Automatic localization method of cerebral blood vessels in brain CTA images based on model registration

Abstract

The present application relates to a method for automatically locating cerebral blood vessels in brain CTA images based on model registration. The method includes: using an image registration algorithm, mapping points in each floating image to the coordinate space of a fixed image, and performing the same mapping operation on the cerebral blood vessel centerline corresponding to each floating image to obtain the transformed cerebral blood vessel centerline; calculating the average grayscale value of each point in the brain CTA image registered to the fixed image coordinate space to obtain a brain CTA model; applying a centerline averaging algorithm to the transformed cerebral blood vessel centerline to obtain a model centerline; obtaining a cerebral blood vessel region of interest through the brain CTA model and the model centerline; obtaining the inertial characteristics of the blood vessel from the model centerline mapped to the image to be located, guiding the centerline in the region of interest to automatically locate the cerebral blood vessel in the image, and outputting the location result of the cerebral blood vessel centerline, thereby realizing automatic location of the cerebral blood vessel.

Description

Technical Field

[0001] The present application relates to the technical field of medical image processing, and in particular to a method for automatically locating cerebral blood vessels in brain CTA images based on model registration. Background Art

[0002] In 1895, X-rays were discovered by German physicist Roentgen. The emergence of medical images has had a great impact on traditional medical diagnosis methods. Today, with the emergence and vigorous development of medical imaging technologies such as fundus photography, magnetic resonance imaging, and computed tomography, diagnosis based on medical images has become one of the important auxiliary methods in clinical diagnosis. Medical images can clearly and accurately reflect the distribution of tissues and pathological conditions inside the human body. This information about the patient provides a lot of help for doctors to make clinical diagnoses of patients. At present, medical images have been widely used.

[0003] Although medical imaging has brought great convenience to diagnosis, there are still some shortcomings.

[0004] At present, most radiologists judge the information presented in medical images based on their own medical knowledge and clinical experience. This method of analysis and diagnosis places high demands on the level of doctors. Moreover, this method is inefficient, time-consuming and labor-intensive. The huge amount of medical images has brought an increasing workload to medical workers. In addition, the body structures of different patients are basically the same, but there are still slight differences. Even medical images of the same part of the same patient at different times will bring certain difficulties to diagnosis due to differences in position, lighting and other factors, which further tests the experience and level of medical workers. These reasons have greatly increased the workload of medical workers, and possible misjudgments will have very serious consequences.

[0005] In recent years, the research methods of vascular localization are mainly divided into traditional methods and deep learning methods, but the traditional methods are still mainly used to achieve vascular localization of brain CTA (CT angiography) images. Traditional methods for cerebral vascular localization include active contour models, level sets, multi-scale spatial filtering and other methods. However, the current vascular localization method is difficult to locate cerebral blood vessels due to the complex structure of brain CTA images and the general localization algorithm requires multiple interactions. Summary of the invention

[0006] Based on this, it is necessary to provide a method for automatic positioning of cerebral vessels in brain CTA images based on model registration to address the above technical problems, which solves the problem that the structure of brain CTA images is complex and it is difficult to locate cerebral vessels and the general positioning algorithm requires multiple interactions.

[0007] A method for automatically locating cerebral blood vessels in a brain CTA image based on model registration, the method comprising:

[0008] Selecting one of the multiple brain CTA images as a fixed image and the rest as floating images, mapping points in each floating image to the coordinate space of the fixed image through an image registration algorithm, and performing the same mapping operation on the cerebral blood vessel center lines corresponding to each floating image to obtain the transformed cerebral blood vessel center lines;

[0009] The average gray value of each point in the brain CTA image registered to the fixed image coordinate space is calculated to obtain the brain CTA model;

[0010] Apply the centerline average algorithm to the transformed cerebral vascular centerline to obtain the model centerline;

[0011] Mapping the brain CTA model to the image to be located by an image registration algorithm, and performing the same mapping on the center line of the model, mapping the center line of the model to the image to be located, obtaining a cerebral vascular region of interest with a fixed radius around the center line of the model mapped to the image to be located, and saving the generated region of interest;

[0012] The inertial characteristics of the blood vessels are obtained from the center line of the model mapped to the image to be located, the inertial characteristics are integrated into the characteristic function of the fast marching algorithm, the center line is guided in the region of interest to automatically locate the cerebral blood vessels in the image, and the positioning result of the cerebral blood vessel center line is output.

