CN118448050B - CT-based large airway stenosis patient lung function prediction method, device, equipment and storage medium - Google Patents

Abstract

The invention belongs to the technical field of medical data intellectualization, and particularly provides a CT-based method, a device, equipment and a storage medium for predicting lung function of a patient with large airway stenosis, wherein the method comprises the following steps: acquiring CT image data of a patient with a large airway stenosis, extracting airway data from the image data, and performing three-dimensional reconstruction by using a convolution surface technology to generate an airway model; the airway model includes a stenosis model and a normal model; inputting the lung function statistical data into a normal model for simulation, generating simulation data of the normal model, and then transplanting the simulation data of the normal model into a narrow model for simulation by taking the simulation data of the normal model as input, so as to generate the simulation data of the narrow model; and inputting the current patient simulation data into the established lung function prediction model, and outputting a predicted value of the lung function of the current large airway stenosis patient. Airway data can be extracted more accurately, and further accuracy of lung function assessment is improved.

Description

A CT-based lung function prediction method, device, equipment and storage medium for patients with large airway stenosis

Technical Field

The present invention relates to the field of medical data intelligent technology, and in particular to a CT-based lung function prediction method, device, equipment and storage medium for patients with large airway stenosis.

Background Art

As an important means of evaluating respiratory diseases, the clinical significance and value of pulmonary function tests cannot be ignored. Pulmonary function tests can provide accurate information about the severity, prognosis, and treatment effects of respiratory diseases, providing a strong basis for doctors to formulate personalized treatment plans. However, in practical applications, pulmonary function tests also face some challenges and limitations.

First of all, pulmonary function tests require a certain degree of cooperation from patients. For patients who are unable to cooperate with the test due to various reasons (such as coma, severe coughing, excessive tension, etc.), the accuracy of pulmonary function tests will be seriously affected. These patients may not be able to complete the breathing movements as instructed by the doctor, or they may not be able to maintain the stability of breathing, resulting in deviations and misleading of pulmonary function data. Therefore, for such patients, we need to find alternative methods to assess their pulmonary function.

Secondly, pulmonary function tests are real-time and can only reflect the patient's current lung function status. However, in some cases, we need to understand the patient's future changes in lung function in order to better formulate treatment plans and predict prognosis. For example, for patients with airway tumor types, as the tumor grows or shrinks, the patient's lung function will also change accordingly. If we can accurately predict these changes, we can take measures in advance to protect the patient's lung function and avoid unnecessary risks and complications. However, traditional pulmonary function test methods cannot directly predict future changes in lung function, which limits their clinical application.

Summary of the invention

In order to solve the above problems, the present invention provides a CT-based lung function prediction method, device, equipment and storage medium for patients with large airway stenosis.

In a first aspect, the technical solution of the present invention provides a method for predicting lung function of patients with large airway stenosis based on CT, comprising the following steps:

Obtain CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model;

Inputting the lung function statistical data into the normal model for simulation to generate simulation data of the normal model, and then using the simulation data of the normal model as input to transplant into the stenosis model for simulation to generate simulation data of the stenosis model;

The current patient simulation data is input into the created lung function prediction model, and the predicted value of the lung function of the current large airway stenosis patient is output.

As a further limitation of the technical solution of the present invention, the steps of obtaining CT image data of a patient with large airway stenosis and extracting airway data from the image data, and using convolution surface technology to perform three-dimensional reconstruction to generate an airway model include:

Acquire CT image data of patients with large airway stenosis, and extract airway data from the image data;

generating skeleton data according to the extracted airway data;

The airway model is generated by three-dimensional reconstruction using convolution surface technology based on the generated skeleton data; the airway model includes a stenosis model consistent with the current state of the patient with large airway stenosis and a normal model expanded according to the current patient.

As a further limitation of the technical solution of the present invention, the steps of obtaining CT image data of a patient with large airway stenosis and extracting airway data from the image data include:

Obtain lung CT imaging data of patients with large airway stenosis;

The airway segmentation function of the medical imaging modeling software was used to select two seed points in the airway; the airway tree was extracted based on the region growing algorithm, and local smoothing and detriangulation were performed; the airway centerline extraction function of the software was used to create an airway tree center path based on the current airway tree to generate an airway tree model;

The airway tree model is exported as a first format file, and the airway tree center path is exported as a second format file, wherein the first format file contains the airway tree surface grid coordinate data and connection relationship, and the second format file contains the airway tree center path point coordinate data and connection relationship, that is, airway data.

As a further limitation of the technical solution of the present invention, the step of generating skeleton data according to the extracted airway data includes:

The airway tree is graded according to the central path of the airway tree; the main airway is grade 0, the branches connected to grade 0 are grade 1, the branches connected to grade 1 are grade 2, and so on;

Connect the first and last points of each branch as the center line of the current branch, draw perpendicular lines from each point on the current branch to the center line of the current branch, and each center line and the perpendicular line to the center line form a plane, obtain the cross-sectional area where the plane intersects with the surface of the current branch of the airway tree, and calculate the equivalent radius of each point on the current branch according to the cross-sectional area; retrieve the radius of the branch connected to each branch point, take the midpoint of each branch as the two end points, keep the equivalent radius of the end point unchanged, and use linear transition control to optimize the equivalent radius of each branch point that is greater than the set value as the radius of each point;

Using the fitting method, the airway tree center path is fitted and interpolated so that each branch contains a set number of coordinate data points, and all the coordinate data points constitute the airway centerline data;

The airway centerline data and the radius of each point together constitute the skeleton data.

