A method for pulmonary vascular perfusion analysis based on watershed dynamics

Abstract

The present invention relates to the technical field of medical image processing. More specifically, it relates to a pulmonary vascular perfusion analysis method based on watershed dynamics, including: acquiring a pulmonary artery vascular tree and a pulmonary arterial CT angiogram; extracting the pulmonary artery vascular tree based on the pulmonary artery vascular tree and the pulmonary arterial CT angiogram; calculating the pressure gradient field inside and on the wall of the pulmonary artery vascular tree based on the pulmonary artery vascular tree; extracting the pulmonary vein watershed based on the pressure gradient field; realizing perfusion analysis by combining dynamic hemodynamic simulation based on the pulmonary vein watershed; evaluating the blood supply of lung tissue based on the perfusion analysis results. By adopting the multi-level connected domain detection method, this method can more accurately extract the complex structure of the pulmonary artery vascular tree. The watershed division method based on the pressure gradient field can more truly reflect the actual distribution of blood flow. The introduction of dynamic blood flow simulation enables this method to comprehensively capture the perfusion changes during the breathing process and provide more comprehensive diagnostic information.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and more specifically, to a method for pulmonary vascular perfusion analysis based on watershed dynamics. Background Art

[0002] With the rapid development of medical imaging technology, pulmonary vascular perfusion analysis plays an increasingly important role in clinical diagnosis and treatment. Traditional pulmonary vascular perfusion analysis methods mainly rely on static CT or magnetic resonance imaging, and evaluate the pulmonary blood flow by observing the distribution of contrast agents. However, these methods often cannot accurately reflect the dynamic changes of pulmonary blood flow during respiration, resulting in certain limitations in the diagnostic results.

[0003] In recent years, with the progress of computer technology and medical image processing technology, pulmonary vascular perfusion analysis methods based on image segmentation and hemodynamic simulation have gradually emerged. These methods perform complex post-processing on CT or magnetic resonance images and combine hemodynamic models to attempt to more accurately simulate and analyze pulmonary blood perfusion. However, there are still some obvious deficiencies in these existing methods.

[0004] First of all, most existing methods still rely on traditional image segmentation algorithms, such as threshold segmentation or region growing method, in the vascular segmentation stage. These algorithms often have problems of over-segmentation or under-segmentation when dealing with complex pulmonary vascular structures, and it is difficult to accurately extract the complete pulmonary artery vascular tree. Secondly, in terms of hemodynamic simulation, existing methods usually adopt simplified hydrodynamic models, ignoring the elastic characteristics of blood vessel walls and the non-Newtonian fluid characteristics of blood, resulting in a large deviation between the simulation results and the actual physiological conditions.

[0005] In addition, when analyzing the pulmonary vein watershed, existing methods often use simple geometric division or empirical division based on anatomical structures, which is difficult to accurately reflect the actual blood flow distribution. This rough division method may lead to incorrect evaluation of the perfusion conditions of different lung lobes and affect the accuracy of clinical diagnosis.

[0006] Finally, most existing pulmonary vascular perfusion analysis methods are limited to static analysis and are difficult to capture the dynamic changes of pulmonary blood flow during respiration. This static analysis method cannot comprehensively reflect the pulmonary perfusion conditions of patients in different respiratory states and may lead to the neglect of certain pathological conditions.

[0007] In view of the problems existing in the above-mentioned prior art, there is an urgent need for a method that can more accurately and comprehensively analyze pulmonary vascular perfusion. The method for pulmonary vascular perfusion analysis based on watershed dynamics proposed by the present invention is designed to solve these technical problems. Summary of the Invention

[0008] The present invention aims to solve the above technical problems and proposes a method for pulmonary vascular perfusion analysis based on watershed dynamics. The method of the present invention innovatively introduces the concept of watershed dynamics, combines advanced image processing technology and accurate hemodynamic simulation, and realizes the all-round and dynamic analysis of pulmonary vascular perfusion. This method not only overcomes the deficiencies of the prior art in aspects such as vascular segmentation, watershed division, and dynamic simulation, but also provides richer and more accurate perfusion information, providing strong support for clinical diagnosis and treatment decision-making.

[0009] To solve the above technical problems, the present invention adopts the following technical solutions:

[0010] A method for pulmonary vascular perfusion analysis based on watershed dynamics, comprising:

[0011] An acquisition step, comprising:

[0012] Acquire the pulmonary artery vascular tree and the pulmonary artery CT angiogram;

[0013] A processing step, comprising:

[0014] Based on the pulmonary artery vascular tree and the pulmonary artery CT angiogram, extract the pulmonary artery vascular tree;

[0015] Based on the pulmonary artery vascular tree, calculate the pressure gradient field inside and on the wall of the pulmonary artery vascular tree;

[0016] Based on the pressure gradient field, extract the pulmonary vein watershed;

[0017] Based on the pulmonary vein watershed, combine dynamic hemodynamic simulation to achieve perfusion analysis;

[0018] An output step, comprising:

[0019] Based on the perfusion analysis result, evaluate the blood supply situation of the lung tissue.