[0013] In one embodiment, the method further comprises:

[0014] The software for three-dimensionally displaying the brain CTA image and the vascular centerline is used to display the positioning result of the cerebral vascular centerline on the display device.

[0015] In one embodiment, the step of selecting one of the multiple brain CTA images as a fixed image and the rest as floating images, mapping points in the floating image to the coordinate space of the fixed image through an image registration algorithm, and performing the same mapping operation on the cerebral blood vessel centerline corresponding to the floating image to obtain the transformed cerebral blood vessel centerline comprises:

[0016] One of the multiple brain CTA images is selected as a fixed image, and the rest are selected as floating images;

[0017] The fixed image and the floating image are roughly aligned, and then the fixed image and the floating image are registered using rigid transformation, and then further registered using affine transformation, and then a rough B-spline free deformation registration is performed, and then a fine B-spline free deformation registration is performed, and the points in the floating image are continuously iteratively projected into the coordinate space of the fixed image, and the cerebral blood vessel centerline corresponding to the floating image is subjected to the same mapping operation to obtain the transformed cerebral blood vessel centerline.

[0018] In one embodiment, the step of applying a centerline averaging algorithm to the transformed cerebral vascular centerline to obtain the model centerline includes:

[0019] Select a line with the smallest distance to other center lines as the middle center line;

[0020] Calculate the cutting plane for each point of the middle center line and obtain the intersection of the cutting plane with other center lines;

[0021] Then, the intersection points are averaged along the middle center line to obtain the average center line;

[0022] The mean centerline is extended, the longest transformed cerebral vascular centerline is selected, and then its extended distal end is aligned with the endpoint of the mean centerline. The extended portion will serve as a supplement to the mean centerline to obtain the model centerline.

[0023] In one embodiment, the step of acquiring the inertial features of the blood vessel from the center line of the model mapped to the image to be located, integrating the inertial features into the characteristic function of the fast marching algorithm, guiding the center line in the region of interest to automatically locate the cerebral blood vessel in the image, and outputting the location result of the center line of the cerebral blood vessel includes:

[0024] Adding the inertial feature of the center line of the model mapped to the image to be positioned to the feature P of the center line of the blood vessel, and introducing it into the eikonal equation;

[0025] Between the manually marked start and end points, combined with the characteristic function of the fast marching algorithm, forward propagation is performed in the entire region of interest to output the positioning result of the cerebral vascular centerline.

[0026] In one embodiment, the step of performing forward propagation in the entire region of interest between the manually marked start point and end point, combining the characteristic function of the fast marching algorithm, and outputting the positioning result of the cerebral blood vessel centerline includes:

[0027] Combined with the characteristic function of the fast marching algorithm, the energy model E is established;

[0028] Through the energy model E, forward propagation is started from the starting point between the manually marked starting point and end point until the U value of each point in the region of interest is obtained, the minimum path energy from the starting point to the end point is derived, and the positioning result of the cerebral vascular centerline is output.

[0029] In one embodiment, the energy model E is:

[0030] E(C)=∫ Ω (P(C(t))+w)dt

[0031] Where C(t) represents the curve on the three-dimensional cerebrovascular image, P(C(t)) represents the characteristics of the current position on the curve, Ω represents the length of the curve, and the domain is [0, L], where L is the length of the cerebrovascular vessel, E(C) is the energy function on the curve path C, and w is a constant.

[0032] In one embodiment, the characteristic function is:

[0033]

[0034] Among them, i(x) represents the inertial characteristics of cerebrovascular vessels, h(x) represents the cerebrovascular voxel characteristics, v(x) represents the cerebrovascular vesselness enhancement, ε is a positive number used to prevent the denominator from being zero, α, β, γ are the exponents of i(x), h(x), and v(x), respectively, and P(x) is the characteristic function.