As a further limitation of the technical solution of the present invention, the step of generating an airway model by three-dimensional reconstruction using convolution surface technology based on the generated skeleton data includes:

Based on the convolution surface technology, skeleton data is used for modeling to reconstruct the large airway stenosis model of the patient's current state, namely, the stenosis model;

Compare the skeleton data with the airway statistical data of the normal population, identify the stenosis site, expand the radius data of the patient's stenosis point to the normal level corresponding to the patient, and establish normal skeleton data based on the current patient;

Based on the convolution surface technology, normal skeleton data is used for modeling to reconstruct the airway model of the patient in a healthy state, that is, the normal model.

As a further limitation of the technical solution of the present invention, the statistical data of lung function is input into a normal model for simulation to generate simulation data of the normal model, and then the simulation data of the normal model is transplanted as input into a stenosis model for simulation, and the steps of generating simulation data of the stenosis model include:

Obtaining lung function statistical data, i.e. lung function data of normal people;

According to the theory of lung function proportion of lung segments, the current patient's lung function statistical data is proportionally distributed to the branch openings of each lung segment as input to obtain the pressure of each branch opening of the normal model;

The branch port pressure obtained from the normal model is then used as input to each branch port corresponding to the stenosis model to obtain the total outlet flow of the stenosis model, that is, the simulation data of the stenosis model.

As a further limitation of the technical solution of the present invention, before the step of inputting the current patient simulation data into the created lung function prediction model and outputting the predicted value of the current lung function of the patient with large airway stenosis includes:

The lung function prediction model is trained based on the historical simulation data of patients with large airway stenosis and the historical lung function data of patients with large airway stenosis.

In a second aspect, the technical solution of the present invention provides a CT-based lung function prediction device for patients with large airway stenosis, comprising a model reconstruction module, a model simulation module and a lung function prediction module;

A model reconstruction module is used to obtain CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model;

A model simulation module is used to input lung function statistical data into a normal model for simulation to generate simulation data of the normal model, and then transplant the simulation data of the normal model as input into a stenosis model for simulation to generate simulation data of the stenosis model;

The pulmonary function prediction module is used to input the current patient simulation data into the created pulmonary function prediction model and output the predicted value of the pulmonary function of the current large airway stenosis patient.

As a further limitation of the technical solution of the present invention, the model reconstruction module includes an airway data extraction unit, a skeleton data generation unit and a model reconstruction unit;

An airway data extraction unit, used to obtain CT image data of a patient with large airway stenosis and extract airway data from the image data;

A skeleton data generating unit, used for generating skeleton data according to the extracted airway data;

The model reconstruction unit is used to generate an airway model by three-dimensional reconstruction using convolution surface technology based on the generated skeleton data; the airway model includes a stenosis model consistent with the current state of the patient with large airway stenosis and a normal model expanded according to the current patient.

As a further limitation of the technical solution of the present invention, the airway data extraction unit includes an image data acquisition submodule, an airway tree model generation submodule and an airway data generation submodule;

An image data acquisition submodule is used to acquire lung CT image data of patients with large airway stenosis;

The airway tree model generation submodule is used to select two seed points in the airway using the airway segmentation function of the medical imaging modeling software; extract the airway tree based on the region growing algorithm, and perform local smoothing and detriangulation processing; use the airway centerline extraction function of the software to create an airway tree center path based on the current airway tree to generate an airway tree model;

The airway data generation submodule is used to export the airway tree model as a first format file and export the airway tree center path as a second format file, wherein the first format file contains the airway tree surface grid coordinate data and connection relationship, and the second format file contains the airway tree center path point coordinate data and connection relationship, that is, airway data.

As a further limitation of the technical solution of the present invention, the skeleton data generation unit includes an airway tree classification submodule, a radius data acquisition submodule and an airway centerline data acquisition submodule;

The airway tree classification submodule is used to classify the airway tree according to the central path of the airway tree; wherein the main airway is level 0, the branches connected to level 0 are level 1, the branches connected to level 1 are level 2, and so on;

The radius data acquisition submodule is used to connect the first and last points of each branch as the center line of the current branch, draw perpendicular lines from each point on the current branch to the center line of the current branch, and each center line and the perpendicular line of the center line form a plane, and obtain the cross-sectional area where the plane intersects with the surface of the current branch of the airway tree, and calculate the equivalent radius of each point on the current branch according to the cross-sectional area; retrieve the radius of the branch connected to each branch point, take the midpoint of each branch as the two end points, keep the equivalent radius of the end point unchanged, and use linear transition control to optimize the equivalent radius of each branch point that is greater than the set value as the radius of each point;

The airway centerline data acquisition submodule is used to fit and interpolate the airway tree center path using a fitting method, so that each branch contains a set number of coordinate data points, and all coordinate data points constitute the airway centerline data; wherein, the airway centerline data and the radius of each point together constitute the skeleton data.