[0020] Preferably, the extraction of the pulmonary artery vascular tree specifically includes:

[0021] Based on the pulmonary artery vascular tree and the pulmonary artery CT angiogram, segment the pulmonary artery vascular tree region;

[0022] Extract a candidate region around the pulmonary artery vascular tree region;

[0023] Extract seed pixel points within the candidate region;

[0024] Based on the seed pixel points, perform region growing to obtain the connected domain of the pulmonary artery vascular tree;

[0025] Perform clustering labeling on the connected domain of the pulmonary artery vascular tree;

[0026] Extract the largest connected domain from the clustered marked area as the pulmonary artery vascular tree.

[0027] Preferably, the calculation of the pressure gradient field inside and on the wall of the pulmonary artery vascular tree specifically includes:

[0028] Perform mesh division on the pulmonary artery vascular tree using the finite element analysis method;

[0029] Set boundary conditions and apply fluid boundary conditions and wall boundary conditions;

[0030] Calculate the pressure gradient vector, momentum flux, fluid pressure distribution, and total pressure of each element;

[0031] Calculate the momentum derivative of the blood flow at the vessel wall.

[0032] Preferably, the extraction of the pulmonary vein basin specifically includes:

[0033] Extract the initial streamline set of the blood flow at the current time point;

[0034] Calculate the first pressure gradient between adjacent streamline clusters in the initial streamline set;

[0035] Calculate the second pressure gradient formed by the connection between the current point and other points in the initial streamline set;

[0036] Compare the absolute value of the difference between the first pressure gradient and the second pressure gradient;

[0037] Based on the comparison result, determine whether the point is in the blood flow transition domain or the blood flow stable domain;

[0038] Based on the judgment result, determine the shunt boundary of the pulmonary vein basin.

[0039] Preferably, the implementation of perfusion analysis by combining dynamic hemodynamic simulation specifically includes:

[0040] Take the pulmonary vein vessel entrance as the outlet point and obtain the streamline information of the previous frame;

[0041] Search for the lowest pressure point on the wall of the pulmonary artery vascular tree as the downstream point;

[0042] Take the downstream point as the starting point of the next frame and determine a new downstream point on the wall of the next frame;

[0043] Repeat the above process until the downstream point of the next frame cannot be found on the new frame;

[0044] Calculate the position where the pressure gradient of the vessel wall is zero as the end point;

[0045] Extract the entire streamline formed from the initial streamline to the end point on the pulmonary artery vessel wall as the perfusion path of the pulmonary artery vascular tree.

[0046] Preferably, the evaluation of the blood supply situation of the lung tissue specifically includes:

[0047] Randomly set multiple seed points outside the boundary of the pulmonary vein basin;

[0048] Use the determined perfusion path as the seed point, and label the seed points inside the vascular tree by the marker search method;

[0049] Extract all perfusion paths containing the labeled points to obtain a set of perfusion paths;

[0050] Calculate the average blood flow velocity of each perfusion path;

[0051] Select the blood flow volume at different lung positions and calculate the range of lung ischemia;

[0052] Calculate the range of lung ischemia at different time points to obtain the dynamic range of lung ischemia during the breathing process of this patient.

[0053] Preferably, it further includes denoising the pulmonary artery CT angiogram, and the denoising process includes:

[0054] Based on the pulmonary artery CT angiogram, extract the largest connected region in the CT angiogram;

[0055] Segment the largest connected region to obtain the vascular tree region;

[0056] Randomly select the background outside the pulmonary artery vascular region for assignment to form a binary image inside and outside the largest connected region;

[0057] Smooth the boundary of the binary image to obtain the denoised pulmonary artery vascular tree region.

[0058] Preferably, it further includes downsampling the pulmonary artery CT angiogram, and the downsampling process includes:

[0059] Perform downsampling on the denoised pulmonary artery vascular tree region;

[0060] Perform edge detection on the downsampled pulmonary artery vascular tree region;

[0061] Based on the watershed algorithm, perform watershed segmentation on the grayscale image after edge detection to obtain each connected domain image;

[0062] Perform region growing on each connected domain image within the vascular region to generate the pulmonary artery CT angiogram.

