterized in that: The prediction step includes: Information acquisition step: acquiring status information of multiple patients and processing the status information to obtain a training set and a test set; Model construction step: construct multiple initial prediction models based on the logistics regression model, decision tree model, random forest model and XGBoost model, and respectively train and test the multiple initial prediction models through the training set and the test set to obtain test results; Model selection step: calculating the test results of each of the initial prediction models to obtain the corresponding test accuracy, and selecting the initial prediction model corresponding to the highest test accuracy as the new coronary plaque prediction model; Prediction result acquisition step: obtaining the prediction result through the new coronary plaque prediction model according to the PCAT volume value and the FAI value of the subject.

7. The method for predicting newly-occurring coronary plaque according to claim 6, characterized in that: The status information includes: age, gender, BMI, interval between two coronary artery CTA examinations, diabetes, hypertension, smoking history, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides, and FAI value.

8. The method for predicting newly-occurring coronary plaque according to claim 6, characterized in that: The model selection steps include: Obtaining the prediction result of whether each patient has coronary artery plaque by using each of the initial prediction models according to the test set; Calculating the prediction results of the plurality of patients obtained by each of the initial prediction models to obtain the test accuracy rate corresponding to the initial prediction model; Sorting the multiple test accuracy rates; The initial prediction model corresponding to the highest test accuracy is selected as the new coronary plaque prediction model.

9. A new coronary plaque prediction system based on coronary artery FAI, characterized in that: The method for predicting newly-developed coronary plaque according to any one of claims 1 to 8 is applied, wherein the system for predicting newly-developed coronary plaque comprises: A scanning unit, which scans the area from below the tracheal carina to above the diaphragm of the subject by means of a CT device with at least 64 slices to obtain a CTA scanning image; A PCAT range acquisition unit, which obtains a PCAT range map of the epicardial coronary artery through the CTA scan image; A parameter acquisition unit, which calculates a PCAT volume value and a FAI value corresponding to the PCAT range map in the CTA scan image; A prediction unit obtains a prediction result by constructing a new coronary plaque prediction model according to the PCAT volume value and the FAI value.

10. The system for predicting newly-occurring coronary plaque according to claim 9, characterized in that: The PCAT range acquisition unit includes: intercepting the CTA scan image according to a preset range to obtain the PCAT range map; wherein the preset range is a proximal 40 mm segment of the epicardial coronary artery, and the epicardial coronary artery includes: the right coronary artery, the anterior descending branch and the circumflex branch, wherein the most proximal 10 mm segment of the right coronary artery is excluded; the PCAT range map includes: a right coronary artery PCAT range map, anterior descending branch PCAT range map and circumflex branch PCAT range map; The parameter acquisition unit comprises: A volume value acquisition module, which calculates and obtains the right coronary artery PCAT volume value, the anterior descending branch PCAT volume value and the circumflex branch PCAT volume value according to the right coronary artery PCAT range map, the anterior descending branch PCAT range map and the circumflex branch PCAT range map; An FAI value acquisition module, which calculates the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value according to the right coronary artery PCAT range map, the anterior descending artery PCAT range map and the circumflex artery PCAT range map; The FAI value processing module obtains the final FAI value through weighted processing according to the right coronary artery FAI value, the anterior descending artery FAI value and the circumflex artery FAI value.

Images