CN112185551A - System and method for predicting coronary artery stenosis resistance based on deep learning - Google Patents
System and method for predicting coronary artery stenosis resistance based on deep learning Download PDFInfo
- Publication number
- CN112185551A CN112185551A CN202011081407.3A CN202011081407A CN112185551A CN 112185551 A CN112185551 A CN 112185551A CN 202011081407 A CN202011081407 A CN 202011081407A CN 112185551 A CN112185551 A CN 112185551A
- Authority
- CN
- China
- Prior art keywords
- model
- blood vessel
- stenosis
- resistance
- parameters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/50—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for simulation or modelling of medical disorders
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Epidemiology (AREA)
- Pathology (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Measuring Pulse, Heart Rate, Blood Pressure Or Blood Flow (AREA)
Abstract
A system and method for predicting coronary stenosis resistance based on deep learning. Belongs to the field of artificial intelligence, and comprises the following steps: constructing a blood vessel model based on the real coronary artery parameters and recording the parameters (entrance area, stenosis length and the like) of the blood vessel model; performing model gridding pretreatment based on the blood vessel model; calculating a preprocessing model based on geometric multi-scale hemodynamics and extracting stenosis resistance of the blood vessel model; establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters; establishing a neural network framework based on the BP neural network; training a narrow resistance training set and performing prediction verification on a prediction set based on deep learning.
Description
Technical Field
The invention belongs to the field of artificial intelligence, and relates to a system and a method for predicting coronary artery stenosis resistance based on deep learning.
Background
FFR is a reliable indicator directly describing the degree of functional myocardial ischemia, and can be obtained by numerical simulation in related studies, and the determination of coronary artery stenosis resistance is particularly important for numerical calculation of FFR. The coronary artery stenosis resistance determination method is intended to provide a calculation basis for non-invasively obtaining the accuracy of the FFR (defined as the ratio of the mean pressure at the distal end of the stenosis (Pd) to the mean pressure at the root of the aorta (Pa)) value.
Machine learning is a multi-field interdiscipline, deep learning is a new research direction in the field of machine learning, is an internal rule and a representation level of learning sample data, and finally aims to enable a machine to have the analysis and learning capacity like a human and to recognize data such as characters, images and shapes. The method combines deep learning and numerical simulation to provide an algorithm for predicting coronary artery stenosis resistance based on deep learning, so that the stenosis resistance is determined based on the coronary artery geometric parameters, and the accuracy of the FFR numerical simulation result is improved in noninvasive calculation. Has certain theoretical value for the research of diagnosis and treatment strategies of individual coronary artery physiological diseases.
Disclosure of Invention
The invention provides an algorithm for predicting coronary artery stenosis resistance based on deep learning, which is quicker and more accurate than numerical simulation and is generally suitable for determining stenosis resistance calculation generated by vascular stenosis. The algorithm for predicting coronary artery stenosis resistance based on deep learning comprises the following steps: the method comprises the steps of construction of a blood vessel model, model grid preprocessing, target parameter extraction, database establishment, neural network establishment and prediction verification.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a system for predicting coronary artery stenosis resistance based on deep learning is characterized in that the system is used for predicting the coronary artery stenosis resistance by the following method, and comprises the following steps:
step A1: constructing a coronary vessel model based on the real coronary anatomical structure parameters and recording the parameters (such as entrance area, narrow length, narrow area and the like) of the coronary vessel model;
step A2: performing model gridding pretreatment based on the blood vessel model;
step A3: calculating a preprocessing model based on geometric multi-scale hemodynamics and extracting stenosis resistance of a blood vessel model;
a4, establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters;
step A5: establishing a neural network framework based on the BP neural network;
step A6: training a narrow resistance training set and performing prediction verification on a prediction set based on deep learning.