[0035] The above-mentioned method for automatic positioning of cerebral blood vessels in brain CTA images based on model registration selects one as a fixed image from multiple brain CTA images and the rest as floating images, maps points in each floating image to the coordinate space of the fixed image through an image registration algorithm, and performs the same mapping operation on the cerebral blood vessel center lines corresponding to each floating image to obtain the transformed cerebral blood vessel center lines; calculates the average grayscale value of each point in the brain CTA image registered to the fixed image coordinate space to obtain a brain CTA model; applies a centerline averaging algorithm to the transformed cerebral blood vessel center lines to obtain a model center line; maps the brain CTA model to the image to be positioned through an image registration algorithm, and The model center line is mapped in the same way, the model center line is mapped to the image to be positioned, a cerebral blood vessel region of interest is obtained with a fixed radius around the model center line mapped to the image to be positioned, and the generated region of interest is saved; the inertial characteristics of the blood vessel are obtained from the model center line mapped to the image to be positioned, the inertial characteristics are integrated into the characteristic function of the fast marching algorithm, the center line is guided in the region of interest to automatically position the cerebral blood vessel in the image, and the positioning result of the cerebral blood vessel center line is output, thereby realizing automatic establishment of the region of interest and automatic positioning of the cerebral blood vessels, solving the problem that the structure of the brain CTA image is complex and it is difficult to position the cerebral blood vessels and the general positioning algorithm requires multiple interactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flowchart of a method for automatically locating cerebral blood vessels in a brain CTA image based on model registration in one embodiment;

[0037] Figure 2 It is a three-section schematic diagram of the grayscale difference between two brain CTA images before registration;

[0038] Figure 3 It is a three-section schematic diagram of the grayscale difference of two brain CTA images after registration;

[0039] Figure 4 It is a point cloud image of the region of interest of the blood vessel;

[0040] Figure 5 This is a schematic diagram of the positioning results of the cerebral vascular centerline. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] In one embodiment, Figure 1As shown, a method for automatically locating cerebral blood vessels in brain CTA images based on model registration is provided, and the method is applied to a terminal as an example for explanation, including the following steps:

[0043] Step S220, select one from the multiple brain CTA images as a fixed image and the rest as floating images, map the points in each floating image to the coordinate space of the fixed image through an image registration algorithm, and perform the same mapping operation on the cerebral vascular centerline corresponding to each floating image to obtain the transformed cerebral vascular centerline.

[0044] The cerebrovascular centerline in the floating image is marked in advance, and each floating image corresponds to a cerebrovascular centerline. In a scene, the points in the floating image A are mapped to the coordinate space of the fixed image, and all mapping operations in the mapping process of the floating image A also act on the cerebrovascular centerline corresponding to the floating image A. The image registration algorithm can be an affine transformation or a B-spline transformation.

[0045] Specifically: Image registration can perform spatial geometric transformation on two or more images taken at different times and angles, so that the points corresponding to the same position in space can be matched one by one, thus achieving the purpose of information fusion. The two images to be registered are called fixed image and moving image, respectively. f and P m To express it, the registration process is the process of iteratively searching for the optimal transformation of the points on the image to be registered to the fixed image. The mapping relationship at the three-dimensional space coordinates (x, y, z) is as follows:

[0046] P f (x,y,z)=P m (T(x,y,z))

[0047] Where T(x,y,z) represents the spatial transformation at the three-dimensional space coordinates (x,y,z).

[0048] The present invention first roughly aligns the two images, then registers them using rigid transformation, and then further registers them using affine transformation. After affine registration, a rough B-spline free-form deformation registration is first performed, and then a fine B-spline free-form deformation registration is performed. The final transformation is obtained by continuous iteration, and this transformation is applied to the marked center line to obtain the transformed cerebrovascular center line.

[0049] Step S240 , calculating the average grayscale value of each point in the brain CTA image registered to the fixed image coordinate space to obtain a brain CTA model.

[0050] Among them, after image registration, the brain CTA images are now in the same model space, and the brain CTA model is obtained by averaging the grayscale values ​​at the same position of each image.

[0051] Step S260, applying a centerline averaging algorithm to the transformed cerebral vascular centerline to obtain a model centerline.

[0052] Specifically: The vascular centerlines of different brain CTA images are also mapped to the same model space. First, a line with the smallest distance to other centerlines is selected, which is called the middle centerline. Calculate the cutting plane for each point on the middle centerline to obtain the intersection of the cutting plane and other centerlines. Then average these intersections along the middle centerline to obtain the average value of the centerline. Finally, it is necessary to extend the centerline mean obtained in the previous step, because sometimes the centerline mean may not be long enough to guide the extraction of the centerline in the image to be processed. To solve this problem, we first select the longest centerline, and then align its distal end with the endpoint of the average centerline we obtained, and the extended part will serve as a supplement to the average centerline. At this time, the centerline belonging to the model is obtained.