As a further limitation of the technical solution of the present invention, the model reconstruction unit includes a normal skeleton data establishment submodule and a model reconstruction submodule;

The model reconstruction submodule uses the skeleton data to build a model based on the convolution surface technology to reconstruct the large airway stenosis model of the patient's current state, that is, the stenosis model; it uses the normal skeleton data to build a model based on the convolution surface technology to reconstruct the airway model of the patient's healthy state, that is, the normal model;

The normal skeleton data establishment submodule is used to compare the skeleton data with the airway statistical data of the normal population, identify the stenosis site, expand the radius data of the patient's stenosis point to the normal level corresponding to the patient, and establish normal skeleton data based on the current patient.

As a further limitation of the technical solution of the present invention, the model simulation module includes a lung function data acquisition unit and a simulation unit;

A lung function data acquisition unit, used to acquire lung function statistical data, i.e., lung function data of normal people;

The simulation unit is used to distribute the current patient's lung function statistical data in proportion to the branch outlets of each lung segment as input according to the segment lung function ratio theory, and obtain the pressure of each branch outlet of the normal model; then use the branch outlet pressure obtained from the normal model as input, and input it into each branch outlet corresponding to the stenosis model to obtain the total outlet flow of the stenosis model, that is, the simulation data of the stenosis model.

In a third aspect, the technical solution of the present invention provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions executable by the at least one processor, and the computer program instructions are executed by the at least one processor so that the at least one processor can execute the CT-based lung function prediction method for patients with large airway stenosis as described in the first aspect.

In a fourth aspect, the technical solution of the present invention also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable the computer to execute the CT-based lung function prediction method for patients with large airway stenosis as described in the first aspect.

It can be seen from the above technical solutions that the present invention has the following advantages: predicting the patient's lung function data based on CT data makes up for the deficiency that lung function tests cannot be performed directly due to the patient's physical condition; at the same time, the lung function of the patient's large airway stenosis degree can be predicted accordingly based on the patient's current airway model, which is of great significance for predicting the lung function of patients with airway tumors throughout the course of the disease and predicting the lung function after airway expansion surgery.

By using convolution surface technology to perform three-dimensional reconstruction to generate an airway model, airway data can be extracted more accurately, thereby improving the accuracy of lung function assessment. By comparing the normal model and the stenosis model, doctors can more intuitively understand the impact of airway stenosis on the patient's lung function, providing more powerful auxiliary support for the diagnosis and treatment of the disease. This method can construct a personalized lung function prediction model based on the patient's specific situation and realize personalized lung function assessment. This helps doctors formulate more accurate treatment plans based on the patient's specific situation.

In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

FIG1 is a schematic flow chart of a method according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of a specific method flow in an embodiment of the present invention.

FIG3 is a diagram of airway-related data, wherein (a) in FIG3 is a diagram of patient CT data, (b) in FIG3 is a diagram of airway data, and (c) in FIG3 is a diagram of skeleton data.

FIG. 4 is an airway model diagram, wherein FIG. 4( a ) is a normal model diagram, and FIG. 4( b ) is a stenosis model diagram.

FIG. 5 is a schematic block diagram of an apparatus according to an embodiment of the present invention.

DETAILED DESCRIPTION

In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions 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 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 should fall within the scope of protection of the present invention.

As shown in FIG. 1 and FIG. 2 , an embodiment of the present invention provides a method for predicting lung function of patients with large airway stenosis based on CT, comprising the following steps:

Step 1: Obtain CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model;

Step 2: Inputting the lung function statistical data into the normal model for simulation to generate simulation data of the normal model, and then using the simulation data of the normal model as input to transplant into the stenosis model for simulation to generate simulation data of the stenosis model;

Step 3: Input the current patient simulation data into the created lung function prediction model, and output the predicted value of the lung function of the current large airway stenosis patient.

In this step, the pulmonary function prediction model is trained using the historical large airway stenosis patient simulation data and the historical large airway stenosis patient pulmonary function data generated in the model simulation module to establish a mapping relationship between the patient simulation data and the patient pulmonary function data.

In some embodiments, the steps of obtaining CT image data of a patient with large airway stenosis, extracting airway data from the image data, and performing three-dimensional reconstruction using convolution surface technology to generate an airway model include:

Step 11: Acquire CT image data of a patient with large airway stenosis, and extract airway data from the image data; here, mainly including airway patch data;

Step 12: Generate skeleton data based on the extracted airway data;

Step 13: Perform three-dimensional reconstruction based on the generated skeleton data using convolution surface technology to generate an airway model; the airway model includes a stenosis model consistent with the current state of the patient with large airway stenosis and a normal model expanded according to the current patient.

In this step, convolution surface technology is used to perform three-dimensional reconstruction to generate an airway model, which can generate a series of transitional airway models from complete occlusion to complete normality, including a stenosis model (a model consistent with the current state of a patient with large airway stenosis) and a normal model (a normal airway model expanded according to the current patient).

