Hemodynamics-based lung drainage basin analysis method

Abstract

The invention relates to the technical field of medical image analysis methods, in particular to a lung drainage basin analysis method based on hemodynamics, which comprises the following steps: acquiring lung CT image data; acquiring lung anatomy data and clinical medical knowledge; generating a pulmonary artery blood vessel tree model based on the lung CT image data; establishing a hemodynamic model according to the pulmonary artery blood vessel tree model; calculating a blood flow velocity field and a pressure gradient field based on the hemodynamic model; according to the pressure gradient field, executing a watershed watershed segmentation algorithm to obtain a lung watershed division result; according to the method, an analysis report containing a lung drainage basin division result is generated, the segmentation precision of a lung blood vessel network is greatly improved by adopting an advanced image processing technology and a blood vessel tracking algorithm, and the method can more accurately reflect individual differences and blood flow characteristics in a pathological state by dynamically adjusting boundary conditions and blood parameters. And meanwhile, an efficient numerical solution algorithm is adopted, so that the calculation complexity is remarkably reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image analysis methods, and more specifically, to a pulmonary flow domain analysis method based on hemodynamics. Background Art

[0002] With the rapid development of medical imaging technology and computer-aided diagnosis, the diagnosis and treatment of lung diseases have made significant progress. However, for complex pulmonary hemodynamic analysis, existing technologies still have many limitations. Traditional lung imaging analysis mainly relies on static images, which are difficult to accurately reflect the dynamic characteristics of blood flow. Although some blood flow imaging technologies based on CT or MRI have emerged in recent years, these methods can often only provide local or fragmentary blood flow information and cannot comprehensively evaluate the blood flow status of the entire lung.

[0003] At present, the closest existing technology usually uses computational fluid dynamics (CFD) simulation combined with medical image segmentation to analyze pulmonary blood flow. Although this method can provide detailed blood flow information in theory, it faces many challenges in practical applications. First, the existing vascular segmentation algorithms often have errors when dealing with complex pulmonary vascular networks, especially for the identification and tracking of small blood vessels. Secondly, the boundary conditions and blood parameters used in CFD simulations are often based on empirical values ​​or simplified assumptions, which are difficult to accurately reflect individual differences and pathological conditions. In addition, the existing methods have high computational complexity and long time consumption in the construction and solution of hemodynamic models, which makes it difficult to meet the needs of clinical real-time analysis.

[0004] More importantly, existing technologies are clearly insufficient in converting hemodynamic analysis results into clinically meaningful diagnostic information. Most methods only provide basic parameters such as blood flow velocity and pressure, and lack systematic identification and classification of abnormal blood flow patterns. At the same time, existing analysis methods often ignore the multi-scale characteristics of pulmonary blood flow, making it difficult to simultaneously analyze large vessels and microcirculation.

[0005] In addition, existing technologies generally lack effective verification mechanisms. Due to the lack of reliable methods to directly measure blood flow in the lungs, the accuracy of CFD simulation results is difficult to fully verify. This seriously limits the application and promotion of these methods in clinical practice.

[0006] In view of the above problems, a new method is urgently needed to comprehensively, accurately and efficiently analyze the hemodynamic characteristics of the lungs. This method should be able to overcome the shortcomings of existing technologies in terms of vascular segmentation, model construction, and abnormality identification, while providing a reliable verification mechanism to meet the needs of clinical diagnosis and research. Summary of the invention

[0007] The present invention aims at these limitations of the existing technology and proposes a pulmonary flow analysis method based on hemodynamics. This method realizes a comprehensive and in-depth analysis of pulmonary blood flow by innovatively combining high-precision vessel segmentation, adaptive CFD simulation, multi-scale flow division and abnormal blood flow identification.

[0008] In order to solve the above technical problems, the present invention adopts the following technical solutions: Pulmonary flow analysis methods based on hemodynamics, including: The acquisition steps include: Acquire lung CT image data; Obtain lung anatomy data and clinical medical knowledge; Processing steps include: Based on the lung CT image data, generating a pulmonary artery vascular tree model; Establishing a hemodynamic model according to the pulmonary artery vascular tree model; Based on the hemodynamic model, calculating the blood flow velocity field and the pressure gradient field; According to the pressure gradient field, a watershed watershed segmentation algorithm is executed to obtain a lung watershed segmentation result; Output steps include: An analysis report including the lung watershed division results is generated.

[0009] Preferably, generating the pulmonary artery vascular tree model specifically includes: reconstructing the lung CT image data into three-dimensional data; extracting pulmonary blood vessels based on the three-dimensional data; According to the pulmonary anatomical data, identifying the bifurcation of the main pulmonary artery; Based on the bifurcation of the pulmonary artery trunk, a pulmonary artery vascular tree structure is generated.

[0010] Preferably, the establishing of the hemodynamic model specifically includes: Set blood physical parameters, including blood density, dynamic viscosity, static viscosity, volume fraction and thermal conductivity; Set the wall conditions, flow field boundary conditions, and lung boundary conditions; Based on Darcy's law, LM equation, continuity equation and Bernoulli equation, the hemodynamic equations were constructed.

[0011] Preferably, the calculating of the blood flow velocity field and the pressure gradient field specifically includes: Dividing the computational domain of the hemodynamic model into two parts, upstream and downstream; applying an inlet boundary condition to the upstream boundary; applying an outlet boundary condition to the downstream boundary; The hemodynamic equations are solved using a computational fluid dynamics (CFD) algorithm to obtain a blood flow velocity field and a pressure gradient field.

[0012] Preferably, the executing of the watershed segmentation algorithm specifically includes: Setting a seed point in the pressure gradient field; Based on the seed point, executing a region growing algorithm; Determine the watershed boundary based on the pressure gradient threshold; Generate lung watershed partition results.