1. As the technical scheme of the invention, the characteristic in the step A1 is that firstly, a coronary vessel model is constructed based on the real coronary anatomical structure parameters, and the coronary vessel model is constructed by Sold Work software. The requirements for constructing the model are as follows: the shape parameters such as entrance area, blood vessel length, stenosis rate, etc. should conform to the parameters of the real blood vessel. For example: a vessel diameter D (2-6mm), a stenosis length ls (2-40mm), a stenosis rate DS% (20-90%), a stenosis diameter DS (1-DS%), a vessel length L ═ ls + D8 + D12);
as a further technical solution of the present invention, the feature described in step a2 is that model meshing preprocessing is performed based on the blood vessel model, and the input form of the software-supported mesh file is calculated, so that the following preprocessing steps are performed before the model meshing:
step B1: deriving an x _ t format for the constructed blood vessel model through Sold Works software;
step B2: and performing pre-calculation meshing on the constructed blood vessel model through an Ansys workbench 15.0 meshing sub-function module, wherein the step requires sensitivity analysis on meshes, and eliminates the influence of mesh factors on calculation results.
As a further technical scheme of the invention, the characteristics in the step A3 are used for extracting the stenosis resistance of the blood vessel model based on a multi-scale haemodynamics calculation preprocessing model. The specific implementation process comprises the following steps:
step C1: giving the same resting flow to determine a corresponding zero-dimensional afterload model, wherein the zero-dimensional afterload model consists of basic electronic component inductors and resistors, and the equivalent relation between the hemodynamic parameters and the physical electrical parameters is shown in table 1.
TABLE 1 hemodynamic parameters and physical electrical parameters equivalence relation
Assuming that the length of a section of blood vessel is l, the cross-sectional area of the blood vessel is a, and the zero-dimensional and three-dimensional corresponding parameter relationship of a section of blood vessel is shown in fig. 5, according to the elastic cavity theory established by Frank, the quantitative relationship among the flow, pressure and related resistance on the section of blood vessel can be expressed as follows:
Δp＝Q*R′#(1)
in the formula, Δ P represents the pressure drop generated in the section of blood vessel, Q represents the blood flow of the branch blood vessel, and R' represents the viscous resistance generated by the section of blood vessel; mu is the blood viscosity value under normal physiology, and the value is 3.5e-3 Pa.s; l represents the length of the section of the blood vessel, and A represents the cross-sectional area of the section of the blood vessel.
The self-inductivity of the inductor is equivalent to the flow inertia of blood flow, and the resistance effect of the resistor is equivalent to the microcirculation resistance of each branch of the coronary artery. The values of the resistance R and the inductance L in the zero-dimensional model are:
in the formula PPowderRepresents the pressure at the end of the section of the blood vessel, and rho represents the blood flow density value and is defined as 1060kg/m3。
Step C2: introducing the gridding model into Ansys-CFX fluid calculation software, and giving fluid calculation parameters including blood flow density and blood flow dynamic viscosity value (namely the blood viscosity value under normal physiology); (ii) a
Step C3: determining boundary conditions and calculating the pressure-flow-afterload relation of three-dimensional model, defining average arterial pressure at the blood vessel inlet, defining virtual flow boundary at the outlet, and coupling zero-dimensional afterload model
Step C4: in the numerical simulation calculation, the setting time is 1.2s, the setting time step length is 0.01s, and the simulation calculation process is executed.
Step C5: after the calculation is finished, the blood vessel pressure is extracted at the position 30mm behind the stenosis, the pressure difference of the pretreatment model is obtained through comparison with the blood vessel inlet pressure, and the stenosis resistance is obtained through dividing the pressure difference by the flow. And extracting blood flow at the stenotic inlet.
As a further technical scheme of the invention, the characteristics in the step A4 are used for establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters; 340 different stenosis parameter blood vessels are calculated through steps A1, A2 and A3, the extracted stenosis resistance and blood flow of the blood vessel model and the geometric parameters of the blood vessel model construct a deep learning data set, and the data set is divided into a training set and a testing set.
As a further technical solution of the present invention, the characteristic described in step a5, establishing a neural network framework based on a BP neural network, specifically includes the following steps:
step D1: establishing a deep learning hidden layer, and setting the number of hidden layers, the number of nodes of the hidden layer, an activation function and the like;
step D2: selecting proper optimizers, Loss functions and other hyper-parameters.
As a further technical scheme of the invention, the characteristics in the step A5 are based on deep learning to train the stenosis resistance training set and to predict and verify the prediction set. The specific implementation process comprises the following steps:
step E1: importing a training set and a testing set into a program through a pandas module;
step E2: the influence of the geometric parameters of the blood vessel on the stenosis resistance of the blood vessel model is analyzed through a correlation coefficient matrix, multiple collinearity among variables is analyzed and avoided, and finally seven variables with strong correlation are obtained, wherein the variables are respectively the entrance area, the stenosis entrance length, the stenosis outlet length, the minimum stenosis length, the stenosis rate and the entrance blood flow.