[0053] The prior model is formed by the brain CTA model and the center line of the model. Since the prior model has certain generalization characteristics, the unknown brain CTA image to be processed can be registered with the prior model to obtain the prior information of the cerebral blood vessels, including the inertial characteristics of the cerebral blood vessels and the region of interest. This information is then combined with the fast marching algorithm to achieve automatic positioning of the cerebral blood vessels.

[0054] Step S280, mapping the brain CTA model to the image to be located through the image registration algorithm, and performing the same mapping on the model centerline, mapping the model centerline to the image to be located, obtaining the cerebral vascular region of interest with a fixed radius around the model centerline mapped to the image to be located, and saving the generated region of interest.

[0055] Among them, the image to be positioned is a brain CTA image to be processed that has not yet been positioned for cerebrovascular positioning. The brain CTA model is aligned to the image to be positioned, and the center line of the model is mapped in the same way. The center line of the model is used as the axis, and the region of interest of the cerebrovascular is established around the center line of the model with a fixed radius from the image to be positioned. Finally, the fast marching algorithm guided by inertial features is used to realize the automatic positioning of cerebrovascular in the region of interest. The region of interest can greatly reduce the search area of ​​the centerline positioning algorithm and improve the efficiency of automatic positioning of cerebrovascular.

[0056] Step S300, obtaining the inertial features of the blood vessels from the center line of the model mapped to the image to be located, integrating the inertial features into the characteristic function of the fast marching algorithm, guiding the center line in the region of interest to automatically locate the cerebral blood vessels in the image, and outputting the positioning result of the cerebral blood vessel center line.

[0057] After obtaining the region of interest of the image to be located, the inertial feature provided by the model centerline is added to the feature P of the blood vessel centerline, as well as the manually marked start and end points, and introduced into the Eikonal Equation. Between the manually marked start and end points, the fast marching algorithm is used to perform forward propagation in the entire region of interest to achieve automatic positioning of cerebral blood vessels.

[0058] The eikonal equation is:

[0059]

[0060] Where x is the current position, F(x) is the speed of propagation at x, and T(x) is the time function to reach the current position x. Represents the gradient of T(x).

[0061] The above-mentioned method for automatic positioning of cerebral blood vessels in brain CTA images based on model registration selects one as a fixed image from multiple brain CTA images and the rest as floating images, maps the points in each floating image to the coordinate space of the fixed image through an image registration algorithm, and performs the same mapping operation on the cerebral blood vessel center lines corresponding to each floating image to obtain the transformed cerebral blood vessel center lines; calculates the average grayscale value of each point in the brain CTA image registered to the fixed image coordinate space to obtain a brain CTA model; applies the centerline averaging algorithm to the transformed cerebral blood vessel center lines to obtain the model center lines; maps the brain CTA model to the image to be positioned through the image registration algorithm, and The same mapping is performed on the model centerline, and the model centerline is mapped to the image to be positioned. The cerebral blood vessel region of interest is obtained with a fixed radius around the model centerline mapped to the image to be positioned, and the generated region of interest is saved; the inertial characteristics of the blood vessel are obtained from the model centerline mapped to the image to be positioned, and the inertial characteristics are integrated into the characteristic function of the fast marching algorithm. The centerline is guided in the region of interest to automatically position the cerebral blood vessels in the image, and the positioning result of the cerebral blood vessel centerline is output, thereby realizing automatic establishment of the region of interest and automatic positioning of the cerebral blood vessels, solving the problem that the complex structure of brain CTA images makes it difficult to locate cerebral blood vessels and the general positioning algorithm requires multiple interactions.

[0062] In one embodiment, the method further comprises:

[0063] The software for three-dimensional display of brain CTA images and vascular centerline is used to display the positioning results of the cerebral vascular centerline on the display device.

[0064] Among them, a software that can display the brain CTA image and the blood vessel centerline in three dimensions is pre-written. After obtaining the positioning result of the cerebrovascular centerline, the software is used to display the positioning result of the image to be positioned and the cerebrovascular centerline. A software that can display the brain CTA image and the blood vessel centerline in three dimensions can be written using Visual Studio, C++, Qt, etc. After obtaining the positioning result of the cerebrovascular centerline of the image to be positioned, the positioning result of the image to be positioned and the cerebrovascular centerline is used as the input of the three-dimensional display software, and it can be seen that the automatic positioning of the cerebrovascular is achieved.