[0063] Preferably, the down - frequency processing adopts the method of minimum mean filtering.

[0064] Preferably, in the step of extracting the pulmonary vein basin area, the following steps are further included:

[0065] Set a threshold value;

[0066] Filter the pressure gradient calculation results according to the threshold value, and only retain the points where the pressure is greater than the pressure gradient threshold value;

[0067] Take the retained points as the outer boundary of the blood basin area;

[0068] Taking the outer boundary as the center, use the bisection method to gradually determine that the pressure gradient at the boundary is zero, and determine the boundary line;

[0069] Take the boundary line as the boundary line of the pulmonary vein blood basin area;

[0070] Randomly select seed points outside the boundary line;

[0071] Connect each seed point on the boundary line to the point on the nearest pulmonary artery blood vessel wall;

[0072] Search outside the boundary until no next - frame downstream point can be found;

[0073] Calculate the position where the pressure gradient of the blood vessel wall is zero as the end point;

[0074] Extract the whole streamline formed from the initial streamline to the end point on the blood vessel wall as the perfusion path of the pulmonary artery blood vessel tree.

[0075] The method of the present invention has the following remarkable technical effects:

[0076] The method of the present invention has made remarkable progress in improving the accuracy of pulmonary blood vessel segmentation, optimizing the basin area division, and realizing dynamic perfusion analysis. By adopting the multi - level connected domain detection method, this method can more accurately extract the complex pulmonary artery blood vessel tree structure. The basin area division method based on the pressure gradient field can more realistically reflect the actual distribution of blood flow. The introduction of dynamic blood flow simulation enables this method to comprehensively capture the perfusion changes during the breathing process and provide more comprehensive diagnostic information.

[0077] In addition, the method of the present invention has also made important innovations in perfusion evaluation. By introducing the evaluation method based on the perfusion path, the present invention can quantitatively analyze the blood supply situation of different lung regions, providing a more objective and accurate basis for clinical diagnosis. This evaluation method can not only identify potential lesion areas, but also evaluate the treatment effect, which is of great significance for guiding clinical treatment.

[0078] Generally speaking, the method for analyzing pulmonary vascular perfusion based on watershed dynamics proposed by the present invention realizes a more accurate, comprehensive and dynamic analysis of pulmonary vascular perfusion through the organic combination of multiple technological innovations. This method not only overcomes many limitations of the existing technologies, but also provides more reliable and rich information support for clinical diagnosis and treatment. It is expected to play an important role in improving the diagnostic accuracy of pulmonary diseases and optimizing the formulation of treatment plans, and has broad clinical application prospects. Brief Description of the Drawings

[0079] Figure 1 It is the overall flowchart of the method of the present invention.

[0080] Figure 2 It is the flowchart for extracting the pulmonary artery vascular tree of the present invention. Detailed Embodiment

[0081] As Figure 1-2 shown, the present invention proposes a method for analyzing pulmonary vascular perfusion based on watershed dynamics. This method makes full use of fluid mechanics and medical image processing technologies to achieve an accurate analysis of pulmonary vascular perfusion. The technical solutions of the present invention will be described in detail below.

[0082] First of all, the method of the present invention includes an acquisition step, a processing step and an output step. In the acquisition step, the pulmonary artery vascular tree and the CT angiogram of the pulmonary artery are acquired. These original data lay the foundation for subsequent analysis.

[0083] Next, in the processing step, the method of the present invention extracts the pulmonary artery vascular tree based on the acquired pulmonary artery vascular tree and the CT angiogram of the pulmonary artery. This step is crucial for accurately analyzing the pulmonary vascular structure. Preferably, the present invention uses the multi-level connected component detection method to extract the pulmonary artery vascular tree, and this method can better process the complex pulmonary vascular branch structure compared with the traditional method.

[0084] After extracting the pulmonary artery vascular tree, the method of the present invention calculates the pressure gradient field inside and on the wall of the pulmonary artery vascular tree based on the extracted pulmonary artery vascular tree. This step is the key to achieving accurate perfusion analysis. The present invention preferably uses the finite element analysis method to calculate the pressure gradient field, and this method can fully consider the elastic characteristics of the blood vessel wall and the non-Newtonian fluid characteristics of the blood, so as to obtain a more accurate pressure distribution.

[0085] After obtaining the pressure gradient field, the method of the present invention extracts the pulmonary venous watershed based on the calculated pressure gradient field. This step is of great significance for distinguishing the perfusion regions of different lung lobes. The present invention innovatively proposes a method based on streamline clusters and pressure gradient analysis to determine the watershed boundary, and this method can more accurately reflect the actual blood flow situation.