[0013] Preferably, the method further comprises the step of abnormal blood flow identification: Calculate outlier discrimination parameters, including flow, density, velocity, vessel radius, and vessel length; Set the abnormal judgment threshold; Based on the abnormal point discrimination parameter and the abnormal judgment threshold, identifying an abnormal blood flow pattern; The abnormal blood flow pattern is added to the analysis report.

[0014] Preferably, the abnormal blood flow pattern includes: When the flow rate is less than the first preset threshold and the density is less than the second preset threshold, it is determined to be an embolism point; When the density is less than the second preset threshold and the blood vessel radius is less than the third preset threshold, it is determined to be a stenosis point; When the density is less than the second preset threshold and the blood vessel radius is greater than the fourth preset threshold, it is determined to be a dilation point.

[0015] As a preferred embodiment, it also includes a dynamic adjustment step: Statistical analysis of velocity distribution factor, pressure drop factor and resistance gradient factor of normal blood flow in the current flow area; calculating a probability distribution of the factor; Based on the maximum value point of the probability distribution, the threshold of the watershed segmentation algorithm is dynamically adjusted.

[0016] Preferably, the lung watershed division result includes: First-order watershed: the area including the right ventricular pulmonary artery; Second-level basin: a branch of the first-level basin; Third-level basin: a branch of the second-level basin; Fourth-level basin: a branch of the third-level basin.

[0017] Preferably, the step of verifying the result is also included: Use dynamic data of fluorescent tracers to locate and quantitatively measure blood flow distribution; Calculating actual blood viscosity based on the dynamic data of the fluorescent tracer; comparing the actual blood viscosity with the blood viscosity parameter in the hemodynamic model; According to the comparison results, the parameters of the hemodynamic model are adjusted.

[0018] The method of the present invention has the following significant technical effects: First, by using advanced image processing technology and vascular tracking algorithms, the segmentation accuracy of the pulmonary vascular network has been greatly improved, especially in the identification of small blood vessels, which has laid a solid foundation for subsequent hemodynamic analysis.

[0019] Secondly, the present invention innovatively introduces adaptive CFD simulation technology. By dynamically adjusting boundary conditions and blood parameters, this method can more accurately reflect individual differences and blood flow characteristics under pathological conditions. At the same time, the use of efficient numerical solution algorithms significantly reduces the computational complexity, making real-time analysis possible.

[0020] More importantly, the multi-scale watershed division method proposed in this paper cleverly solves the problem of unifying macrovascular and microcirculatory analysis. By dividing the pulmonary vascular network into four levels of watersheds, this method can capture hemodynamic characteristics at different scales, providing a new perspective for the comprehensive assessment of pulmonary blood flow conditions.

[0021] In terms of abnormal blood flow identification, the present invention has developed a systematic parameter system and judgment mechanism that can automatically identify and classify a variety of abnormal blood flow patterns, such as embolism, stenosis, dilation, etc. This greatly improves the clinical value of the analysis results and provides important support for early diagnosis and precise treatment.

[0022] The present invention also innovatively introduces a verification mechanism based on fluorescent tracers. By comparing and optimizing the CFD simulation results with the actual measurement data, the reliability and accuracy of the analysis results are significantly improved. This not only solves the verification difficulty problem in the prior art, but also provides a way for the continuous optimization of the model.

[0023] In summary, the hemodynamic-based pulmonary flow analysis method proposed in the present invention realizes a comprehensive, accurate and efficient analysis of pulmonary blood flow through the organic combination of multiple innovative technologies. This method not only overcomes many limitations of existing technologies, but also has achieved significant improvements in accuracy, efficiency, clinical practicality, etc. This provides a powerful tool for the diagnosis, treatment evaluation and research of lung diseases, and is expected to play an important role in clinical practice and promote the further development of pulmonary medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1The figure is an overall flow chart of the method of the present invention.

[0025] Figure 2 The flowchart of the present invention is to generate a pulmonary artery vascular tree model.

[0026] Figure 3 The flowchart of establishing the hemodynamic model of the present invention is shown in FIG.

[0027] Figure 4 The flowchart of the watershed segmentation algorithm of the present invention is shown. DETAILED DESCRIPTION

[0028] like Figure 1-4 As shown, the present invention provides a pulmonary flow area analysis method based on hemodynamics. The method comprises the following steps: First, obtain lung CT image data and related lung anatomy data and clinical medical knowledge. Preferably, the resolution of the CT image data should be no less than 512x512 pixels, and the layer thickness should not exceed 1mm to ensure the accuracy of subsequent analysis. In one embodiment of the present invention, a multi-row CT scanner can be used to obtain spiral CT data of the entire lung, and the scanning parameters are set to 120kV, 250mAs. This setting can reduce the radiation dose while ensuring image quality.

[0029] Next, based on the acquired CT image data, a pulmonary artery vascular tree model is generated. Specifically, the method first reconstructs the two-dimensional CT image into a three-dimensional data set. Then, a vascular enhancement algorithm based on the Hessian matrix is ​​used to enhance the vascular structure. The mathematical expression of the algorithm is as follows: , in, Represent the second-order partial derivatives of the image in the x, y, and z directions, respectively. , , Represents mixed partial derivatives. By analyzing the eigenvalues ​​of the Hessian matrix, vascular structures and other tissues can be effectively distinguished.

[0030] Preferably, the present invention adopts a multi-scale analysis method to calculate the Hessian matrix at different Gaussian scales to adapt to blood vessels of different calibers. The scale range is usually set to 0.5 mm to 2 mm, which covers the diameter of most blood vessels in the lungs.

[0031] After the blood vessel enhancement, the method uses the region growing algorithm to extract the blood vessel structure. The selection of the seed point is crucial. The present invention preferably starts from the pulmonary artery trunk and automatically selects the point with the highest gray value as the initial seed point. The growth threshold can be adaptively adjusted according to the contrast of the image, and is generally set to the mean gray value plus 1.5 times the standard deviation, which can effectively distinguish blood vessels from surrounding tissues.