[0065] In one embodiment, one is selected from a plurality of brain CTA images as a fixed image, and the rest are used as floating images. The points in the floating image are mapped to the coordinate space of the fixed image through an image registration algorithm, and the cerebral blood vessel centerline corresponding to the floating image is subjected to the same mapping operation to obtain the transformed cerebral blood vessel centerline, including the following steps: selecting one from a plurality of brain CTA images as a fixed image, and the rest are used as floating images; roughly aligning the fixed image and the floating image, and then registering the fixed image and the floating image using a rigid transformation, further registering them using an affine transformation, and then performing a rough B-spline free-form deformation registration, and then performing a fine B-spline free-form deformation registration, and continuously and iteratively mapping the points in the floating image to the coordinate space of the fixed image, and performing the same mapping operation on the cerebral blood vessel centerline corresponding to the floating image to obtain the transformed cerebral blood vessel centerline.

[0066] In one embodiment, the steps of applying the centerline averaging algorithm to the transformed cerebral blood vessel centerline to obtain the model centerline include: selecting a line with the smallest distance from other centerlines as the middle centerline; calculating the cutting plane of each point on the middle centerline to obtain the intersection of the cutting plane and other centerlines; then averaging the intersections along the middle centerline to obtain the average centerline; extending the average centerline to select the longest transformed cerebral blood vessel centerline, then aligning its extended distal end with the endpoint of the average centerline, and the extended portion will serve as a supplement to the average centerline to obtain the model centerline.

[0067] In one embodiment, the steps of acquiring the inertial features of the blood vessel from the center line of the model mapped to the image to be positioned, integrating the inertial features into the characteristic function of the fast marching algorithm, guiding the center line in the region of interest to automatically position the cerebral blood vessel in the image, and outputting the positioning result of the cerebral blood vessel center line include: adding the inertial features of the center line of the model mapped to the image to be positioned to the feature P of the blood vessel center line, and introducing it into the eikonal equation; performing forward propagation in the entire region of interest in combination with the characteristic function of the fast marching algorithm between the manually marked starting point and end point, and outputting the positioning result of the cerebral blood vessel center line.

[0068] In one embodiment, the steps of forward propagation in the entire region of interest between the manually marked starting point and the end point, combined with the characteristic function of the fast marching algorithm, and outputting the positioning result of the cerebral blood vessel centerline include: establishing an energy model E in combination with the characteristic function of the fast marching algorithm; using the energy model E, forward propagating from the starting point between the manually marked starting point and the end point until the U value of each point in the region of interest is obtained, deriving the minimum path energy from the starting point to the end point, and outputting the positioning result of the cerebral blood vessel centerline.

[0069] Among them, the characteristic P of the blood vessel centerline takes a smaller value near the cerebral blood vessel characteristics. The constant w is introduced to keep the expression of the energy model E positive. By finding the path with the smallest integral of P+w, the centerline positioning problem is transformed into a problem of finding the global minimum. The energy model E is:

[0070] E(C)=∫ Ω (P(C(t))+W)dt

[0071] Where C(t) represents the curve on the three-dimensional cerebrovascular image, P(C(t)) represents the characteristics of the current position on the curve, Ω represents the length of the curve, and the domain is [0, L], where L is the length of the cerebrovascular vessel, E(C) is the energy function on the curve path C, and w is a constant.

[0072] Starting from the starting point, forward propagation is performed until the U value of each point in the area of ​​interest is obtained. Then, the minimum path energy from the starting point to the end point can be derived, which is defined as:

[0073]

[0074] Among them, U is defined as the minimum value of the path integral from the starting point to any point, represents the set of all paths between the starting point and the end point, C(t) represents the curve on the three-dimensional cerebrovascular image, P(C(t)) represents the characteristics of the current position on the curve, and U(p) is the minimum value of the integral of all paths between the starting point and the end point.

[0075] In one embodiment, the characteristic function is:

[0076]

[0077] Among them, i(x) represents the inertial characteristics of cerebrovascular vessels, h(x) represents the cerebrovascular voxel characteristics, v(x) represents the cerebrovascular vesselness enhancement, ε is a positive number used to prevent the denominator from being zero, α, β, γ are the exponents of i(x), h(x), and v(x), respectively, and P(x) is the characteristic function.