[0086] Subsequently, the method of the present invention realizes perfusion analysis based on the extracted pulmonary vein watershed, in combination with dynamic hemodynamic simulation. This step takes into account the hemodynamic changes during respiration and can obtain a more realistic perfusion situation. The present invention adopts a streamline-tracing-based method to analyze the perfusion path, which can visually display the blood flow process in the lungs.

[0087] Finally, in the output step, the method of the present invention evaluates the blood supply of the lung tissue based on the obtained perfusion analysis results. This step provides an important basis for clinical diagnosis and treatment. The present invention proposes an evaluation method based on the perfusion path, which can quantitatively analyze the blood supply of different lung regions.

[0088] Next, the present invention will describe in detail the specific steps of extracting the pulmonary artery vascular tree. First, based on the obtained pulmonary artery vascular tree and the pulmonary artery CT angiogram, the pulmonary artery vascular tree region is segmented. This step is usually implemented by using image segmentation algorithms such as threshold segmentation or region growing. Preferably, the present invention adopts an adaptive threshold segmentation algorithm, and the threshold can be automatically adjusted according to the gray distribution of the image. The typical threshold range is 100 - 200 HU (Hounsfield Unit).

[0089] After segmenting the pulmonary artery vascular tree region, the method of the present invention extracts a candidate region around the segmented pulmonary artery vascular tree region. Preferably, the size of the extracted candidate region is 3×3 pixels. The selection of this size is based on experience and can reduce the computational amount while ensuring the extraction accuracy.

[0090] Subsequently, the method of the present invention extracts seed pixel points within the extracted candidate region. Usually, pixel points with a gray value of zero are selected as seed pixel points. This is because in CT images, the HU value of air is close to -1000, and after normalization, its gray value is close to zero.

[0091] Based on the extracted seed pixel points, the method of the present invention performs region growing to obtain the connected domain of the pulmonary artery vascular tree. During the region growing process, the growth threshold is usually set to ±10% of the gray value of the seed points. The selection of this range can effectively control the growth process and avoid overgrowth or undergrowth.

[0092] After obtaining the connected domain, the method of the present invention performs clustering labeling on the obtained connected domain of the pulmonary artery vascular tree. Preferably, the K-means clustering algorithm is adopted, and the number of clusters K is usually set to 3 - 5. The selection of this range is based on the anatomical structure of the pulmonary blood vessels and can better distinguish the main vascular branches.

[0093] Finally, the method of the present invention extracts the largest connected domain within the marked clustering region as the pulmonary artery vascular tree. This step can effectively remove some small isolated regions and obtain a complete pulmonary artery vascular tree structure.

[0094] Next, the present invention will detail the specific steps for calculating the pressure gradient field inside and on the wall of the pulmonary artery vascular tree. First, the extracted pulmonary artery vascular tree is meshed using the finite element analysis method. Preferably, tetrahedral meshes are used, and the mesh size is typically set to 0.1 - 0.5 mm. This range of selection can achieve a balance between ensuring calculation accuracy and calculation efficiency.

[0095] After the meshing is completed, the method of the present invention sets boundary conditions and applies fluid boundary conditions and wall boundary conditions. Generally, a constant flow boundary condition is set at the pulmonary artery inlet, and the flow value can be set according to the patient's physiological parameters, with a typical value of 4 - 6 L / min. A pressure boundary condition is set at the pulmonary vein outlet, and the pressure value is usually set to 0 - 5 mmHg.

[0096] Subsequently, the method of the present invention calculates the pressure gradient vector, momentum flux, fluid pressure distribution, and total pressure for each element. These calculations are based on the Navier - Stokes equations, and the specific form is as follows:

[0097] ,

[0098] ,

[0099] Among them, is the blood density, is the blood velocity vector, is the time, is the pressure, is the blood viscosity, is the body force. Finally, the method of the present invention calculates the momentum derivative of the blood flow at the wall. This step is of great significance for evaluating the shear stress on the blood vessel wall. The calculation formula for the momentum derivative is as follows:

[0100] ,

[0101] Among them, is the wall normal vector, 、 、 are the components of the velocity in the x, y, and z directions respectively.

[0102] Through the above steps, the method of the present invention can obtain the detailed pressure gradient field inside and on the wall of the pulmonary artery vascular tree, laying a foundation for subsequent perfusion analysis. This analysis method based on watershed dynamics can more accurately reflect the actual situation of pulmonary blood perfusion compared with traditional methods, providing important support for clinical diagnosis and treatment. Next, the present invention will describe in detail the specific steps for extracting the pulmonary vein watershed. First, the method of the present invention extracts the initial streamline set of the blood flow at the current time point. This step usually uses numerical integration methods, such as the fourth-order Runge-Kutta method, to track the trajectories of blood particles. Preferably, in one embodiment of the present invention, the number of initial streamlines is set to 100 - 500, and this range can achieve a good balance between computational efficiency and result accuracy.