[0032] Subsequently, the method uses the minimum spanning tree algorithm to construct the vascular connection relationship and form a complete pulmonary artery vascular tree structure. In this process, the Kruskal algorithm is used to construct the minimum spanning tree, and its time complexity is O(ElogE), where E is the number of edges. This method can effectively handle complex vascular branching structures.

[0033] After obtaining the pulmonary artery vascular tree model, the method of the present invention further establishes a hemodynamic model. The model is constructed based on classical fluid mechanics theory and combined with the particularity of pulmonary blood flow. Specifically, the method considers the following key equations: 1. Continuity equation: , in, is the blood density, v is the velocity vector, For time.

[0034] 2. Momentum equation (Navier-Stokes equation): , For pressure, is the dynamic viscosity of blood, is the volume force.

[0035] 3. Energy equation: , in, is the specific heat capacity, is the temperature, is the thermal conductivity coefficient, is the viscous dissipation function.

[0036] When setting the parameters of these equations, the present invention preferably adopts the following empirical values: blood density kg / , dynamic viscosity Pa·s. These values ​​are based on a large number of physiological studies and can better reflect the characteristics of human blood.

[0037] It is worth noting that one of the innovations of the present invention is that the non-Newtonian properties of blood are taken into account. The Carreau-Yasuda model is used to describe the shear thinning properties of blood: , in, and are the viscosities at infinite shear rate and zero shear rate, respectively, is the characteristic time, is the shear rate, and is the model parameter. This model can more accurately describe the rheological properties of blood in different blood vessel diameters and flow rates.

[0038] In terms of boundary condition setting, this method uses periodic flow boundary conditions at the inlet to simulate the effect of heart beats on pulmonary artery blood flow. Zero pressure gradient conditions are used at the outlet. The vascular wall is regarded as a rigid no-slip boundary, which takes into account the relatively hard characteristics of the pulmonary artery wall.

[0039] Based on the above hemodynamic model, the method then calculates the blood flow velocity field and pressure gradient field. In this step, the finite volume method is used to discretize the control equations. When generating the grid, it is preferred to use a boundary layer grid near the vessel wall to better capture the velocity gradient near the wall. The number of grids is usually between 1 million and 5 million, depending on the complexity of the vascular tree.

[0040] During the calculation process, SIMPLE (Semi-Implicit Method for Pressure Linked Equations) algorithm is used to solve the velocity field and pressure field. The selection of the time step is crucial to the accuracy and stability of the calculation results. The present invention recommends setting the time step to 1 / 100 of the cardiac cycle, which can better capture the transient characteristics of blood flow.

[0041] After obtaining the blood flow velocity field and the pressure gradient field, the method executes the watershed segmentation algorithm to obtain the lung watershed division result. The core idea of ​​the watershed algorithm is to regard the image as a topographic map, and the grayscale value represents the altitude. The present invention innovatively applies this idea to the pressure gradient field, and regards the magnitude of the pressure gradient as the altitude.

[0042] The specific steps of the watershed algorithm are as follows: 1. Initialization: Mark the local minimum points in the pressure gradient field with different labels.

[0043] 2. Immersion process: gradually increase the water level to simulate the flooding process.

[0044] 3. When two areas with different labels are about to meet, a dam is built between them, which is the watershed line.

[0045] 4. Repeat steps 2 and 3 until the entire image is completely submerged.

[0046] In practical applications, in order to avoid over-segmentation, this method uses a marker-controlled watershed algorithm. Marker points are pre-set according to the hierarchical structure of vascular branches, and vascular bifurcation points are usually selected as marker points. This ensures that the segmentation results are consistent with the actual anatomical structure.

[0047] Finally, this method generates an analysis report containing the results of pulmonary watershed division. The report includes not only the spatial distribution map of the watershed, but also the quantitative analysis results of each watershed, such as watershed volume, average blood flow velocity, average pressure, etc. This information can help clinicians understand the patient's pulmonary blood flow status more comprehensively.

[0048] Through the above steps, the method provided by the present invention can accurately analyze and divide the pulmonary blood flow based on the principle of hemodynamics. Compared with the traditional method based only on anatomical structure, this method can provide more functional information, which helps to improve the diagnostic accuracy and treatment effect evaluation of lung diseases. The method of the present invention further includes specific steps for generating a pulmonary artery vascular tree model. These steps are designed to accurately extract the pulmonary artery structure from CT image data, laying the foundation for subsequent hemodynamic analysis.

[0049] First, the method reconstructs the lung CT image data into three-dimensional data. Preferably, a voxel-based reconstruction technique is used, such as a volume rendering algorithm (VolumeRendering) or a surface rendering algorithm (SurfaceRendering). In one embodiment of the present invention, the volume rendering algorithm uses the following transfer function: ,

[0050] in, is the output image intensity, and are the color and opacity of the i-th voxel, respectively, and n is the number of voxels that the light passes through. This method can well preserve the three-dimensional structural information of the blood vessels.

[0051] Next, the method extracts pulmonary vessels based on the reconstructed three-dimensional data. In this step, the present invention preferably uses a multi-scale vessel enhancement filter. The filter is based on the eigenvalue analysis of the Hessian matrix, and its response function can be expressed as: , in, is the eigenvalue of the Hessian matrix .

[0052] is a control parameter, usually taken as half of the maximumHessian norm.

[0053] After the blood vessel extraction, the method further identifies the bifurcation of the pulmonary artery trunk. Preferably, a method based on skeleton extraction and topological analysis is used. First, a thinning algorithm is used to extract the blood vessel skeleton, and then the bifurcation point is identified by analyzing the connectivity of the skeleton points. After the blood vessel extraction, the method further identifies the bifurcation of the pulmonary artery trunk. Preferably, a method based on skeleton extraction and topological analysis is used. First, a thinning algorithm is used to extract the blood vessel skeleton, and then the bifurcation point is identified by analyzing the connectivity of the skeleton points. In one embodiment of the present invention, the Zhang-Suen thinning algorithm is used, and the iterative process of the algorithm is as follows: 1. For each boundary point , check its 8 neighboring points .