Step E3: and importing the processed data into a neural network framework to adjust the hyper-parameters.
Step E4: and saving the training model and executing the prediction of the stenosis resistance of the blood vessel model.
The system of the invention can predict coronary artery stenosis resistance accurately with small error.
Description of the drawings:
FIG. 1 is a flow chart of the method of the present invention
FIG. 2: blood vessel model diagram
FIG. 3: zero-dimensional three-dimensional coupling model diagram
FIG. 4: parameter matrix correlation analysis
FIG. 5: neural network architecture
FIG. 6: comparison graph of predicted value and calculated value
Detailed Description
The present invention will be further illustrated with reference to the following examples, but the present invention is not limited to the following examples.
Example 1
A system for predicting coronary artery stenosis resistance based on deep learning is used for predicting the coronary artery stenosis resistance through the following method, and comprises the following steps:
step A1: constructing a coronary vessel model based on the real coronary anatomical structure parameters and recording the coronary vessel model parameters (entrance area, stenosis entrance length, stenosis exit length, minimum stenosis length, stenosis rate);
step A2: performing model gridding pretreatment based on the blood vessel model;
step A3: calculating a preprocessing model based on geometric multi-scale hemodynamics and extracting stenosis resistance of a blood vessel model;
a4, establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters;
step A5: establishing a neural network framework based on the BP neural network;
step A6: training a narrow resistance training set and performing prediction verification on a prediction set based on deep learning.
As the technical scheme of the invention, the characteristic in the step A1 is that firstly, a coronary vessel model is constructed based on the real coronary anatomical structure parameters, and the coronary vessel model is constructed by Sold Work software. The requirements for constructing the model are as follows: the shape parameters such as entrance area, blood vessel length, stenosis rate, etc. should conform to the parameters of the real blood vessel. For example: blood vessel diameter D (2-6mm), stenosis length ls (2-40mm), stenosis rate DS (20-90%), stenosis diameter Ds (1-DS%), blood vessel length L (ls + D8 + D12)
As a further technical solution of the present invention, the feature described in step a2 is that model meshing preprocessing is performed based on the blood vessel model, and the input form of the software-supported mesh file is calculated, so that the following preprocessing steps are performed before the model meshing:
step B1: deriving an x _ t format for the constructed blood vessel model through Sold Works software;
step B2: and performing pre-calculation meshing on the constructed blood vessel model through an Ansys workbench 15.0 meshing sub-function module, wherein the step requires sensitivity analysis on meshes, and eliminates the influence of mesh factors on calculation results.
As a further technical scheme of the invention, the characteristics in the step A3 are used for extracting the stenosis resistance of the blood vessel model based on a multi-scale haemodynamics calculation preprocessing model. The specific implementation process comprises the following steps:
step C1: giving the same resting flow to determine a corresponding zero-dimensional afterload model, wherein the zero-dimensional afterload model consists of basic electronic component inductors and resistors, and the equivalent relation between the hemodynamic parameters and the physical electrical parameters is shown in table 1.
TABLE 1 hemodynamic parameters and physical electrical parameters equivalence relation
Assuming that the length of a section of blood vessel is l, the cross-sectional area of the blood vessel is a, and the zero-dimensional and three-dimensional corresponding parameter relationship of a section of blood vessel is shown in fig. 3, according to the elastic cavity theory established by Frank, the quantitative relationship among the flow, pressure and related resistance on the section of blood vessel can be expressed as follows:
Δp＝Q*R′#(1)
in the formula, Δ P represents the pressure drop generated in the section of blood vessel, Q represents the blood flow of the branch blood vessel, and R' represents the viscous resistance generated by the section of blood vessel; mu is the blood viscosity value under normal physiology, and the value is 3.5e-3 Pa.s; l represents the length of the section of the blood vessel, and A represents the cross-sectional area of the section of the blood vessel.