[0103] After obtaining the initial streamline set, the method of the present invention calculates the first pressure gradient between adjacent streamline clusters in the initial streamline set. The calculation formula for the pressure gradient is as follows:

[0104] ,

[0105] where, is the pressure, respectively represent the three spatial directions.

[0106] Subsequently, the method calculates the second pressure gradient formed by the connection line between the current point and other points in the initial streamline set. The purpose of this step is to capture the local pressure change situation. Preferably, in an embodiment of the present invention, points within a range of 5 - 10 mm from the current point are selected for calculation, and this range selection can effectively reflect the local pressure change without introducing too much noise.

[0107] Next, the method of the present invention compares the absolute value of the difference between the first pressure gradient and the second pressure gradient. This step is the key to judging the blood flow state. Specifically, the method uses the following judgment criterion:

[0108] ,

[0109] where, is the first pressure gradient, is the second pressure gradient, is the threshold. Preferably, is set to 0.1 mmHg / mm. The selection of this threshold is based on a large amount of experimental data and can effectively distinguish the stable domain and the transition domain of blood flow.

[0110] Based on the above comparison results, the method of the present invention determines whether a point is in the blood flow transition region or the blood flow stable region. If the absolute value of the difference is less than or equal to the threshold, it is determined that the point is in the blood flow transition region; otherwise, it is determined that the point is in the blood flow stable region. This determination method can effectively identify different states of blood flow and provide an important basis for subsequent watershed analysis.

[0111] Finally, based on the determination result, the method determines the shunt boundary of the pulmonary vein watershed. Preferably, in one embodiment of the present invention, the level set method is used to track the evolution of the shunt boundary. The level set function has the following evolution equation:

[0112] ,

[0113] where is the velocity field, which can be constructed according to the pressure gradient information.

[0114] Through the above steps, the method of the present invention can accurately extract the pulmonary vein watershed, laying a foundation for subsequent perfusion analysis. This analysis method based on watershed dynamics can divide the perfusion regions of different lung lobes more precisely than traditional methods, which helps to improve the accuracy of diagnosis.

[0115] Next, the present invention will detail the specific steps for realizing perfusion analysis in combination with dynamic hemodynamic simulation. First, the method takes the pulmonary vein vessel entrance as the outlet point to obtain the streamline information of the previous frame. This step provides the initial conditions for the dynamic simulation. Preferably, in the embodiment of the present invention, the particle tracking method is used to obtain the streamline information, and the number of particles is usually set to 1000 - 5000, and this range can achieve a good balance between computational efficiency and result accuracy.

[0116] Subsequently, the method of the present invention searches for the lowest pressure point on the wall of the pulmonary artery vascular tree as the downstream point. This step is based on the principle of fluid mechanics, and the lowest pressure point usually corresponds to the position with the fastest blood flow velocity. Preferably, the gradient descent algorithm is used for the search, and the search step size is set to 0.1 - 0.5 mm, and this range can ensure the accuracy and efficiency of the search.

[0117] Next, the method takes the downstream point as the starting point of the next frame and determines a new downstream point on the wall of the next frame. This step realizes the continuity of the dynamic simulation. Preferably, in one embodiment of the present invention, the frame interval time is set to 0.01 - 0.05 seconds, and this range can better capture the dynamic changes of blood flow.

[0118] The method of the present invention repeats the above process until the downstream points of the next frame cannot be found on the new frame. This termination condition ensures the integrity of the simulation process. Preferably, when no new downstream points can be found for 5 consecutive frames, it is considered that the termination condition is reached.

[0119] After the termination condition is reached, the method of the present invention calculates the position where the pressure gradient of the blood vessel wall is zero as the end point. This step is based on the theory of fluid mechanics. The position where the pressure gradient is zero usually corresponds to the position of blood flow stagnation or bifurcation. Specifically, the calculation formula of the pressure gradient is as follows:

[0120] ,

[0121] where, is a very small positive number, usually set to 0.01 mmHg / mm.

[0122] Finally, the method of the present invention extracts the entire streamline formed from the initial streamline to the end point on the pulmonary artery blood vessel wall as the perfusion path of the pulmonary artery blood vessel tree. These perfusion paths comprehensively reflect the blood flow in the lungs and provide an important basis for evaluating the blood supply to the lung tissue.