[0054] 2. If the following conditions are met, then mark To be deleted: a) ; b) ; c) or ; d) or ,in, for The number of non-zero neighbors of For arrive The number of 0-1 transitions when traversing in order.

[0055] Finally, the method generates a complete pulmonary artery vascular tree structure based on the identified pulmonary artery trunk bifurcation. In this step, an adaptive region growing algorithm is preferably used. The core idea of ​​the algorithm is to start from the trunk bifurcation point and gradually expand outward until a preset stop condition is reached.

[0056] The method of the present invention pays special attention to processing small blood vessels and diseased areas when generating a pulmonary artery vascular tree model. For blood vessels with a diameter of less than 1 mm, super-resolution reconstruction technology is used to improve their visibility. For areas where there may be lesions, such as stenosis or dilation, the method combines morphological operations and local contrast enhancement technology to ensure that these areas can be accurately identified and modeled.

[0057] Through the above steps, the method of the present invention can generate a high-precision and high-completeness pulmonary artery vascular tree model from CT image data, providing a reliable geometric basis for subsequent hemodynamic analysis.

[0058] Next, the method of the present invention further elaborates on the process of establishing the hemodynamic model, which involves the setting of multiple key parameters and is crucial for accurately simulating pulmonary blood flow.

[0059] First, the method sets the physical parameters of blood. Under standard physiological conditions, the blood density is usually 1060 kg / m³. However, taking into account the individual differences of different patients, in a preferred embodiment of the present invention, the density value can be adjusted within the range of 1050-1070 kg / m³. The dynamic viscosity of blood is a more complex parameter because blood exhibits non-Newtonian fluid properties. At low shear rates, the blood viscosity is higher, while the viscosity decreases at high shear rates. In order to accurately describe this behavior, the method uses the Carreau-Yasuda model: , in, is the viscosity at infinite shear rate (about 0.0035 Pa·s), is the viscosity at zero shear rate (about 0.016 Pa·s), is the characteristic time (about 3.313s), is the shear rate, and These parameter values ​​are obtained based on a large amount of experimental data fitting and can well describe the rheological properties of human blood.

[0060] Static viscosity is another important parameter, which reflects the viscosity of blood in a static state. In this method, the static viscosity is set to 1.5 times the dynamic viscosity, which takes into account the effect of red blood cell aggregation in a static state.

[0061] The volume fraction of blood mainly refers to the hematocrit. Under normal physiological conditions, this value is about 45%. However, in certain lung diseases, such as pulmonary edema or anemia, this value may change significantly. Therefore, the method of the present invention allows this parameter to be adjusted according to the specific conditions of the patient, and the range is usually between 35% and 55%.

[0062] Finally, the thermal conductivity of blood is usually taken as 0.52 W / (m·K). This parameter is kept constant in most blood flow simulations unless the study is specifically interested in heat transfer processes.

[0063] Next, the method sets the wall conditions. In the standard case, the vessel wall is regarded as a rigid no-slip boundary. However, considering the elastic properties of actual blood vessels, a preferred embodiment of the present invention adopts an elastic wall model. This model is based on the following equation: , in, is the vascular wall density, is the displacement vector, is the stress tensor, is the volume force. The Young's modulus of the blood vessel wall is usually taken as 1MPa, and the Poisson's ratio is 0.45.

[0064] When setting the flow field boundary conditions, this method uses the velocity inlet condition at the inlet. The velocity profile can be described using the Womersley analytical solution: , in, is the zero-order Bessel function, is the Womersley number, is the vessel radius, is the angular frequency.

[0065] At the outlet boundary, this method adopts the pressure outlet condition. Considering the particularity of pulmonary circulation, the outlet pressure is set to 10-25 mmHg, which covers the normal physiological state to mild pulmonary hypertension.

[0066] Finally, this method also takes into account the special boundary conditions of the lungs. The pulmonary blood vessels are affected by respiratory motion, so a periodic external pressure change is introduced into the model. This pressure change can be described by the following function: , in, is the average thoracic pressure (about -4 mmHg), is the pressure wave amplitude (about 2 mmHg), is the breathing rate (about 0.25Hz).

[0067] By setting these detailed parameters and conditions, the method of the present invention can establish a highly accurate pulmonary hemodynamic model. This model not only takes into account the complex rheological properties of blood, but also includes the unique physiological characteristics of the lungs, providing a reliable basis for subsequent blood flow analysis.

[0068] After establishing the hemodynamic model, the method of the present invention further elaborates the calculation process of the blood flow velocity field and the pressure gradient field. This process is the core of the entire analysis and directly affects the accuracy of the subsequent flow domain division.

[0069] First, the method uses computational fluid dynamics (CFD) technology to solve the hemodynamic equations. Among the many CFD methods, the present invention preferably uses the finite volume method (FVM). The core idea of ​​FVM is to discretize the computational domain into a series of control volumes, and then apply conservation laws to each control volume. This method naturally guarantees the conservation of mass, momentum and energy, and is particularly suitable for processing vascular networks with complex geometries.

[0070] In the discretization process, this method uses a strategy that combines structured and unstructured grids. For the main blood vessels, hexahedral structured grids are used, which have high computational efficiency and small numerical diffusion. For branches and curved parts, tetrahedral unstructured grids are used to better adapt to complex geometric shapes. When generating the grid, special attention is paid to encrypting the grid near the blood vessel wall to accurately capture the boundary layer flow characteristics. The typical grid size is about 0.1 mm in the center of the blood vessel and can reach 0.01 mm near the wall.