[0078] The inertial feature i(x) of the cerebrovascular vessel is obtained through the model centerline. During the cerebrovascular centerline tracking process, the positioning results of the model centerline and the cerebrovascular centerline are divided into a certain number of segments for processing. When locating the first segment of the cerebrovascular centerline, a characteristic function without inertial features is used for positioning. The positioning of the remaining segments is based on the positioning result of a segment on the cerebrovascular centerline, and the corresponding model centerline is mapped to the positioning result of the previous segment, and the same mapping is applied to the model centerline of the current segment to guide the subsequent positioning process.

[0079]

[0080] Where, e is a natural constant, d(x) is the distance between the given position in the fast marching algorithm and the starting point of the current segment, and μ d is the distance between the guidance position provided by the mapped brain CTA model and the model centerline and the starting point of the current segment, σ d Normalize the distance parameter.

[0081] The cerebrovascular voxel feature h(x) is defined as the feature of the voxel gray value at each position in the region of interest of the image to be located.

[0082]

[0083] Where M(x) is the current value of the grayscale of the voxel at a given position in the image to be located, μ g is the mean value of the voxel grayscale of the entire region of interest in the image to be located.

[0084] Compared with the existing technology that requires manual annotation of the region of interest and interactive actions such as marking the start and end points, the above-mentioned method for automatic positioning of cerebral blood vessels in brain CTA images based on model registration only requires one interaction of marking the start and end points to achieve automatic positioning of cerebral blood vessels, solving the problem that there are many cerebral blood vessels in brain CTA images with complex structures, cerebral blood vessels are difficult to locate, and general positioning algorithms require multiple interactions. By establishing a brain CTA model and a model centerline, the cerebral blood vessel region of interest in the brain CTA image to be processed is obtained, and by integrating the inertial features provided by the brain CTA model and the model centerline into the characteristic function of the fast marching algorithm, automatic positioning of cerebral blood vessels is achieved.

[0085] To further demonstrate the beneficial effects of the above-mentioned method for automatic positioning of cerebral vessels in brain CTA images based on model registration, see Figure 2-Figure 5 , Figure 2 The figure shows two brain CTA images in the same coordinate space before the profile alignment. It can be seen that the positions of the skull and other parts of the two brain CTA images are quite different. Figure 3 It can be seen that after the two brain CTA images are registered in the same coordinate space, their contours overlap, achieving a good registration effect. Figure 4 It can be seen that the region of interest is obtained from the brain CTA image, and cerebrovascular localization in the region of interest can greatly reduce the search area of ​​the centerline localization algorithm and improve the efficiency of automatic cerebrovascular localization. Figure 5 It can be seen that the automatic positioning of the cerebral vascular centerline was successfully achieved.