[0123] Through the above steps, the method of the present invention realizes the organic combination of dynamic blood flow simulation and perfusion analysis. Compared with the static analysis method, it can more realistically reflect the changes in pulmonary blood perfusion during the breathing process, which helps to improve the accuracy and comprehensiveness of diagnosis.

[0124] Next, the present invention will describe in detail the specific steps for evaluating the blood supply to the lung tissue. First, the method randomly sets multiple seed points outside the boundary of the pulmonary vein basin. These seed points will serve as reference points for evaluating the blood supply. Preferably, in one embodiment of the present invention, the number of seed points is set to 100 - 500, and this range can achieve a good balance between calculation efficiency and evaluation accuracy.

[0125] Subsequently, the method of the present invention uses the determined perfusion path as the seed point and marks the seed points inside the blood vessel tree through the marker search method. This step can effectively identify different perfusion regions. Preferably, the method uses the breadth - first search algorithm for marking, and the search radius is set to 1 - 3 mm, and this range can better cover the local blood vessel structure.

[0126] Next, the method extracts all perfusion paths containing marked points to obtain a set of perfusion paths. These sets of perfusion paths comprehensively reflect the blood flow in the lungs. Preferably, in the embodiment of the present invention, the perfusion paths are smoothed, and the cubic spline interpolation method is used, and the control point spacing is set to 0.5 - 1 mm, and this range can achieve a balance between maintaining the path shape and reducing noise.

[0127] Then, the method of the present invention calculates the average blood flow velocity of each perfusion path. The calculation of blood flow velocity is based on Poiseuille's law:

[0128] ,

[0129] where, is the average flow velocity, is the blood vessel radius, is the blood viscosity, is the pressure difference, is the blood vessel length.

[0130] Next, this method selects the blood flow in different lung positions and calculates the range of lung ischemia. Preferably, in one embodiment of the present invention, the area where the blood flow is less than 50% of the normal value is defined as the ischemic area. The selection of this threshold is based on a large amount of clinical data and can effectively identify potential lesion areas.

[0131] Finally, the method of the present invention calculates the range of lung ischemia at different time points to obtain the dynamic range of lung ischemia during the patient's breathing process. This step can comprehensively reflect the blood supply situation of the patient's lungs throughout the respiratory cycle. Preferably, this method samples and analyzes 10 - 20 respiratory time points, and this sampling frequency can better capture the blood supply changes during the breathing process.

[0132] Through the above steps, the method of the present invention realizes a comprehensive assessment of the blood supply situation of lung tissue. This assessment method based on dynamic perfusion analysis can more accurately reflect the actual physiological state of the patient compared with the traditional static analysis method, providing an important basis for clinical diagnosis and treatment. Next, the present invention will describe in detail the specific steps of downsampling the pulmonary artery CT angiogram. First, this method performs downsampling on the denoised pulmonary artery vascular tree region. The purpose of this step is to reduce the high-frequency noise in the image while retaining important structural information. Preferably, in one embodiment of the present invention, a Gaussian low-pass filter is used for downsampling, and the standard deviation σ of the filter is set to 0.5 - 1.5 pixels. The selection of this range can achieve a good balance between the denoising effect and the retention of details.

[0133] Subsequently, the method of the present invention performs edge detection on the downsampled pulmonary artery vascular tree region. Edge detection is an important basis for subsequent segmentation. Preferably, this method uses the Canny edge detection operator, and its mathematical expression is as follows:

[0134] ,

[0135] where, [[ID=3"]] is the image gray function, is the edge strength. The low threshold of the Canny operator is usually set to 1 / 2 or 1 / 3 of the high threshold, and this setting can effectively detect the main edges and suppress noise.

[0136] Next, the method of the present invention performs watershed segmentation on the grayscale image after edge detection based on the watershed algorithm to obtain each connected domain image. The watershed algorithm can effectively process complex vascular branch structures. Preferably, in the embodiments of the present invention, the marker-controlled watershed algorithm is adopted, and the selection of marker points is based on local minima, and this method can effectively avoid the over-segmentation problem.

[0137] Finally, the present method performs region growing on each connected domain image within the vascular region to generate pulmonary artery CT angiograms. During the region growing process, the growth criterion is usually based on pixel gray values and spatial continuity. Preferably, in an embodiment of the present invention, the growth threshold is set to ±15% of the gray value of the seed point, and this range can effectively control the growth process and avoid missed segmentation or over-segmentation.

[0138] Through the above steps, the method of the present invention realizes effective downsampling processing of pulmonary artery CT angiograms. This processing method can not only effectively remove image noise, but also retain important vascular structure information, providing high-quality image data for subsequent perfusion analysis.