[0071] In the solution process, this method uses SIMPLE (Semi-Implicit Method for Pressure-Linked Equations) algorithm. This is an iterative algorithm, and its basic steps are as follows: 1. Assume a pressure field; 2. Solve the momentum equation to obtain the velocity field; 3. Solve the pressure correction equation; 4. Correct pressure and speed; 5. Solve other scalar equations (such as turbulence equations); 6. Check convergence and return to step 2 if it has not converged.

[0072] In order to improve the computational efficiency, a preferred embodiment of the present invention adopts a multi-grid technique. This technique can significantly accelerate the convergence process by alternately solving on grids of different resolutions. The core idea of ​​the multi-grid method can be expressed by the following error equation: , , in, and denote the discrete operators on the fine grid and the coarse grid respectively, and is the corresponding error, is the residual, is the restriction operator.

[0073] In terms of time discretization, this method adopts a second-order implicit format. The selection of the time step is crucial and requires a balance between computational accuracy and efficiency. In an embodiment of the present invention, the time step is set to 1 / 100 of the cardiac cycle, which can well capture the transient characteristics of blood flow.

[0074] For turbulence simulation, this method uses the k-ωSST (Shear Stress Transport) model. This model combines the advantages of the k-ε model in the free flow region and the k-ω model in the near-wall region, and is particularly suitable for simulating complex flows with adverse pressure gradients and flow separation. The control equations of the k-ωSST model are as follows: , , in, is the turbulent kinetic energy, is the specific dissipation rate, and is the effective diffusion coefficient, and To generate items, and is the dissipation term, is the cross diffusion term, and A user-defined source item.

[0075] By solving the above equations, this method obtains detailed blood flow velocity field and pressure gradient field. These fields contain rich hemodynamic information and lay the foundation for subsequent flow domain division.

[0076] Next, the method of the present invention further describes the specific implementation process of the watershed basin segmentation algorithm. The core idea of ​​this algorithm is to regard the pressure gradient field as a topographic map and identify different basins by simulating the flooding process.

[0077] First, the method sets seed points in the pressure gradient field. The selection of seed points directly affects the quality of the segmentation results. In a preferred embodiment of the present invention, an automatic seed point selection strategy based on vascular bifurcation points is adopted. Specifically, for each major vascular bifurcation point, the point with the smallest pressure gradient around it is selected as the seed point. This ensures that each major vascular branch has a corresponding flow basin.

[0078] After the seed point is selected, the method executes the region growing algorithm. The core idea of ​​the algorithm is to start from the seed point and gradually expand the region until it encounters other regions or reaches the preset stop condition. In the present invention, the process of region growing can be described by the following mathematical model: , in, represents the pixel set of the ith region at the tth iteration, express The neighborhood pixel set of represents the pressure gradient at pixel p, is the threshold. The choice of is crucial. This method adopts an adaptive threshold strategy: , in, and are the average pressure gradient and standard deviation of the current area, To adjust the parameter, the value is usually between 1.5 and 2.5.

[0079] In the process of region growing, when two different regions meet, the method establishes a watershed line between them. In order to avoid the over-segmentation problem, the present invention introduces a region merging strategy. Specifically, when the difference in the average pressure gradient of two adjacent regions is less than a preset threshold, the two regions will be merged. The merging condition can be expressed as: , in, and is the average pressure gradient between two adjacent regions, and is the corresponding standard deviation, It is a merging parameter, usually with a value between 0.5 and 1.

[0080] In order to further improve the robustness of the segmentation, the method of the present invention also introduces a morphological post-processing step. First, an opening operation is applied to remove small isolated regions: , Among them, A is the segmentation result, B is the structural element, and Denote the erosion and dilation operations respectively. Then, a closing operation is applied to fill the small holes: , Through these steps, the method of the present invention can obtain a smooth and coherent lung flow domain division result.

[0081] Finally, this method generates an analysis report containing the results of lung watershed delineation. This report includes not only a visual watershed distribution map, but also quantitative analysis results for each watershed. Specifically, for each identified watershed, the report contains the following information: 1. Basin volume: calculated by integration, in cubic millimeters.

[0082] 2. Average blood flow velocity: The average velocity of all points in the flow basin, in meters per second.

[0083] 3. Average pressure: The average pressure of all points in the basin, expressed in millimeters of mercury.

[0084] 4. Wall shear stress (WSS): Calculate the average shear stress of the blood vessel wall in Pascals.

[0085] 5. Oscillatory Shear Index (OSI): An indicator that reflects the change in the direction of wall shear stress, dimensionless.

[0086] These quantitative indices provide clinicians with tools to comprehensively assess pulmonary blood flow conditions. For example, abnormally high wall shear stress may indicate the risk of aneurysm, while a high oscillatory shear index may be associated with atherosclerosis.

[0087] In addition, the method of the present invention also adds abnormal detection results to the report. By comparing the difference between the indicators of each watershed and the normal reference value, the area where there may be problems is automatically marked. For example, if the average blood flow velocity in a watershed is less than 50% of the normal value, the area will be marked as a potential blood flow obstruction area.

[0088] To make it easier for doctors to understand and use, the report adopts a hierarchical structure. The first page provides an overall overview, including a 3D watershed distribution map and a summary of the main abnormal areas. Subsequent pages list the specific parameters and analysis results of each watershed in detail. The report also includes interactive features that allow doctors to get more detailed information by clicking on the area of ​​interest.

[0089] Through this comprehensive, detailed and intuitive report, the method of the present invention provides a powerful tool for pulmonary hemodynamic analysis, which is expected to significantly improve the diagnosis and treatment of related diseases. The method of the present invention further includes an abnormal blood flow identification step, which is a key innovation that can directly convert the hemodynamic analysis results into clinically useful information.

[0090] First, the method calculates the outlier discrimination parameters. These parameters include flow, density, velocity, vessel radius, and vessel length. The calculation of each parameter is based on the hemodynamic simulation results obtained in the previous steps. Specifically: 1. The flow rate Q is calculated using the following formula: , in, is the velocity vector, and A is the cross-sectional area of ​​the blood vessel. In actual calculations, the integral can be discretized as: , Where N is the number of mesh elements in the cross section, and are the velocity and area of ​​the ith unit respectively.