In some embodiments, the steps of obtaining CT image data of a patient with large airway stenosis and extracting airway data from the image data include:

Step 111: Acquire lung CT image data of a patient with large airway stenosis;

Step 112: using the airway segmentation function of the medical imaging modeling software to select two seed points in the airway; extracting the airway tree based on the region growing algorithm, and performing local smoothing and detriangulation processing; using the airway centerline extraction function of the software, creating an airway tree center path based on the current airway tree to generate an airway tree model;

Specifically, in this step, the airway tree is automatically or semi-automatically extracted using medical imaging modeling software such as Mimics, Simpleware, and 3D-DOCTOR. The airway segmentation function of the software is utilized. Different tissues have different HU values in the imaging data. Seed points in two airways are selected, and the airway tree is extracted based on the region growing algorithm. Surface optimization processing such as local smoothing and detriangulation is performed. The airway centerline extraction function of the software is used to create an airway tree center path based on the current airway tree to generate an airway tree model.

Step 113: export the airway tree model into a first format file, and export the airway tree center path into a second format file, wherein the first format file contains the airway tree surface grid coordinate data and connection relationship, and the second format file contains the airway tree center path point coordinate data and connection relationship, that is, airway data.

The airway tree model was exported as a .stl format file (including the airway tree surface mesh coordinate data and connection relationship), that is, the airway data, and the airway tree center path was exported as a .igs format file (including the airway tree center path point coordinate data and connection relationship).

In some embodiments, the step of generating skeleton data based on the extracted airway data includes:

Step 121: classifying the airway tree according to the central path of the airway tree; wherein the main airway is level 0, the branches connected to level 0 are level 1, the branches connected to level 1 are level 2, and so on;

Step 122: Connect the first and last points of each branch as the center line of the current branch, draw perpendicular lines from each point on the current branch to the center line of the current branch, and each center line and the perpendicular line to the center line form a plane, obtain the cross-sectional area where the plane intersects with the surface of the current branch of the airway tree, and calculate the equivalent radius of each point on the current branch according to the cross-sectional area; retrieve the radius of the branch connected to each branch point, take the midpoint of each branch as the two end points, keep the equivalent radius of the end point unchanged, and use linear transition control to optimize the equivalent radius of each branch point that is greater than the set value as the radius of each point;

Step 123: Using a fitting method, the airway tree center path is fitted and interpolated so that each branch contains a set number of coordinate data points, and all coordinate data points constitute airway centerline data; the airway centerline data and the radius of each point together constitute skeleton data.

When skeleton data is generated, the specific implementation process is:

As shown in Figure 3, (a) in Figure 3 is the patient CT data, and (b) in Figure 3 is the airway data. The patient CT data classifies the airway tree according to the central path of the airway tree, where the main airway is level 0, the branch connected to level 0 is level 1, the branch connected to level 1 is level 2, and so on. Connect the first and last points of each branch as the center line of the current branch _Li , and make a perpendicular line _Lij from each point on the current branch to the center line _Li of the current branch. _Li and _Lij form a plane _Aij , and the plane _Aij intersects with the surface of the current branch of the airway tree. The cross-sectional area _Sij is obtained, _Rij = As the equivalent radius of each point on the current branch, retrieve the radius of the branch connected to each branch point, take the midpoint of each branch as the two end points, keep the equivalent radius of the endpoint unchanged, and use linear transition control to optimize the oversized equivalent radius of each branch point as the radius of each point. Use fitting methods such as Nurbs fitting and polynomial fitting to fit and interpolate the central path of the airway tree so that each branch contains 1,000 coordinate data points, and all coordinate data points constitute the final airway centerline data. The airway centerline and the radius of each point together constitute the skeleton data, as shown in (c) in Figure 3, which is the skeleton data.

In some embodiments, the step of generating an airway model by performing three-dimensional reconstruction using convolutional surface technology based on the generated skeleton data includes:

Step 131: Modeling is performed using skeleton data based on convolutional surface technology to reconstruct a large airway stenosis model of the patient's current state, that is, a stenosis model;

Step 132: Compare the skeleton data with the airway statistical data of a normal population, identify the stenosis site, expand the radius data of the stenosis point of the patient to a normal level corresponding to the patient, and establish normal skeleton data based on the current patient;

Step 133: Modeling is performed using normal skeleton data based on convolutional surface technology to reconstruct the airway model of the patient in a healthy state, that is, a normal model.

The specific implementation process of this step is as follows:

Based on the convolution surface technology, the skeleton data is used to build a model and reconstruct the stenosis model, that is, the large airway stenosis model of the patient's current state. The patient's skeleton data is compared with the airway statistical data of the normal population to identify the stenosis site, and the radius data of the patient's stenosis point is expanded to the normal level corresponding to the patient. The normal skeleton data based on the current patient is established, and the normal skeleton data is used to build a model based on the convolution surface technology to reconstruct the normal model, that is, the airway model of the patient's healthy state. The core parameters of the convolution surface technology used are as follows. Assuming that the point P (x, y, z) in the three-dimensional space, Q is a point on the skeleton, and the truncated quartic polynomial function is used:

As a kernel function, it has local support and can support larger analytical expression of skeleton potential function, where r is the distance between P and Q, r=||PQ||, R>0 is called the effective radius, and the ball with P as the center and R as the radius is called the clipping ball. Assuming that the intersection interval of the straight line segment equation L(t)=P ₁ +t(P ₂ -P ₁ ) and the clipping ball is [t ₁ ,t ₂ ], the corresponding straight line segment skeleton potential function is as follows:

in, =‖P ₂ -P ₁ ‖, a=(P ₂ -P ₁ )•(PP ₁ ), h=R ² -‖PP ₁ ‖ ² .