[0139] Next, the present invention will describe in detail the minimum mean filtering method adopted in the downsampling processing. The minimum mean filtering is a non-linear filtering method, which can effectively remove impulse noise in the image while keeping the edge information of the image. Specifically, the mathematical expression of the minimum mean filtering is as follows:

[0140] ,

[0141] Among them, is the original image, is the filtered image, is the filtering window, is the number of pixels within the window. Preferably, in the embodiments of the present invention, the size of the filtering window is set to 3×3 or 5×5 pixels. The selection of this window size can achieve a good balance between the denoising effect and the calculation efficiency.

[0142] The core idea of the minimum mean filtering is to calculate the means of multiple sub-windows within the filtering window, and then select the minimum mean as the new value of the central pixel. This method can effectively remove larger positive impulse noise while keeping the overall structure of the image.

[0143] It should be noted that the minimum mean filtering has special advantages in processing CT images. Since blood vessels in CT images usually appear as high-brightness regions while the gray values of surrounding tissues are relatively low, the minimum mean filtering can effectively preserve the blood vessel structure while smoothing the surrounding tissues. This is of great significance for subsequent blood vessel segmentation and perfusion analysis.

[0144] Finally, the present invention will detail some key operations in the step of extracting the pulmonary vein basin. First, the method sets a threshold. This threshold is used to filter the pressure gradient calculation results and is a key parameter for determining the boundary of the blood basin area. Preferably, in an embodiment of the present invention, the threshold is set to 20% - 30% of the average pressure gradient. The selection of this range is based on a large amount of experimental data and can effectively distinguish the main blood flow region from the marginal region.

[0145] Subsequently, the method of the present invention filters the pressure gradient calculation results according to the threshold, and only retains the points where the pressure is greater than the pressure gradient threshold. This step can effectively remove some noise points and weak blood flow regions. Preferably, the method further processes the filtered results using morphological operations, such as performing an opening operation with a structuring element having a radius of 2 - 3 pixels, which can remove some small isolated regions.

[0146] Next, the method takes the retained points as the outer boundary of the blood basin area. These points form the preliminary basin boundary. Preferably, in an embodiment of the present invention, spline interpolation is performed on these boundary points to obtain a smooth boundary curve. When interpolating, the distance between control points is usually set to 3 - 5 pixels, and this range can achieve a balance between maintaining boundary details and smoothness.

[0147] Then, the method of the present invention takes the outer boundary as the center and uses the bisection method to gradually determine that the pressure gradient at the boundary is zero to determine the boundary line. This step can accurately locate the actual boundary of the basin area. Preferably, the method sets the number of iterations of the bisection method to 10 - 15 times, and this number of iterations can achieve a good balance between computational efficiency and accuracy.

[0148] Finally, the method takes the boundary line as the boundary line of the pulmonary vein blood basin area and randomly selects seed points outside the boundary line. These seed points will be used for subsequent perfusion path analysis. Preferably, in an embodiment of the present invention, the number of seed points is set to 1 / 10 to 1 / 5 of the boundary line length, and this range can ensure sufficient sampling density without introducing excessive computational burden.

[0149] Through the above steps, the method of the present invention realizes the accurate extraction of the pulmonary vein basin area. This method based on pressure gradient and boundary optimization can more accurately reflect the actual blood flow distribution compared with traditional image segmentation methods, providing a reliable basis for subsequent perfusion analysis.

[0150] The above are only the preferred specific embodiments of the present invention. However, the protection scope of the present invention is not limited thereto. Any person familiar with the art within the scope disclosed by the present invention, making equivalent substitutions or changes according to the solution and improved concept of the present invention, shall be covered by the protection scope of the present invention.

Claims

1. A method for pulmonary vascular perfusion analysis based on watershed dynamics, characterized in that Including: An acquisition step, including: Acquiring a pulmonary artery vascular tree and a pulmonary artery CT angiogram; A processing step, including: Extracting the pulmonary artery vascular tree based on the pulmonary artery vascular tree and the pulmonary artery CT angiogram; Calculating the pressure gradient field inside and on the wall of the pulmonary artery vascular tree based on the pulmonary artery vascular tree; Extracting the pulmonary vein watershed based on the pressure gradient field; Implementing perfusion analysis by combining dynamic hemodynamic simulation based on the pulmonary vein watershed; An output step, including: Evaluating the blood supply situation of the lung tissue based on the perfusion analysis result.