[0091] 2. The density ρ is directly extracted from the blood flow simulation results and is usually about 1060 kg / m³ under normal physiological conditions.

[0092] 3. The velocity v is taken as the average value in the basin: , where V is the volume of the basin.

[0093] 4. The vessel radius R can be estimated by the volume and length of the basin: , Where L is the length of the vessel.

[0094] 5. The vessel length L is calculated using a skeletonization algorithm, specifically the Zhang-Suen thinning algorithm.

[0095] Next, the method sets abnormality judgment thresholds. The selection of these thresholds is based on a large amount of clinical data and expert experience. In a preferred embodiment of the present invention, the following thresholds are used: Traffic threshold : 50% of normal value; density threshold :1000kg / m³; speed threshold : 30% of normal value; radius threshold : 70% (stenosis) and 150% (dilation) of normal value; length threshold : 120% of normal value; These thresholds can be adjusted according to specific clinical needs. For example, the thresholds can be set to be more sensitive for early screening and more specific for confirmatory diagnosis.

[0096] Based on the above parameters and thresholds, this method identifies abnormal blood flow patterns. The specific judgment logic is as follows: 1. Embolism point: when and This is because embolism can cause a significant reduction in blood flow, and due to the formation of thrombus, the local blood density will also decrease.

[0097] 2. Narrow point: When and Stenosis will cause the local blood vessel radius to decrease, and due to the increase in flow velocity, the local density may decrease.

[0098] 3. Expansion point: When and Vascular dilation will cause local blood flow to slow down, density may decrease, and the vascular radius will increase significantly.

[0099] 4. Diversion point: When and This is usually found at the bifurcation of the vessel and is characterized by a short vessel length and a small radius.

[0100] 5. Infusion deficiency: when and In this case, the vessel is shorter but has a larger radius, which may be due to compensatory dilation caused by upstream blood flow obstruction.

[0101] 6. Overfilling: When This may be due to abnormal angiogenesis or arteriovenous malformations.

[0102] In order to improve the accuracy of judgment, this method also introduces a fuzzy logic system. For example, for the judgment of narrow points, the following fuzzy rules can be used: IF (density is low) AND (radius is small) THEN (stenosis is high) Among them, low, small and high are all fuzzy sets, which can be described by membership functions. This method can better handle the continuous changes of parameters and avoid the judgment errors that may be caused by hard thresholds.

[0103] Finally, the method adds the identified abnormal blood flow patterns to the analysis report. In the report, the abnormal points are marked with different colors on the 3D model, with detailed parameter information and possible pathological explanations. For example, for an identified stenosis point, the report will provide information such as the exact location of the point, the vessel radius, the local flow velocity, the pressure gradient, and the possible causes of the stenosis (such as atherosclerosis, thrombosis, etc.).

[0104] In this way, the method of the present invention not only provides quantitative analysis of hemodynamics, but also directly gives abnormal identification results with clinical significance, greatly improving the practicality and interpretability of the analysis results.

[0105] Next, the method of the present invention introduces an innovative dynamic adjustment step, which aims to improve the adaptability and robustness of the method, especially when dealing with different patients or different pathological conditions.

[0106] First, this method counts the velocity distribution factor, pressure drop factor and resistance gradient factor of normal blood flow in the current flow domain. The definitions of these factors are as follows: 1. Velocity distribution factor : ,in, is the standard deviation of velocity in the basin, is the average speed. This factor reflects the uniformity of the speed distribution.

[0107] 2. Pressure drop factor : ,in, is the pressure drop in the flow domain, and L is the characteristic length of the flow domain. This factor reflects the pressure loss per unit dynamic pressure.

[0108] 3. Resistance gradient factor : , where R is the vascular resistance, which can be estimated by the Hagen-Poiseuille equation: , μ is the blood viscosity, and r is the blood vessel radius.

[0109] Next, the method calculates the probability distribution of these factors. In a preferred embodiment of the present invention, a kernel density estimation (KDE) method is used to estimate the probability density function. The mathematical expression of KDE is as follows: , Where K is the kernel function and h is the bandwidth parameter. This method uses the Gaussian kernel function: , The choice of bandwidth parameter h is crucial. This method adopts Silverman's rule of thumb: , in, is the sample standard deviation, and n is the sample size.

[0110] Based on the estimated probability density function, this method extracts the maximum value point as the dynamic adjustment threshold. Specifically: 1. For velocity distribution factor , take the maximum value of its probability density function , adjust the speed threshold in the watershed algorithm to: , Here α is a tuning parameter, which usually takes a value between 0.5 and 1.5.

[0111] 2. For the pressure drop factor , take the maximum value of its probability density function , adjust the pressure gradient threshold to: , Where β is a tuning parameter, which usually takes a value between 0.8 and 1.2.

[0112] 3. For the resistance gradient factor , take the maximum value of its probability density function , adjust the blood vessel radius change threshold to: , Here γ is a tuning parameter, which usually takes a value between 0.3 and 0.7.

[0113] Through this dynamic adjustment mechanism, the method of the present invention can adaptively adjust the threshold of the watershed segmentation algorithm to better adapt to the individual differences and different pathological conditions of different patients. For example, for patients with pulmonary hypertension, due to the high overall pressure gradient, a fixed threshold may lead to over-segmentation. The dynamically adjusted threshold can more accurately reflect the actual situation of the patient.

[0114] In addition, this method introduces a feedback mechanism. After each analysis is completed, the results are compared with the clinical diagnosis. If there is a significant difference, the system automatically adjusts the three parameters α, β, and γ to optimize the next analysis result. This mechanism makes this method have the ability to learn and can continuously improve its accuracy as it is used.