The self-inductivity of the inductor is equivalent to the flow inertia of blood flow, and the resistance effect of the resistor is equivalent to the microcirculation resistance of each branch of the coronary artery. The values of the resistance R and the inductance L in the zero-dimensional model are:
in the formula PPowderRepresents the pressure at the end of the section of the blood vessel, and rho represents the blood flow density value and is defined as 1060kg/m3。
Step C2: introducing the gridding model into Ansys-CFX fluid calculation software, and giving fluid calculation parameters including blood flow density and blood flow dynamic viscosity values;
step C3: determining boundary conditions and calculating the pressure-flow-afterload relation of the three-dimensional model, giving mean arterial pressure at a fluid inlet, defining a virtual flow boundary at an outlet, and coupling the zero-dimensional afterload model.
Step C4: in the numerical simulation calculation, the setting time is 1.2s, the setting time step length is 0.01s, and the simulation calculation process is executed.
Step C5: after the calculation is finished, the blood vessel pressure is extracted at the position 30mm behind the stenosis, the pressure difference of the pretreatment model is obtained through comparison with the inlet pressure, and the stenosis resistance is obtained through dividing the pressure difference by the flow. And extracting blood flow at the stenotic inlet.
As a further technical scheme of the invention, the characteristics in the step A4 are used for establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters; 340 different stenosis parameter blood vessels are calculated through steps A1, A2 and A3, the extracted stenosis resistance and blood flow of the blood vessel model and the geometric parameters of the blood vessel model construct a deep learning data set, and the data set is divided into a training set and a testing set.
As a further technical solution of the present invention, the characteristic described in step a5, establishing a neural network framework based on a BP neural network, specifically includes the following steps:
step D1: step D3: establishing a deep learning hidden layer and selecting a Dense layer of Keras, setting an input layer to be 7 variables and an output layer to be 1 narrow resistance, setting the hidden layer to be 7 layers, and using an activation function tanh function
Step D2: selecting an adam optimizer to adjust the hyper-parameters such as lr, beta _1, beta _2, epsilon and the like, and selecting an MSE (mean square error) as a Loss function:
as a further technical scheme of the invention, the characteristics in the step A5 are based on deep learning to train the stenosis resistance training set and to predict and verify the prediction set. The specific implementation process comprises the following steps:
step D1: importing a training set and a testing set into a program through a pandas module;
step D2: the influence of the geometric parameters of the blood vessel on the stenosis resistance of the blood vessel model is analyzed through a correlation coefficient matrix, multiple collinearity among variables is analyzed and avoided, and finally seven variables with strong correlation are obtained, wherein the variables are respectively the entrance area, the stenosis entrance length, the stenosis outlet length, the minimum stenosis length, the stenosis rate and the entrance blood flow.
Step D3: importing the processed data into a neural network framework to adjust parameters such as batch _ size, epochs, estimation _ split, estimation _ freq and the like.
Step D5: and saving the training model and executing the prediction of the stenosis resistance of the blood vessel model.
The invention aims to realize the prediction of the stenosis resistance of a stenosis vessel, numerical simulation calculation is carried out by using an ideal model when a data set is established to obtain the stenosis resistance, real case CTA data is used for model reconstruction when the data set is actually applied to carry out the extraction of the vessel length, the inlet area, the stenosis inlet length, the stenosis rate, the stenosis minimum length, the stenosis outlet length and the vessel flow parameters of the stenosis vessel, and then the stenosis resistance of the stenosis vessel is predicted by using a neural network model to obtain a non-invasive FFR value. The method carries out MSE (mean square error) analysis and verification on the narrow resistance obtained by numerical simulation and the narrow resistance predicted by the neural network. The final MSE result is 3.06%, and the MSE result meets the application standard of less than 5%.
Claims (8)
1. A system for predicting coronary artery stenosis resistance based on deep learning is characterized in that the system is used for predicting the coronary artery stenosis resistance by the following method, and comprises the following steps:
step A1: constructing a coronary vessel model based on the real coronary anatomical structure parameters and recording the parameters (such as entrance area, narrow length, narrow area and the like) of the coronary vessel model;
step A2: performing model gridding pretreatment based on the blood vessel model;
step A3: calculating a preprocessing model based on geometric multi-scale hemodynamics and extracting stenosis resistance of a blood vessel model;
step A4: establishing a stenosis resistance training set and a prediction set based on data extraction and blood vessel model parameters;
step A5: establishing a neural network framework based on the BP neural network;
step A6: training a narrow resistance training set and performing prediction verification on a prediction set based on deep learning.