2. The method according to claim 1, characterized in that, The specific extraction of the pulmonary artery vascular tree includes: Segmenting the pulmonary artery vascular tree region based on the pulmonary artery vascular tree and the pulmonary artery CT angiogram; Extracting a candidate region around the pulmonary artery vascular tree region; Extracting seed pixel points within the candidate region; Performing region growing based on the seed pixel points to obtain the connected domain of the pulmonary artery vascular tree; Performing clustering labeling on the connected domain of the pulmonary artery vascular tree; Extracting the largest connected domain from the clustered labeled regions as the pulmonary artery vascular tree.

3. The method according to claim 1, wherein The specific calculation of the pressure gradient field inside and on the wall of the pulmonary artery vascular tree includes: Performing mesh division on the pulmonary artery vascular tree using the finite element analysis method; Setting boundary conditions and applying fluid boundary conditions and wall boundary conditions; Calculating the pressure gradient vector, momentum flux, fluid pressure distribution, and total pressure of each element; Calculating the momentum derivative of the blood flow at the vessel wall.

4. The method according to claim 1, wherein The specific extraction of the pulmonary vein watershed includes: Extracting the initial streamline set of the blood flow at the current time point; Calculating the first pressure gradient between adjacent streamline clusters in the initial streamline set; Calculating the second pressure gradient formed by the connection between the current point and other points in the initial streamline set; Comparing the absolute value of the difference between the first pressure gradient and the second pressure gradient; Based on the comparison result, determining whether the current point is in the blood flow transition domain or the blood flow stable domain; Based on the determination result, determining the shunt boundary of the pulmonary vein watershed.

5. The method according to claim 1, wherein The specific implementation of perfusion analysis by combining dynamic hemodynamic simulation includes: Taking the pulmonary vein vessel entrance as the exit point to obtain the streamline information of the previous frame; Searching for the lowest pressure point on the wall of the pulmonary artery vascular tree as the downstream point; Taking the downstream point as the starting point of the next frame and determining a new downstream point on the wall of the next frame; Repeating the above process until no downstream point for the next frame can be found on the new frame; Calculating the position where the pressure gradient of the vessel wall is zero as the end point; Extracting the entire streamline formed from the initial streamline to the end point on the pulmonary artery vessel wall as the perfusion path of the pulmonary artery vascular tree.

6. The method according to claim 1, wherein The specific evaluation of the blood supply situation of the lung tissue includes: Randomly setting multiple seed points outside the pulmonary vein watershed boundary; Using the determined perfusion path as a seed point and marking the seed points inside the vascular tree through the marker search method; Extracting all perfusion paths containing marked points to obtain a perfusion path set; Calculating the average blood flow velocity of each perfusion path; Selecting the blood flow volumes at different lung positions and calculating the lung ischemia range; Calculating the lung ischemia ranges at different time points to obtain the dynamic lung ischemia range during the patient's breathing process.

7. The method according to claim 1, wherein It also includes denoising the pulmonary artery CT angiogram, and the denoising process includes: Based on the pulmonary artery CT angiogram, extracting the largest connected region in the CT angiogram; Segmenting the largest connected region to obtain the vascular tree region; Randomly selecting a background outside the pulmonary artery vascular region for assignment, so as to form a binary image inside and outside the largest connected region; Smoothing the boundary of the binary image to obtain the denoised pulmonary artery vascular tree region.

8. The method according to claim 1, characterized in that, It also includes downsampling the pulmonary artery CT angiogram, and the downsampling process includes: Performing downsampling on the denoised pulmonary artery vascular tree region; Performing edge detection on the downsampled pulmonary artery vascular tree region; Performing watershed segmentation on the grayscale image after edge detection based on the watershed algorithm to obtain each connected domain image; Performing region growing on each connected domain image within the vascular region to generate the pulmonary artery CT angiogram.

9. The method according to claim 8, characterized in that, The downsampling process uses the method of minimum mean filtering.

10. The method according to claim 1, characterized in that In the step of extracting the pulmonary vein watershed, it also includes: Setting a threshold; Filtering the pressure gradient calculation result according to the threshold, and only retaining the points with a pressure greater than the threshold; Taking the retained points as the outer boundary of the blood watershed; Taking the outer boundary as the center, gradually determining that the pressure gradient at the boundary is zero by using the dichotomy method to determine the boundary line; Taking the boundary line as the pulmonary vein blood watershed boundary line; Randomly selecting seed points outside the boundary line; Connecting each seed point on the boundary line to the nearest point on the pulmonary artery vascular wall; Searching outside the boundary until no next downstream point can be found; Calculating the position where the pressure gradient of the vascular wall is zero as the end point; Extracting the whole streamline formed from the initial streamline to the end point on the vascular wall as the perfusion path of the pulmonary artery vascular tree.

Images