[0115] Through this dynamic adjustment and feedback mechanism, the method of the present invention significantly improves its applicability and reliability in different clinical scenarios, providing a powerful tool for individualized pulmonary hemodynamic analysis. The method of the present invention further refines the results of pulmonary flow basin division and divides it into four levels. This hierarchical structure can not only more accurately reflect the complexity of pulmonary blood flow, but also provide more valuable information for clinical diagnosis.

[0116] The first-level flow domain includes the area including the right ventricular exit pulmonary artery. This is the starting point of the entire pulmonary circulation and is crucial for understanding the overall pulmonary hemodynamics. In a preferred embodiment of the present invention, the identification of the first-level flow domain adopts the following steps: 1. Based on anatomical knowledge, locate the main pulmonary artery at the outlet of the right ventricle.

[0117] 2. Using the vessel tracking algorithm, start from the main pulmonary artery and track downstream until the first major bifurcation is encountered.

[0118] 3. Define the traced area as the first-level watershed.

[0119] This area usually has the largest blood vessel diameter and the highest blood flow velocity. The present method pays special attention to the changes in pressure and flow velocity in this area because they may reflect early signs of right heart function and pulmonary hypertension.

[0120] The second-order drainage basins are branches of the first-order drainage basins. These usually correspond to the major branches of the left and right pulmonary arteries. The procedure for identifying the second-order drainage basins is as follows: 1. Starting from the end point of the first-order basin (i.e., the main bifurcation point), continue vessel tracing.

[0121] 2. Each branch is traced until the next significant bifurcation point is encountered or the vessel diameter decreases below a threshold.

[0122] 3. Define each traced branch as an independent second-level watershed.

[0123] In one embodiment of the present invention, a significant bifurcation point is defined as a bifurcation with a ratio of the diameter of the daughter vessel to the diameter of the parent vessel greater than 0.7. The vessel diameter threshold is usually set to 3 mm, which is determined based on the normal adult pulmonary artery anatomy.

[0124] The third-level basins are offshoots of the second-level basins. These usually correspond to vessels at the lobar and segmental levels. The identification process is similar to the second-level basins, with the following differences: 1. The criteria for determining bifurcation points are more relaxed, and a significant bifurcation is considered when the ratio of the diameter of the daughter vessel to the diameter of the parent vessel is greater than 0.5.

[0125] 2. The blood vessel diameter threshold is reduced to 1.5 mm.

[0126] The fourth-order basins are branches of the third-order basins and correspond to smaller intrapulmonary vessels. Since the size of these vessels is close to the resolution limit of CT scanning, a special processing strategy is used in this method: 1. Use super-resolution reconstruction technology to improve image resolution. The present invention preferably uses a super-resolution algorithm based on deep learning, such as SRCNN (Super-Resolution Convolutional Neural Network).

[0127] 2. Apply a vessel enhancement filter, such as the Frangi filter, to highlight small vessel structures.

[0128] 3. Use more flexible criteria for determining bifurcation, as long as it can be clearly identified as a branching structure.

[0129] 4. The blood vessel diameter threshold is further reduced to 0.5mm.

[0130] In order to ensure the accuracy and consistency of watershed division, this method introduces an innovative verification mechanism. This mechanism is based on the following principles: 1. Conservation of mass: The total flow of an upstream basin should be equal to the sum of the flows of all its downstream basins.

[0131] 2. Pressure continuity: The pressure between adjacent river basins should change continuously and no sudden changes should occur.

[0132] 3. Morphological consistency: Watersheds of the same level should have similar morphological characteristics (such as length, number of branches, etc.).

[0133] If violations of these principles are found, the method automatically performs local re-division. For example, if a third-level watershed is found to have unusually high flow, the system will re-examine whether the area should be divided into more sub-basins.

[0134] In addition, the method of the present invention also takes into account special circumstances under pathological conditions. For example, for patients with pulmonary embolism, some blood vessels may be completely occluded. In this case, the traditional blood flow-based segmentation method may fail. For this reason, the present method introduces a morphology-based auxiliary segmentation strategy. Even in the absence of obvious blood flow, reasonable flow basin segmentation can be performed based on the anatomical structure of the blood vessels.

[0135] Through this multi-level, adaptive flow domain partitioning method, the present invention can provide a comprehensive and detailed description of pulmonary blood flow, providing a powerful tool for clinical diagnosis and research.

[0136] Finally, the method of the present invention introduces an innovative result verification step to further improve the reliability and accuracy of the analysis results. This step uses the dynamic data of the fluorescent tracer to locate and quantitatively measure the blood flow distribution, and uses the actual measurement results to verify and optimize the hemodynamic model.

[0137] First, the method uses a fluorescent tracer for dynamic imaging. In a preferred embodiment of the present invention, indocyanine green (ICG) is used as a tracer because it has good biocompatibility and rapid plasma clearance. The injection and imaging process of ICG is as follows: 1. Rapidly inject ICG solution (dose 0.1 mg / kg body weight) through a peripheral vein.

[0138] 2. Use a near-infrared fluorescence imaging system to continuously collect images with a sampling rate of 30 frames / s and a duration of 60 seconds.

[0139] Next, the method analyzes the obtained dynamic image sequence to extract blood flow distribution information. The specific steps include: 1. Image preprocessing: including denoising, background correction and motion compensation.

[0140] 2. Region of interest (ROI) definition: Based on the watershed division results obtained in the previous step, define the corresponding ROI on the fluorescence image.

[0141] 3. Time-intensity curve (TIC) extraction: For each ROI, extract the curve of fluorescence intensity changing with time.

[0142] 4. Calculation of blood flow parameters: Calculate various blood flow parameters based on TIC, including: Time to Peak (TTP); Maximum Slope (MS); Mean Transit Time (MTT); Blood Flow Index (BFI); The calculation formula of BFI is:

[0143] here, is the maximum fluorescence intensity, is the baseline intensity and TTP is the time to peak.