2. The system for predicting coronary artery stenosis resistance based on deep learning of claim 1, wherein the features of step a1 are obtained by first constructing a coronary vessel model based on the real coronary anatomy parameters, constructing a coronary vessel model by using a Sold word software, and constructing the model so that the shape parameters such as the requirement of the model should be in accordance with the parameters of the real vessel.
3. The system for predicting coronary artery stenosis resistance based on deep learning as claimed in claim 1, wherein the features in step a2 are pre-processed by model meshing based on the vessel model, the input form of the mesh file is supported by the computing software, and therefore the following pre-processing steps are performed before the model meshing:
step B1: deriving an x _ t format for the constructed blood vessel model through Sold Works software;
step B2: and performing pre-calculation meshing on the constructed blood vessel model through an Ansys workbench 15.0 meshing sub-function module, wherein the step requires sensitivity analysis on meshes, and eliminates the influence of mesh factors on calculation results.
4. The system for predicting coronary artery stenosis resistance based on deep learning of claim 1, wherein the features of step a3 are used for extracting the stenosis resistance of the blood vessel model based on a multi-scale hemodynamic computation preprocessing model; the specific implementation process comprises the following steps:
step C1: giving the same resting flow to determine a corresponding zero-dimensional afterload model, wherein the zero-dimensional afterload model consists of basic electronic component inductors and resistors;
assuming a segment of a blood vessel with a length l and a cross-sectional area a, the quantitative relationship between flow, pressure and associated resistance over the segment of the blood vessel can be expressed as:
Δp＝Q*R′#(1)
in the formula, Δ P represents the pressure drop generated in the section of blood vessel, Q represents the blood flow of the branch blood vessel, and R' represents the viscous resistance generated by the section of blood vessel; mu is the blood viscosity value under normal physiology, and the value is 3.5e-3 Pa.s; l represents the length of the section of the blood vessel, and A represents the cross-sectional area of the section of the blood vessel;
the self-inductivity of the inductor is equivalent to the flow inertia of blood flow, and the resistance effect of the resistor is equivalent to the microcirculation resistance of each branch of the coronary artery. The values of the resistance R and the inductance L in the zero-dimensional model are:
in the formula PPowderRepresents the pressure at the end of the section of the blood vessel, and rho represents the blood flow density value and is defined as 1060kg/m3。
Step C2: introducing the gridding model into Ansys-CFX fluid calculation software, and giving fluid calculation parameters including blood flow density and blood flow dynamic viscosity value (namely the blood viscosity value under normal physiology);
step C3: determining boundary conditions and calculating the pressure-flow-afterload relation of the three-dimensional model, giving mean arterial pressure at the entrance of a blood vessel, defining a virtual flow boundary at the exit, and coupling a zero-dimensional afterload model;
step C4: in the numerical simulation calculation, setting time to be 1.2s and setting time step length to be 0.01s, and executing a simulation calculation process;
step C5: after calculation, extracting blood vessel pressure at a position 30mm behind the stenosis, obtaining the pressure difference of the pretreatment model through comparison with the blood vessel inlet pressure, and obtaining the stenosis resistance through dividing the pressure difference by the flow; and extracting blood flow at the stenotic inlet.
5. The system for predicting coronary artery stenosis resistance based on deep learning of claim 1, wherein the features of step a4 are used to establish a stenosis resistance training set and a prediction set based on data extraction and vessel model parameters; 340 different stenosis parameter blood vessels are calculated through steps A1, A2 and A3, the extracted stenosis resistance and blood flow of the blood vessel model and the geometric parameters of the blood vessel model construct a deep learning data set, and the data set is divided into a training set and a testing set.