The convolution surface performs linear transition control on the surface radius, and models are built based on different skeleton data to reconstruct the normal model and stenosis model of the patient. As shown in Figure 4, (a) in Figure 4 is a normal model, and (b) in Figure 4 is a stenosis model. At the same time, the airway radius of the patient's lesion position is controlled, and a series of airway models of the patient from complete occlusion to complete normality are established using convolution surface technology.

In some embodiments, the lung function statistical data is input into a normal model for simulation to generate simulation data of the normal model, and then the simulation data of the normal model is transplanted into a stenosis model for simulation using the simulation data of the normal model as input. The step of generating the simulation data of the stenosis model includes:

Step 21: Obtaining lung function statistical data, that is, lung function data of normal people;

Step 22: According to the segmental lung function ratio theory, the current patient's lung function statistical data is proportionally distributed to the branch openings of each lung segment as input to obtain the pressure of each branch opening of the normal model;

Step 23: The branch port pressure obtained from the normal model is used as input to each branch port corresponding to the stenosis model to obtain the total outlet flow of the stenosis model, that is, the simulation data of the stenosis model.

In this step, obtain lung function statistical data, that is, lung function data of normal people (including but not limited to FVC, FEV1, MVV, SVC and other lung function data as well as age, height, weight, and ethnicity data), use fluid simulation software such as Fluent, Comsol or Xflow, first simulate and analyze the normal model, and according to the classic segmental lung function ratio theory, distribute the current patient's FEV1 proportionally to the branch ports of each lung segment as input to obtain the pressure of each branch port of the normal model. Then use the branch port pressure obtained from the normal model as input, input it to each branch port corresponding to the stenosis model, and obtain the flow rate of the total outlet of the stenosis model.

In some embodiments, the step of inputting the current patient simulation data into the created lung function prediction model and outputting the predicted value of the current lung function of the patient with large airway stenosis includes:

A pulmonary function prediction model is obtained by training based on historical simulation data of patients with large airway stenosis and historical pulmonary function data of patients with large airway stenosis, wherein the training methods include but are not limited to linear regression model prediction algorithm, BP neural network, SVM support vector machine, RBF radial basis neural network and CNN convolutional neural network.

Specifically, historical lung function data of patients with large airway stenosis and historical simulation data of patients with large airway stenosis generated in the model simulation module are used for training to establish a mapping relationship between patient simulation data and patient lung function data, such as a univariate mapping relationship between flow and FEV1, or a multivariate mapping relationship between flow, pressure and FVC, FEV1.

The current patient simulation data is input into the created lung function prediction model, and the predicted value of the current lung function of the patient with large airway stenosis is output. In the embodiment of the present invention, an indicator of lung function, FEV1, is used for prediction, which specifically includes:

(1) Definition of measured ratio , simulation ratio The a and b data of four patients with airway stenosis are shown in Table 1.

(2) Determine the degree of linear correlation between a and b. Using the comparative T test on a and b, the linear correlation between the two is 0.994, with a significance of 0.006 and a Sig value of less than 0.05, indicating that at a confidence level of 95%, there is a significant statistical difference between the two and the degree of linear correlation between the two is high.

(3) Using the univariate linear regression analysis prediction method, a linear regression model of a and b was established. The linear regression prediction equation established was a=-0.2023+1.0546b, with a fitting effect of RSS=0.00181 and R-Square=0.98747.

Table 1

Finally, the total outlet flow data obtained by simulating the current large airway stenosis patients is input into the lung function prediction model, and the lung function FEV1 data of the current large airway stenosis patients is output.

At the same time, according to the method of step 1, a series of transition airway models from complete occlusion to complete normal are established, and input into step 2 for simulation to generate simulation data of the transition model, which is input into the lung function prediction model in step 3. This can realize the prediction of the patient's lung function throughout the course of the disease and the prediction of the postoperative lung function of patients undergoing airway expansion surgery.

As shown in FIG5 , an embodiment of the present invention provides a CT-based lung function prediction device for patients with large airway stenosis, including a model reconstruction module, a model simulation module, and a lung function prediction module;

A model reconstruction module is used to obtain CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model;

A model simulation module is used to input lung function statistical data into a normal model for simulation to generate simulation data of the normal model, and then transplant the simulation data of the normal model as input into a stenosis model for simulation to generate simulation data of the stenosis model;

The pulmonary function prediction module is used to input the current patient simulation data into the created pulmonary function prediction model and output the predicted value of the pulmonary function of the current large airway stenosis patient.