[0086] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0087] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for automatically locating cerebral blood vessels in brain CTA images based on model registration, characterized in that: The method comprises: Selecting one of the multiple brain CTA images as a fixed image and the rest as floating images, mapping points in each floating image to the coordinate space of the fixed image through an image registration algorithm, and performing the same mapping operation on the cerebral blood vessel center lines corresponding to each floating image to obtain the transformed cerebral blood vessel center lines; The average gray value of each point in the brain CTA image registered to the fixed image coordinate space is calculated to obtain the brain CTA model; Apply the centerline average algorithm to the transformed cerebral vascular centerline to obtain the model centerline; Mapping the brain CTA model to the image to be located by an image registration algorithm, and performing the same mapping on the center line of the model, mapping the center line of the model to the image to be located, obtaining a cerebral vascular region of interest with a fixed radius around the center line of the model mapped to the image to be located, and saving the generated region of interest; Acquire the inertial features of the blood vessels from the center line of the model mapped to the image to be located, integrate the inertial features into the characteristic function of the fast marching algorithm, guide the center line in the region of interest to automatically locate the cerebral blood vessels in the image, and output the location result of the center line of the cerebral blood vessels; The method further comprises: Using a three-dimensional display software for brain CTA images and vascular centerlines, the positioning result of the cerebral vascular centerline is displayed on a display device; The step of selecting one from a plurality of brain CTA images as a fixed image and the rest as floating images, mapping points in the floating image to the coordinate space of the fixed image through an image registration algorithm, and performing the same mapping operation on the cerebral blood vessel centerline corresponding to the floating image to obtain the transformed cerebral blood vessel centerline comprises: One of the multiple brain CTA images is selected as a fixed image, and the rest are selected as floating images; Roughly aligning the fixed image and the floating image, registering the fixed image and the floating image using rigid transformation, further registering them using affine transformation, performing rough B-spline free deformation registration, and then performing fine B-spline free deformation registration, iteratively mapping points in the floating image to the coordinate space of the fixed image, and performing the same mapping operation on the cerebral blood vessel centerline corresponding to the floating image to obtain the transformed cerebral blood vessel centerline; The step of applying a centerline averaging algorithm to the transformed cerebral blood vessel centerline to obtain the model centerline includes: Select a line with the smallest distance to other center lines as the middle center line; Calculate the cutting plane for each point of the middle center line and obtain the intersection of the cutting plane with other center lines; Then, the intersection points are averaged along the middle center line to obtain the average center line; Extending the mean centerline, selecting the longest transformed cerebrovascular centerline, and then aligning the extended distal end with the endpoint of the mean centerline, and the extended portion will serve as a supplement to the mean centerline to obtain the model centerline; The steps of acquiring the inertial features of the blood vessels from the center line of the model mapped to the image to be positioned, integrating the inertial features into the characteristic function of the fast marching algorithm, guiding the center line in the region of interest to automatically position the cerebral blood vessels in the image, and outputting the positioning result of the center line of the cerebral blood vessels include: The inertial feature of the model centerline mapped to the image to be positioned is added to the feature P of the blood vessel centerline, and introduced into the eikonal equation, which is: Where x is the current position, F(x) is the speed of propagation at x, and T(x) is the time function to reach the current position x. represents the gradient of T(x); Between the manually marked start and end points, the characteristic function of the fast marching algorithm is combined to perform forward propagation in the entire region of interest and output the positioning result of the cerebral vascular centerline; The step of performing forward propagation in the entire region of interest between the manually marked starting point and end point, combining the characteristic function of the fast marching algorithm, and outputting the positioning result of the cerebral blood vessel centerline includes: Combined with the characteristic function of the fast marching algorithm, the energy model E is established; Through the energy model E, between the manually marked start and end points, forward propagation is started from the start point until the U value of each point in the region of interest is obtained, the minimum path energy from the start point to the end point is derived, and the positioning result of the cerebral vascular centerline is output; The energy model E is: E(C)=∫ Ω (P(C(t))+w)dt Where C(t) represents the curve on the three-dimensional cerebrovascular image, P(C(t)) represents the characteristics of the current position on the curve, Ω represents the length of the curve, and the domain is [0, L], where L is the length of the cerebrovascular vessel, E(C) is the energy function on the curve path C, and w is a constant; Propagate forward from the starting point until the U value of each point in the area of ​​interest is obtained, and the minimum path energy from the starting point to the end point is derived, which is defined as: Among them, U is defined as the minimum value of the path integral from the starting point to any point, represents the set of all paths between the starting point and the end point, C(t) represents the curve on the three-dimensional cerebrovascular image, P(C(t)) represents the characteristics of the current position on the curve, and U(p) is the minimum value of the integral of all paths between the starting point and the end point; The characteristic function is: Among them, i(x) represents the inertial characteristics of cerebrovascular vessels, h(x) represents the voxel characteristics of cerebrovascular vessels, v(x) represents the enhancement of cerebrovascular vesselness, ε is a positive number used to prevent the denominator from being zero, α, β, γ are the exponents of i(x), h(x), v(x), respectively, and P(x) is the characteristic function; The inertial feature i(x) of the cerebrovascular vessel is obtained through the model centerline. In the process of cerebrovascular centerline tracking, the positioning results of the model centerline and the cerebrovascular centerline are divided into a certain number of segments for processing. When locating the first segment of the cerebrovascular centerline, a characteristic function without inertial features is used for positioning. The positioning of the remaining segments is based on the positioning result of a segment on the cerebrovascular centerline. The corresponding model centerline is mapped to the positioning result of the previous segment, and the same mapping is applied to the model centerline of the current segment to guide the subsequent positioning process, where: Where, e is a natural constant, xd(x) is the distance between the given position in the fast marching algorithm and the starting point of the current segment, and μ d is the distance between the guidance position provided by the mapped brain CTA model and the model centerline and the starting point of the current segment, σ d Normalize the parameters for distance; The cerebrovascular voxel feature h(x) is defined as the feature of the voxel grayscale value at each position in the region of interest of the image to be located, where: Among them, g(x) is the current value of the grayscale of the voxel at a given position in the image to be located, μ g is the mean value of the voxel grayscale of the entire region of interest in the image to be located.

Images