6. The system for predicting coronary artery stenosis resistance based on deep learning as claimed in claim 1, wherein the step a5 of building a neural network framework based on the BP neural network comprises the following steps:
step D1: establishing a deep learning hidden layer, and setting the number of hidden layers, the number of nodes of the hidden layer, an activation function and the like;
step D2: selecting a proper optimizer, a Loss function and other hyper-parameters;
and B, training a stenosis resistance training set and performing prediction verification on a prediction set based on the characteristics in the step A5. The specific implementation process comprises the following steps:
step F1: importing a training set and a testing set into a program through a pandas module;
step E2: analyzing the influence of the geometric parameters of the blood vessel on the stenosis resistance of the blood vessel model through a correlation coefficient matrix, analyzing and avoiding multiple collinearity among variables, and finally obtaining seven variables with strong correlation, namely an inlet area, a stenosis inlet length, a stenosis outlet length, a minimum stenosis length, a stenosis rate and an inlet blood flow;
step E3: importing the processed data into a neural network framework to adjust the hyper-parameters;
step E4: and saving the training model and executing the prediction of the stenosis resistance of the blood vessel model.
7. The system for predicting coronary artery stenosis resistance based on deep learning of claim 6, wherein the neural network framework is established based on a BP neural network, and the implementation comprises the following steps:
step D1: establishing a deep learning hidden layer and selecting a Dense layer of Keras, setting an input layer to be 7 variables and an output layer to be 1 narrow resistance, setting the hidden layer to be 7 layers, and using an activation function tanh function
Step D2: selecting an adam optimizer to adjust the hyper-parameters such as lr, beta _1, beta _2, epsilon and the like, and selecting an MSE (mean square error) as a Loss function:
8. the system for predicting coronary artery stenosis resistance based on deep learning of claim 6, wherein the features of step a5 are trained on a stenosis resistance training set and verified on a prediction set based on deep learning; the specific implementation process comprises the following steps:
step D1: importing a training set and a testing set into a program through a pandas module;
step D2: analyzing the influence of the geometric parameters of the blood vessel on the stenosis resistance of the blood vessel model through a correlation coefficient matrix, analyzing and avoiding multiple collinearity among variables, and finally obtaining seven variables with strong correlation, namely an inlet area, a stenosis inlet length, a stenosis outlet length, a minimum stenosis length, a stenosis rate and an inlet blood flow;
step D3: importing the processed data into a neural network framework to adjust parameters such as batch _ size, epochs, estimation _ split, estimation _ freq and the like;
step D5: and saving the training model and executing the prediction of the stenosis resistance of the blood vessel model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011081407.3A CN112185551B (en) | 2020-10-10 | 2020-10-10 | System and method for predicting coronary artery stenosis resistance based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011081407.3A CN112185551B (en) | 2020-10-10 | 2020-10-10 | System and method for predicting coronary artery stenosis resistance based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112185551A true CN112185551A (en) | 2021-01-05 |
| CN112185551B CN112185551B (en) | 2023-04-07 |
Family
ID=73948027
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011081407.3A Active CN112185551B (en) | 2020-10-10 | 2020-10-10 | System and method for predicting coronary artery stenosis resistance based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112185551B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113128139A (en) * | 2021-04-21 | 2021-07-16 | 北京工业大学 | Method and system for rapidly calculating fractional flow reserve based on simplified coronary artery zero-dimensional model and stenosis resistance prediction model |
| CN116453697B (en) * | 2022-12-30 | 2023-12-19 | 徐州医科大学 | Coronary artery stenosis hemodynamic simulation method and system based on FFR fitting |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106456078A (en) * | 2013-10-17 | 2017-02-22 | 西门子保健有限责任公司 | Method and system for machine learning based assessment of fractional flow reserve |
| CN109256205A (en) * | 2017-07-12 | 2019-01-22 | 西门子保健有限责任公司 | Method and system for the clinical decision support carried out using local and remote analytics |
| EP3564963A1 (en) * | 2018-05-02 | 2019-11-06 | Siemens Healthcare GmbH | System and methods for fast computation of computed tomography based fractional flow reserve |