[0144] Based on these actual measured blood flow parameters, the present method was validated and optimized in the following steps: 1. Blood flow distribution verification: Compare the blood flow distribution obtained from CFD simulation with the fluorescence imaging measurement results. If the difference in a certain area exceeds 20%, the area will be marked for further investigation.

[0145] 2. Blood viscosity correction: Using the dynamic data of the fluorescent tracer, the local blood viscosity can be estimated. The estimation formula is: , in, is the pressure difference, D is the vessel diameter, and L is the vessel length. This estimated viscosity value is used to update the blood viscosity parameter in the CFD model.

[0146] 3. Boundary condition optimization: Optimize the boundary condition settings of the CFD model by comparing the simulation results with the actual measured inlet / outlet flow rates.

[0147] 4. Model parameter adjustment: Use machine learning algorithms (such as Bayesian optimization) to automatically adjust key parameters of the CFD model so that the simulation results better match the actual measurement data.

[0148] Through this measured data-driven verification and optimization process, the method of the present invention significantly improves the accuracy and reliability of hemodynamic analysis. In particular, this method can capture individual differences and special circumstances under pathological conditions, providing a solid foundation for personalized diagnosis and treatment.

[0149] It is worth noting that this method fully considers patient safety when using fluorescent tracers. The dose of ICG used is far below the clinical safety limit, and the entire imaging process does not involve ionizing radiation, so the additional health risks are extremely low. This makes this verification step safe to be applied in routine clinical practice.

[0150] In general, by combining computational fluid dynamics simulation and fluorescent tracer measured data, the method of the present invention creatively constructs a virtual-real pulmonary blood flow analysis system. This system not only provides highly accurate hemodynamic information, but also has the ability of self-verification and continuous optimization, representing an important advancement in the field of pulmonary blood flow analysis.

[0151] The above description is only a preferred specific implementation manner of the present invention; however, the protection scope of the present invention is not limited thereto; any person familiar with the art who, within the scope disclosed by the present invention, makes equivalent replacements or changes based on the scheme and improved concepts of the present invention shall be covered within the protection scope of the present invention.

Claims

1. A pulmonary flow area analysis method based on hemodynamics, characterized in that: include: The acquisition steps include: Acquire lung CT image data; Obtain lung anatomy data and clinical medical knowledge; Processing steps include: Based on the lung CT image data, generating a pulmonary artery vascular tree model; Establishing a hemodynamic model according to the pulmonary artery vascular tree model; Based on the hemodynamic model, calculating the blood flow velocity field and the pressure gradient field; According to the pressure gradient field, a watershed watershed segmentation algorithm is executed to obtain a lung watershed segmentation result; Output steps include: An analysis report including the lung watershed division results is generated.

2. The method according to claim 1, characterized in that: The generating of the pulmonary artery vascular tree model specifically comprises: reconstructing the lung CT image data into three-dimensional data; extracting pulmonary blood vessels based on the three-dimensional data; According to the pulmonary anatomical data, identifying the bifurcation of the main pulmonary artery; Based on the bifurcation of the pulmonary artery trunk, a pulmonary artery vascular tree structure is generated.

3. The method according to claim 1, characterized in that The establishment of the hemodynamic model specifically includes: Set blood physical parameters, including blood density, dynamic viscosity, static viscosity, volume fraction and thermal conductivity; Set the wall conditions, flow field boundary conditions, and lung boundary conditions; Based on Darcy's law, LM equation, continuity equation and Bernoulli equation, the hemodynamic equations were constructed.

4. The method according to claim 3, characterized in that The calculation of the blood flow velocity field and the pressure gradient field specifically includes: Dividing the computational domain of the hemodynamic model into two parts, upstream and downstream; applying an inlet boundary condition to the upstream boundary; applying an outlet boundary condition to the downstream boundary; The computational fluid dynamics (CFD) algorithm is used to solve the hemodynamic equations to obtain the blood flow velocity field and the pressure gradient field.

5. The method according to claim 1, characterized in that The execution of the watershed segmentation algorithm specifically includes: Setting a seed point in the pressure gradient field; Based on the seed point, executing a region growing algorithm; Determine the watershed boundary based on the pressure gradient threshold; Generate lung watershed partition results.

6. The method according to claim 1, characterized in that It also includes abnormal blood flow identification steps: Calculate outlier discrimination parameters, including flow, density, velocity, vessel radius, and vessel length; Set the abnormal judgment threshold; Based on the abnormal point discrimination parameter and the abnormal judgment threshold, identifying an abnormal blood flow pattern; The abnormal blood flow pattern is added to the analysis report.

7. The method according to claim 6, characterized in that The abnormal blood flow patterns include: When the flow rate is less than the first preset threshold and the density is less than the second preset threshold, it is determined to be an embolism point; When the density is less than the second preset threshold and the blood vessel radius is less than the third preset threshold, it is determined to be a stenosis point; When the density is less than the second preset threshold and the blood vessel radius is greater than the fourth preset threshold, it is determined to be a dilation point.

8. The method according to claim 1, characterized in that It also includes dynamic adjustment steps: Statistical analysis of velocity distribution factor, pressure drop factor and resistance gradient factor of normal blood flow in the current flow area; calculating a probability distribution of the factor; Based on the maximum value point of the probability distribution, the threshold of the watershed segmentation algorithm is dynamically adjusted.

9. The method according to claim 1, characterized in that: The lung watershed division results include: First-order watershed: the area including the right ventricular pulmonary artery; Second-level basin: a branch of the first-level basin; Third-level basin: a branch of the second-level basin; Fourth-level basin: a branch of the third-level basin.

10. The method according to claim 1, characterized in that It also includes the result verification step: Use dynamic data of fluorescent tracers to locate and quantitatively measure blood flow distribution; Calculating actual blood viscosity based on the dynamic data of the fluorescent tracer; comparing the actual blood viscosity with the blood viscosity parameter in the hemodynamic model; According to the comparison results, the parameters of the hemodynamic model are adjusted.

Images