In some embodiments, the model reconstruction module includes an airway data extraction unit, a skeleton data generation unit, and a model reconstruction unit;

An airway data extraction unit, used to obtain CT image data of a patient with large airway stenosis and extract airway data from the image data;

A skeleton data generating unit, used for generating skeleton data according to the extracted airway data;

The model reconstruction unit is used to generate an airway model by three-dimensional reconstruction using convolution surface technology based on the generated skeleton data; the airway model includes a stenosis model consistent with the current state of the patient with large airway stenosis and a normal model expanded according to the current patient.

In some embodiments, the airway data extraction unit includes an image data acquisition submodule, an airway tree model generation submodule, and an airway data generation submodule;

An image data acquisition submodule is used to acquire lung CT image data of patients with large airway stenosis;

The airway tree model generation submodule is used to select two seed points in the airway using the airway segmentation function of the medical imaging modeling software; extract the airway tree based on the region growing algorithm, and perform local smoothing and detriangulation processing; use the airway centerline extraction function of the software to create an airway tree center path based on the current airway tree to generate an airway tree model;

The airway data generation submodule is used to export the airway tree model as a first format file and export the airway tree center path as a second format file, wherein the first format file contains the airway tree surface grid coordinate data and connection relationship, and the second format file contains the airway tree center path point coordinate data and connection relationship, that is, airway data.

In some embodiments, the skeleton data generation unit includes an airway tree classification submodule, a radius data acquisition submodule, and an airway centerline data acquisition submodule;

The airway tree classification submodule is used to classify the airway tree according to the central path of the airway tree; wherein the main airway is level 0, the branches connected to level 0 are level 1, the branches connected to level 1 are level 2, and so on;

The radius data acquisition submodule is used to connect the first and last points of each branch as the center line of the current branch, draw perpendicular lines from each point on the current branch to the center line of the current branch, and each center line and the perpendicular line to the center line form a plane, and obtain the cross-sectional area where the plane intersects with the surface of the current branch of the airway tree, and calculate the equivalent radius of each point on the current branch according to the cross-sectional area; retrieve the radius of the branch connected to each branch point, take the midpoint of each branch as the two end points, keep the equivalent radius of the end point unchanged, and use linear transition control to optimize the equivalent radius of each branch point that is greater than the set value as the radius of each point;

The airway centerline data acquisition submodule is used to fit and interpolate the airway tree center path using a fitting method, so that each branch contains a set number of coordinate data points, and all coordinate data points constitute the airway centerline data; wherein, the airway centerline data and the radius of each point together constitute the skeleton data.

In some embodiments, the model reconstruction unit includes a normal skeleton data establishment submodule and a model reconstruction submodule;

The model reconstruction submodule uses the skeleton data to build a model based on the convolution surface technology to reconstruct the large airway stenosis model of the patient's current state, that is, the stenosis model; it uses the normal skeleton data to build a model based on the convolution surface technology to reconstruct the airway model of the patient's healthy state, that is, the normal model;

The normal skeleton data establishment submodule is used to compare the skeleton data with the airway statistical data of the normal population, identify the stenosis site, expand the radius data of the patient's stenosis point to the normal level corresponding to the patient, and establish normal skeleton data based on the current patient.

In some embodiments, the model simulation module includes a lung function data acquisition unit and a simulation unit;

A lung function data acquisition unit, used to acquire lung function statistical data, i.e., lung function data of normal people;

The simulation unit is used to distribute the current patient's lung function statistical data in proportion to the branch outlets of each lung segment as input according to the segment lung function ratio theory, and obtain the pressure of each branch outlet of the normal model; then use the branch outlet pressure obtained from the normal model as input, and input it into each branch outlet corresponding to the stenosis model to obtain the total outlet flow of the stenosis model, that is, the simulation data of the stenosis model.

The embodiment of the present invention also provides an electronic device, which includes: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus. The communication bus can be used for information transmission between the electronic device and the sensor. The processor can call the logic instructions in the memory to execute the following method: Step 1: Acquire CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model; Step 2: Input the statistical data of lung function into the normal model for simulation, generate simulation data of the normal model, and then use the simulation data of the normal model as input to transplant it into the stenosis model for simulation, and generate simulation data of the stenosis model; Step 3: Input the current patient simulation data into the created lung function prediction model, and output the predicted value of the current lung function of the patient with large airway stenosis.

In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

An embodiment of the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the method provided by the above method embodiment, for example, including: Step 1: Acquire CT image data of patients with large airway stenosis and extract airway data from the image data, and use convolution surface technology to perform three-dimensional reconstruction to generate an airway model; the airway model includes a stenosis model and a normal model; Step 2: Input lung function statistical data into the normal model for simulation to generate simulation data of the normal model, and then use the simulation data of the normal model as input to transplant it into the stenosis model for simulation to generate simulation data of the stenosis model; Step 3: Input the current patient simulation data into the created lung function prediction model, and output the predicted value of the lung function of the current patient with large airway stenosis.

It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

The above-mentioned embodiments provide an embodiment of a CT-based device for predicting lung function for patients with large airway stenosis. The device and the CT-based method for predicting lung function for patients with large airway stenosis in the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the CT-based device for predicting lung function for patients with large airway stenosis, reference can be made to the above-mentioned embodiment of the method for predicting lung function for patients with large airway stenosis based on CT.

The CT-based lung function prediction device for patients with large airway stenosis is a unit and algorithm step of each example described in combination with the embodiments disclosed herein, and can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

Those skilled in the art will appreciate that various aspects of the method for predicting lung function in patients with large airway stenosis based on CT can be implemented as a system, method or program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software aspects, which can be collectively referred to as "circuit", "module" or "system" here.

Although the present invention has been described in detail with reference to the accompanying drawings and in combination with preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily conceive of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims (4)

1. The CT-based method for predicting the lung function of the patient with the large airway stenosis is characterized by comprising the following steps:

Acquiring CT image data of a patient with a large airway stenosis, extracting airway data from the image data, and performing three-dimensional reconstruction by using a convolution surface technology to generate an airway model; the airway model includes a stenosis model and a normal model;

Inputting the lung function statistical data into a normal model for simulation, generating simulation data of the normal model, and then transplanting the simulation data of the normal model into a narrow model for simulation by taking the simulation data of the normal model as input, so as to generate the simulation data of the narrow model;

inputting the current patient simulation data into the established lung function prediction model, and outputting a predicted value of the lung function of the current large airway stenosis patient;

the method for obtaining CT image data of a patient with a large airway stenosis and extracting airway data from the image data comprises the following steps of:

CT image data of a patient with a large airway stenosis is obtained, and airway data is extracted from the image data;

generating skeleton data according to the extracted airway data;

performing three-dimensional reconstruction by using a convolution surface technology according to the generated skeleton data to generate an airway model; the airway model comprises a stenosis model consistent with the current state of a large airway stenosis patient and a normal model expanded according to the current patient;

The step of obtaining CT image data of a patient with a large airway stenosis and extracting airway data from the image data comprises the steps of:

Acquiring lung CT image data of a patient with a large airway stenosis;

Selecting two seed points in the airway by utilizing the airway segmentation function of medical image modeling software; extracting an airway tree based on a region growing algorithm, and carrying out local smoothing and deltoiding treatment; creating an airway tree center path based on the current airway tree by utilizing the airway center line extraction function of software to generate an airway tree model;

The method comprises the steps of exporting an airway tree model into a first format file, exporting an airway tree central path into a second format file, wherein the first format file comprises airway tree surface grid coordinate data and a connection relation, and the second format file comprises airway tree central path point coordinate data and the connection relation, namely airway data;

the step of generating skeleton data from the extracted airway data includes:

classifying the airway tree according to the central path of the airway tree; wherein the main air pipe is of 0 level, the branch connected with 0 level is of 1 level, the branch connected with 1 level is of 2 level, and the like;

Connecting the head and tail points of each branch to serve as the central line of the current branch, making a perpendicular line from each point on the current branch to the central line of the current branch, forming a plane by each central line and the perpendicular line of the central line, obtaining the cross-sectional area of the plane intersecting with the current branch surface of the airway tree, and calculating the equivalent radius of each point on the current branch according to the cross-sectional area; searching the radius of the branch connected with each branching point, taking the midpoint of each branch as two end points, keeping the equivalent radius of the end points unchanged, and optimizing the equivalent radius of each branching point which is larger than a set value by utilizing linear transition control to serve as the radius of each point;

Fitting interpolation is carried out on the central path of the airway tree by using a fitting method, so that each branch section contains a set number of coordinate data points, and all the coordinate data points form airway central line data;

The central line data of the air passage and the radius of each point form skeleton data together;

the step of generating the airway model by three-dimensional reconstruction by utilizing a convolution surface technology according to the generated skeleton data comprises the following steps:

Modeling is carried out by utilizing skeleton data based on a convolution surface technology, and a large airway stenosis model in the current state of a patient, namely a stenosis model, is reconstructed;

Comparing the skeleton data with the airway statistics data of normal people, identifying a narrow part, expanding the radius data of the narrow point position of a patient to a normal level corresponding to the patient, and establishing normal skeleton data based on the current patient;

modeling is carried out by utilizing normal skeleton data based on a convolution surface technology, and an airway model of the health state of a patient, namely a normal model is reconstructed;

Inputting the lung function statistical data into a normal model for simulation, generating simulation data of the normal model, and transplanting the simulation data of the normal model into a narrow model for simulation by taking the simulation data of the normal model as input, wherein the step of generating the simulation data of the narrow model comprises the following steps:

acquiring lung function statistical data, namely lung function data of a normal person;

According to the lung function proportion theory of the lung segments, distributing the lung function statistical data of the current patient to each lung segment branch port in proportion as input to obtain the pressure intensity of each branch port of the normal model;

Taking the branch port pressure obtained by the normal model as input, and inputting the branch port pressure into each branch port corresponding to the narrow model to obtain the total outlet flow of the narrow model, namely simulation data of the narrow model;

The step of inputting the current patient simulation data into the created lung function prediction model and outputting the predicted value of the lung function of the current large airway stenosis patient comprises the following steps:

and training according to the historical large airway stenosis patient simulation data and the historical large airway stenosis patient lung function data to obtain a lung function prediction model.

2. The CT-based large airway stenosis patient lung function prediction device is characterized by comprising a model reconstruction module, a model simulation module and a lung function prediction module;

The model reconstruction module is used for acquiring CT image data of a patient with a large airway stenosis, extracting airway data from the image data, and performing three-dimensional reconstruction by using a convolution surface technology to generate an airway model; the airway model includes a stenosis model and a normal model;

the model simulation module is used for inputting the lung function statistical data into the normal model for simulation, generating simulation data of the normal model, and then transplanting the simulation data of the normal model into the narrow model for simulation by taking the simulation data of the normal model as input, so as to generate simulation data of the narrow model;

The lung function prediction module is used for inputting the current patient simulation data into the created lung function prediction model and outputting a predicted value of the lung function of the current large airway stenosis patient; training according to the simulation data of the historical large airway stenosis patient and the lung function data of the historical large airway stenosis patient to obtain a lung function prediction model;

the model reconstruction module comprises an airway data extraction unit, a skeleton data generation unit and a model reconstruction unit;

The airway data extraction unit is used for acquiring CT image data of a patient with a large airway stenosis and extracting airway data from the image data;

the framework data generating unit is used for generating framework data according to the extracted airway data;

the model reconstruction unit is used for performing three-dimensional reconstruction by utilizing a convolution surface technology according to the generated skeleton data to generate an airway model; the airway model comprises a stenosis model consistent with the current state of a large airway stenosis patient and a normal model expanded according to the current patient;

the airway data extraction unit comprises an image data acquisition sub-module, an airway tree model generation sub-module and an airway data generation sub-module;

the image data acquisition sub-module is used for acquiring lung CT image data of a patient with a large airway stenosis;

The airway tree model generation sub-module is used for selecting two seed points in the airway by utilizing the airway segmentation function of the medical image modeling software; extracting an airway tree based on a region growing algorithm, and carrying out local smoothing and deltoiding treatment; creating an airway tree center path based on the current airway tree by utilizing the airway center line extraction function of software to generate an airway tree model;

The airway data generation sub-module is used for exporting an airway tree model into a first format file and exporting an airway tree central path into a second format file, wherein the first format file comprises airway tree surface grid coordinate data and a connection relation, and the second format file comprises airway tree central path point coordinate data and the connection relation, namely airway data;

The skeleton data generation unit comprises an airway tree classification sub-module, a radius data acquisition sub-module and an airway center line data acquisition sub-module;

The airway tree grading submodule is used for grading the airway tree according to the central path of the airway tree; wherein the main air pipe is of 0 level, the branch connected with 0 level is of 1 level, the branch connected with 1 level is of 2 level, and the like;

The radius data acquisition sub-module is used for connecting the head and tail points of each branch to serve as the central line of the current branch, each point on the current branch is perpendicular to the central line of the current branch, each central line and the perpendicular line of the central line form a plane, the cross-sectional area of the plane intersecting with the surface of the current branch of the airway tree is acquired, and the equivalent radius of each point on the current branch is calculated according to the cross-sectional area; searching the radius of the branch connected with each branching point, taking the midpoint of each branch as two end points, keeping the equivalent radius of the end points unchanged, and optimizing the equivalent radius of each branching point which is larger than a set value by utilizing linear transition control to serve as the radius of each point;

the airway central line data acquisition sub-module is used for carrying out fitting interpolation on the airway tree central path by utilizing a fitting method, so that each section of branch comprises a set number of coordinate data points, and all the coordinate data points form airway central line data; the airway center line data and the radiuses of all points form skeleton data together;

the model reconstruction unit comprises a normal skeleton data establishment sub-module and a model reconstruction sub-module;

The model reconstruction sub-module is used for modeling by utilizing skeleton data based on a convolution surface technology and reconstructing a large airway stenosis model in the current state of a patient, namely a stenosis model; modeling is carried out by utilizing normal skeleton data based on a convolution surface technology, and an airway model of the health state of a patient, namely a normal model is reconstructed;

the normal skeleton data establishing sub-module is used for comparing the skeleton data with the airway statistics data of normal people, identifying a narrow part, expanding the radius data of the narrow point position of a patient to a normal level corresponding to the patient, and establishing normal skeleton data based on the current patient;

The model simulation module comprises a lung function data acquisition unit and a simulation unit;

A lung function data acquisition unit for acquiring lung function statistical data, i.e., lung function data of a normal person;

The simulation unit is used for distributing the lung function statistical data of the current patient to each lung segment branch port according to the lung segment lung function duty ratio theory as input to obtain the pressure intensity of each branch port of the normal model; and then taking the branch port pressure obtained by the normal model as input, and inputting the branch port pressure into each branch port corresponding to the narrow model to obtain the total outlet flow of the narrow model, namely the simulation data of the narrow model.

3. An electronic device, the electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; the memory stores computer program instructions executable by the at least one processor to enable the at least one processor to perform the CT-based large airway stenosis patient lung function prediction method of claim 1.

4. A non-transitory computer-readable storage medium storing computer instructions that cause the computer to perform the CT-based large airway stenosis patient lung function prediction method of claim 1.