| CN110731789A (en) * | 2019-10-25 | 2020-01-31 | 上海杏脉信息科技有限公司 | coronary artery microvascular resistance index acquisition method, medium and equipment |
| CN111356406A (en) * | 2017-10-06 | 2020-06-30 | 埃默里大学 | Method and system for determining hemodynamic information for one or more arterial segments |
-
2020
- 2020-10-10 CN CN202011081407.3A patent/CN112185551B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106456078A (en) * | 2013-10-17 | 2017-02-22 | 西门子保健有限责任公司 | Method and system for machine learning based assessment of fractional flow reserve |
| CN109256205A (en) * | 2017-07-12 | 2019-01-22 | 西门子保健有限责任公司 | Method and system for the clinical decision support carried out using local and remote analytics |
| CN111356406A (en) * | 2017-10-06 | 2020-06-30 | 埃默里大学 | Method and system for determining hemodynamic information for one or more arterial segments |
| EP3564963A1 (en) * | 2018-05-02 | 2019-11-06 | Siemens Healthcare GmbH | System and methods for fast computation of computed tomography based fractional flow reserve |
| CN110444275A (en) * | 2018-05-02 | 2019-11-12 | 西门子医疗有限公司 | System and method for quickly calculating blood flow reserve score |
| CN110731789A (en) * | 2019-10-25 | 2020-01-31 | 上海杏脉信息科技有限公司 | coronary artery microvascular resistance index acquisition method, medium and equipment |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113128139A (en) * | 2021-04-21 | 2021-07-16 | 北京工业大学 | Method and system for rapidly calculating fractional flow reserve based on simplified coronary artery zero-dimensional model and stenosis resistance prediction model |
| CN113128139B (en) * | 2021-04-21 | 2024-03-29 | 北京工业大学 | Method for calculating FFR (fringe field switching) based on coronary artery zero-dimensional model and stenosis resistance prediction model |
| CN116453697B (en) * | 2022-12-30 | 2023-12-19 | 徐州医科大学 | Coronary artery stenosis hemodynamic simulation method and system based on FFR fitting |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112185551B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109965862B (en) | Construction method of sleeveless long-time continuous blood pressure non-invasive monitoring model | |
| CN106980899A (en) | The deep learning model and system of flow characteristic on prediction vascular tree blood flow paths | |
| CN109559326B (en) | Hemodynamic parameter calculation method and system and electronic device | |
| US11871995B2 (en) | Patient-specific modeling of hemodynamic parameters in coronary arteries | |
| CN107123159A (en) | blood flow state analysis system and method | |
| CN106650267B (en) | System for calculating fractional flow reserve and method for setting boundary conditions | |
| CN112185551B (en) | System and method for predicting coronary artery stenosis resistance based on deep learning | |
| CN105249954A (en) | Method and system for prediction of post-stenting hemodynamic metrics for treatment planning of arterial stenosis | |
| CN105474219A (en) | Systems and methods for estimating physiological heart measurements from medical images and clinical data | |
| CA3126313C (en) | Patient-specific modeling of hemodynamic parameters in coronary arteries | |
| CN108665449A (en) | Predict the deep learning model and device of the flow characteristic in blood flow vector path | |
| CN111241759A (en) | FFR rapid calculation system model based on zero-dimensional hemodynamic model | |
| CN113015497A (en) | Method and device for simulating blood flow of blood vessel inherent to object | |
| CN114947910A (en) | Coronary artery end microvascular resistance calculation method and FFR calculation method and system | |
| CN114530252A (en) | Coronary artery blood flow dynamics simulation method and device | |
| CN113128139B (en) | Method for calculating FFR (fringe field switching) based on coronary artery zero-dimensional model and stenosis resistance prediction model | |
| Nair et al. | Hemodynamics in patients with aortic coarctation: A comparison of in vivo 4D-flow MRI and FSI simulation | |
| CN110457765B (en) | Method for calculating odd vein shunt ratio in half-ventricle model based on geometric multi-scale model | |
| EP3438989A1 (en) | Method and apparatus for predicting fluid flow through a subject conduit | |
| EP3770921A1 (en) | Method and system for pressure autoregulation based synthesizing of photoplethysmogram signal | |
| JPWO2019168474A5 (en) | ||
| CN114947909A (en) | Method and system for calculating FFR (flow field noise ratio) based on blood flow ratio before and after stenosis | |
| CN110742688A (en) | Blood vessel model establishing method and device and readable storage medium | |
| CN113936805A (en) | Method for calculating coronary artery geometric multi-scale model based on fluid-solid coupling | |
| CN111833434A (en) | Method for simply and reliably determining left and right coronary artery shunt fraction and method for calculating FFR (fractional flow rate) |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |