A tracheal tree extraction method combining region growing and morphological analysis

Abstract

A trachea tree extraction method combining region growing and morphological analysis, which divides the extraction of trachea tree into main trachea extraction and fine trachea extraction, adopts adaptive high threshold to determine the growing condition in the main trachea extraction, improves the extraction efficiency of main trachea tree; in the extraction of fine trachea, the skeleton and branch point of main trachea tree are extracted in advance, and the fine trachea is segmented based on the branch point of the skeleton, which reduces the shortcomings of pure morphological and pattern recognition extraction, improves the success rate, reduces the operation amount, and improves the extraction efficiency; after the main trachea and fine trachea are combined, the fine trachea in the broken state is made up based on the fine trachea splicing method, and the false trachea is deleted, so as to improve the overall quality of trachea tree extraction.

Description

Technical Field

[0001] This invention relates to the field of computer image processing technology, specifically to a method for extracting tracheal trees that combines region growing and morphological analysis. Background Technology

[0002] With the development of CT (Computed Tomography) technology, the scanning resolution of CT scans has become increasingly higher. This has made it possible to extract relevant characteristic parameters of the trachea from the scanned images, thereby enabling the diagnosis of other pulmonary diseases such as pulmonary nodules and chronic obstructive pulmonary disease. However, the development of CT has also led to a rapid increase in the number of scans, with a single case now containing hundreds of scans. When detecting tracheal lesions, radiologists must read a large number of scans, and the probability of missed diagnoses and misdiagnoses increases with the number of CT scans. Using CAD (Computer-Aided Design) systems for tracheal disease detection can help reduce the time doctors spend reading images and improve the objectivity of the tracheal detection and quantification process.

[0003] The distribution characteristics of trachea in lung CT images dictate that tracheal segmentation methods can be either gray-scale region-based or edge-based. Many algorithms exist for both methods, but each has its own limitations. In existing gray-scale region-based methods, the main trachea appears as a bright black color in CT images, significantly different from the gray-scale values ​​within the lumen. Therefore, region growing can be used to segment the main trachea. However, a large portion of the bronchioles extends into the lung parenchyma, and during region growing, the lung parenchyma is often grown along with them. To avoid this, strict control of growth constraints is necessary, but this comes at the cost of losing some actual trachea, resulting in incomplete tracheal tree segmentation. Therefore, ensuring both complete segmentation and preventing leakage is a challenge. Edge-based segmentation methods can avoid tracheal leakage, but when segmenting bronchioles, many trachea and the main trachea are often missed and broken. There is no simple and effective method to connect them, and this difficulty is further increased in today's high-precision environments.

[0004] Therefore, how to avoid the shortcomings of various bronchial tree extraction methods, extract the main bronchial tree and the fine bronchial tree separately according to their characteristics, and simultaneously splice the main bronchial tree and the fine bronchial tree, or the fine bronchial trees, so as to be compatible with different extraction methods, have the advantages of both, avoid simply merging the two, reduce the amount of computation, and improve the extraction efficiency has become a technical problem that needs to be solved by the existing technology. Summary of the Invention

[0005] The purpose of this invention is to propose a tracheal tree extraction method that combines regional growth and morphological analysis. This method can improve the extraction methods of the main trachea and the trachea, while avoiding simple merging operations between the two, reducing the computational load of the system, and improving extraction efficiency.

[0006] To achieve this objective, the present invention adopts the following technical solution:

[0007] A method for extracting tracheal trees that combines regional growth and morphological analysis, characterized by comprising the following steps:

[0008] CT image preprocessing step S110:

[0009] Preprocessing CT images reduces image noise, minimizes interfering tissue, and extracts and segments the lung parenchyma region;

[0010] Main tracheal tree segmentation step S120:

[0011] The centroid method is used to select a seed point as the seed point for region growth. The neighboring pixels around the seed point are judged through a 3D 26-neighborhood. A low threshold T1 and a high threshold T2 are set, where T2 = T1 + ΔT. The region growth is completed step by step in an iterative manner. After each iteration, leakage detection is performed. If no leakage occurs, the values ​​of T1 and T2 are increased by ΔT respectively; otherwise, the growth is stopped.

[0012] Skeleton and branch point extraction step S130:

[0013] Extract the 3D skeleton map of the tracheal tree, convert the 3D skeleton map into a 2D image, and define a threshold length T. h Using the threshold length T h Length less than T h The branches are removed, and then the processed two-dimensional image is converted back into a three-dimensional image for trimming. The trimmed three-dimensional image is converted back into a two-dimensional image, and the branch points (Nodes) and branches of the trachea are displayed on the two-dimensional image. The coordinate positions of each branch point are saved.

[0014] Step S140: Extraction of fine tracheal trees

[0015] Using the coordinates of the branch points in step S123 as the center point, morphological processing is performed layer by layer along the position range of the branches to segment the bronchiolar region of the single-layer image. Finally, the segmentation results are reconstructed in three dimensions to synthesize the three-dimensional lung bronchiolar region.

[0016] Step S150 for determining and connecting the main trachea and the trachea:

[0017] The extracted main airways and fine airways are synthesized into the lung airway structure. In the lung CT images of each layer, distance judgment and angle judgment are respectively performed on the endpoints of the main airways and the endpoints of the fine airways, and the endpoints of the main airways and the endpoints of the fine airways that pass through the above two judgments are connected.

[0018] Optionally, in step S110,

[0019] Select a threshold of -600 HU for binary processing of the CT image; then use a 7×7×7 spherical structuring element to perform morphological closing on the binary image to fill the holes in blood vessels and tracheal walls; finally, use this as a mask to isolate the lung parenchyma region from the original image.

[0020] Optionally, the step S120 specifically includes:

[0021] Automatic determination of seed points step S121:

[0022] Select the 10th axial slice at the upper end of the lung parenchyma, and define a rectangular search area in the 2D slice, where the length and width are respectively half of the length and width of the slice; perform binary processing on the search area with a threshold of -750 HU, find the largest connected domain, which corresponds to the main airway, and its centroid is the center of the tracheal cross-section, serving as the seed point for region growing.

[0023] Adaptive threshold increase and iterative growth step S122:

[0024] Judgment is made on the adjacent pixel points around the selected seed point through a three-dimensional 26-neighborhood: the adjacent pixel points with density value L≤T1 are marked as Confirm, that is, the confirmed tracheal region; the adjacent pixel points with density value T1<L<T2 are marked as Ready, and they are regarded as candidate tracheal regions; the pixel points with T2<L are marked as Away, belonging to the unconfirmed region; the region growing is gradually completed in an iterative manner, and low threshold T1 and high threshold T2 are set, where T2 = T1 + ΔT. Each time the iteration is completed, leakage detection is first performed. If there is no leakage, growth continues; if leakage occurs, the next iteration is terminated, and the values of T1 and T2 are respectively increased by ΔT, and the previous Ready region is converted into the Confirm region.

[0025] Optionally, in step S120,

[0026] Introduce a local region mutation index:

[0027] ΔV = V k+1 - V k (1)

[0028] In formula (1), V k and V k+1Let be the tracheal region volume in the k-th and (k+1)-th iterations, respectively. ΔV represents the change in tracheal region volume between adjacent iterations, i.e., the number of Ready pixels in the k-th iteration. Define a ratio coefficient δ.

[0029]

[0030] After each iteration, the ratio coefficient δ is calculated, and leakage occurs when δ > 5%.

[0031] Optionally, in step S120, the initial value T1 = -980HU, ΔT = 5HU.

[0032] Optionally, in step S140, a disk structure unit with rotational invariance is used for morphological processing.

[0033] Optionally, in step S150,

[0034] The distance determination is as follows: Let D be the distance between the endpoint of the bronchiole and the endpoint of the main trachea. 12 If D 12 At the distance threshold D t If the two endpoints are within the range, it is assumed that there may be a break between them.

[0035] The angle determination is as follows: α is the angle α between the slope at the end of the main trachea and the slope at the end of the narrow trachea. t The included angle threshold is defined when α ≤ α t At that time, it was believed that there was a possibility that the bronchioles and the main trachea were connected.

[0036] Optionally, in the angle determination in step S150,

[0037] The calculation method for the angle α between the slopes is as follows: Points A2 and B1 are represented by the centroids of the cross sections at the fracture points of the main trachea and the bronchioles. A1 is the centroid of the tracheal cross section above the fracture point of the main trachea, and B2 is the centroid of the tracheal cross section below the fracture point of the bronchioles. The straight line and the relevant slope are determined by the two points. The slope from point A1 to point A2 is K1, and the slope from point B1 to point B2 is K2. Formula (3) is obtained by the slope dot product theory. The angle α between the two slopes K1 and K2 is calculated by formula (3).

[0038]

[0039] Optionally, in step S150, for those that do not meet the distance and angle criteria, they are identified as pseudo-tracheal regions. The pseudo-tracheal regions are removed by using region growing or all voxel points other than the tracheal regions are set as non-tracheal points. After processing, the final result of lung tracheal tree segmentation is obtained.

[0040] The present invention further discloses a storage medium for storing computer-executable instructions, characterized in that:

[0041] The computer-executable instructions, when executed by the processor, perform the aforementioned tracheal tree extraction method that combines region growth and morphological analysis.

[0042] In summary, the present invention has the following advantages:

[0043] (1) An adaptive high threshold is used to determine the growth conditions in the extraction of the main trachea, thereby realizing the extraction of the main trachea tree and successfully segmenting the main trachea under low leakage conditions.

[0044] (2) The skeleton and branch points of the main tracheal tree are extracted in advance. Based on the branch points of the skeleton, morphological and pattern recognition methods are used to segment the trachea. This not only reduces the shortcomings of simple morphological and pattern recognition extraction and improves the success rate, but also reduces the amount of computation and improves the extraction efficiency.

[0045] (3) Based on the distance criterion and the included angle criterion, qualified tracheal ends and main tracheal ends are determined, and the main trachea and trachea are spliced ​​together. Compared with simply performing an OR operation to merge the two, the presentation quality of the tracheal tree extraction is improved.

[0046] (4) Use region growth to remove pseudo-trachea or set all voxel points except trachea regions as non-trachea points to remove pseudo-trachea regions and improve the overall quality of trachea tree extraction. Attached Figure Description

[0047] Figure 1 This is a flowchart of a tracheal tree extraction method that combines region growth and morphological analysis according to a specific embodiment of the present invention.

[0048] Figure 2 This is a schematic diagram of tracheal tree branch point extraction according to a specific embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram showing the distance between the main trachea and the bronchioles according to a specific embodiment of the present invention.

[0050] Figure 4 This is a schematic diagram of the angled splicing of the main trachea and the bronchioles according to a specific embodiment of the present invention. Detailed Implementation

[0051] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0052] The main feature of this invention is that it divides the extraction of the tracheal tree into main tracheal extraction and fine tracheal extraction, and improves the extraction methods for both. In the extraction of the main tracheal tree, an adaptive high threshold is used to determine the growth conditions, thereby achieving the extraction of the main tracheal tree. In the extraction of the fine tracheal trees, the skeleton and branch points of the main tracheal tree are extracted in advance. Based on the branch points of the skeleton, morphological methods are used to segment the fine tracheal trees, which not only reduces the shortcomings of simple morphological and pattern recognition extraction and improves the success rate, but also reduces the amount of computation and improves the extraction efficiency. After merging the extracted main tracheal trees and fine tracheal trees, a fine tracheal splicing method is used to compensate for the fine tracheal trees that are in a broken state, thereby improving the overall quality of the extracted tracheal tree.

[0053] For details, see Figure 1 The flowchart illustrates a method for extracting tracheal trees based on a combination of regional growth and morphological analysis according to the present invention, comprising the following steps:

[0054] CT image preprocessing step S110:

[0055] Preprocessing of CT images reduces image noise, minimizes interfering tissues, and extracts and segments the lung parenchyma region.

[0056] In one specific embodiment, the present invention selects a threshold of -600HU to binarize the CT image; then, a 7×7×7 spherical structuring element is used to perform a morphological closure operation on the binary image to fill the cavities in the blood vessels and tracheal walls; finally, this is used as a mask to isolate the lung parenchyma region from the original image.

[0057] Main tracheal tree segmentation step S120:

[0058] Since the gray value changes of the lumen of the thicker trachea in CT images are relatively small and are generally wrapped by a complete tracheal wall, this invention improves the region growth method by first setting a relatively conservative low threshold T1, and then adaptively raising the threshold T2 by using the inherent connectivity relationship in the growth process and the mutation index of the local region to determine the growth termination condition in order to achieve tracheal tree pre-segmentation.

[0059] This step includes selecting a seed point as the seed point for region growth using the centroid method, judging the surrounding neighboring pixels using a 3D 26-neighborhood with the seed point as the center, setting a low threshold T1 and a high threshold T2, where T2 = T1 + ΔT, and gradually completing the region growth in an iterative manner. After each iteration, a leakage detection is performed. If no leakage occurs, the values ​​of T1 and T2 are increased by ΔT respectively; otherwise, the growth is stopped.

[0060] Step S120 specifically includes the following sub-steps:

[0061] Automatic seed point determination step S121:

[0062] Select the 10th axial slice at the upper end of the lung parenchyma, and define a rectangular search area in the 2D slice, where the length and width are respectively half of the length and width of the slice. Binarize the search area with a threshold of -750 HU, find the largest connected domain, and this connected region corresponds to the main trachea. Its centroid is the center of the tracheal cross-section, which serves as the seed point for region growing.

[0063] Adaptive threshold increase iterative growth step S122:

[0064] Judge the adjacent pixel points around the selected seed point through a three-dimensional 26-neighborhood: The adjacent pixel points with density value L ≤ T1 are marked as Confirm, that is, the confirmed tracheal area; the adjacent pixel points with density value T1 < L < T2 are marked as Ready and regarded as candidate tracheal areas; the pixel points with T2 < L are marked as Away and belong to the unconfirmed area. Gradually complete the region growth in an iterative manner, set a lower low threshold T1 and a high threshold T2, where T2 = T1 + ΔT. After each iteration is completed, first perform leakage detection. If there is no leakage, continue to grow; if there is leakage, terminate the next iteration, increase the values of T1 and T2 by ΔT respectively, and convert the previous Ready area into the Confirm area.

[0065] In a specific embodiment, set the initial value T1 = -980 HU and ΔT = 5 HU.

[0066] To effectively detect leakage, the present invention introduces a local region mutation index:

[0067] ΔV = V k+1 -V k (1)

[0068] In formula (1), V k and V k+1 are the volumes of the tracheal region in the k-th and (k + 1)-th iterations respectively. ΔV is the change in the volume of the tracheal region between adjacent iterations, that is, the number of Ready pixel points in the k-th iteration. Furthermore, define the ratio coefficient δ.

[0069]

[0070] In the present invention, through specific experiments, it is obtained that leakage occurs when δ > 5%.

[0071] Skeleton and branch point extraction step S130:

[0072] To improve the quality of the extraction of the fine tracheal tree, the present invention first extracts the skeleton and branch points of the main tracheal tree extracted in step S120, and uses the skeleton and branch points as the basis for the next extraction of the fine tracheal tree.

[0073] Specifically, this step involves: extracting the 3D skeleton of the tracheal tree, converting the 3D skeleton into a 2D image, and defining a threshold length T. h Using the threshold length T h Make the length less than T h The branches are removed, and then the processed 2D image is converted back to a 3D image for trimming. The trimmed 3D image is then converted back to a 2D image, and the branching points (Nodes) and branches of the trachea are displayed on this 2D image. The coordinates of each branching point are saved. In this step, the last 2D image refers to the 2D image of each layer.

[0074] like Figure 2 As shown, spheres represent branch points and lines represent branches, in order to count the number of branch points (Node Count) and the number of branches (Branch Count).

[0075] Step S140: Extraction of fine tracheal trees

[0076] Using the coordinates of the branch point (Node) in step S123 as the center point, morphological processing is performed layer by layer along the location range of the branch (Branch) to segment the bronchiolar region of the single-layer image. Finally, the segmentation results are reconstructed in three dimensions to synthesize the three-dimensional lung bronchiolar region.

[0077] In this step, morphological processing is performed based on the coordinates of the branch points and the range of the number of branches, which improves the processing efficiency of the fine tracheal tree and reduces the amount of computation.

[0078] Furthermore, in this step, since the size and shape of the bronchiolar region vary depending on the direction, a disk-shaped structural unit with invariable rotation can be used for morphological processing to improve the accuracy of identification.

[0079] Step S150 for determining and connecting the main trachea and the trachea:

[0080] The main bronchus and bronchioles were obtained through the above steps, and then synthesized into the tracheal structure of the lungs using an OR operation. After synthesis, some bronchioles are still in a broken state. In order to obtain a complete and connected tracheal structure, it is necessary to splice the bronchioles to connect them to the main bronchus. When splicing the bronchioles and the main bronchus, the two ends of the bronchioles and the main bronchus should first be identified.

[0081] This invention determines the two endpoints of the bronchioles and the main bronchus based on distance and angle, and connects the main bronchus and any missed bronchioles through these two determinations.

[0082] This step involves synthesizing the extracted main trachea and trachea into a lung tracheal structure. In the lung CT images at each layer, the distance and angle of the endpoints of the main trachea and the trachea are determined. The endpoints of the main trachea and the trachea that pass the above two determinations are then connected.

[0083] Starting from the topology of the trachea and observing and analyzing the actual fracture sites, it can be found that the size of the broken tracheal region is generally relatively small, and the distance between the two broken ends is relatively close. Therefore, the distance can be used to determine whether the trachea and the main trachea are broken.

[0084] For details, see Figure 3 The distance determination is as follows: the distance between the end point of the bronchiole and the end point of the main trachea is set as D. 12 If D 12 At the distance threshold D t If the two endpoints are within the range, it is assumed that there may be a break between them.

[0085] In one specific embodiment, the distance threshold D t Adjustable, depending on the location, preferably 18mm.

[0086] By observing and analyzing the fracture points of the bronchioles and main trachea in CT images, it can be seen that the bronchioles and main trachea are a complete trachea, which appears as a smooth curve connection in morphology, and the slopes on both sides of the fracture point should not be significantly different.

[0087] See Figure 4 The angle determination is as follows: determining the angle α between the slope at the end of the main trachea and the slope at the end of the narrow trachea. t The included angle threshold is defined when α ≤ α t If the bronchioles and main bronchus are considered to be connected, then the remaining bronchioles and main bronchus need to be re-evaluated using distance and angle criteria.

[0088] Specifically, the calculation method for the angle α between the slopes is as follows: Points A2 and B1 are represented by the centroids of the cross sections at the fracture points of the main trachea and the bronchioles. A1 is taken as the centroid of the tracheal cross section above the fracture point of the main trachea, and B2 is taken as the centroid of the tracheal cross section below the fracture point of the bronchioles. The straight line and the relevant slope are determined by these two points. The slope from point A1 to point A2 is K1, and the slope from point B1 to point B2 is K2. Formula (3) is obtained through the slope dot product theory. The angle α between the two slopes K1 and K2 (0°≤α≤180°) is calculated using formula (3).

[0089]

[0090] Regarding the connection between the main trachea and the severed bronchioles, subjective experience suggests that there is a certain topological similarity at the site of the rupture. The main trachea and bronchioles differ in thickness, and the endpoints of the rupture points often have varying shapes. A simple straight-line connection would disrupt the original topological structure of the trachea. The tracheal segmentation is used as a reference for diagnosis and treatment; therefore, it is crucial to restore the original structure as much as possible. Starting from the main trachea, the shape of the cross-section at the rupture point gradually changes towards the cross-section at the bronchiole rupture point; it is essentially a process of changing shape from one point to another.

[0091] Therefore, in this invention, the connection between the end points of the main trachea and the end points of the fine trachea is specifically achieved by connecting them using a two-dimensional shape interpolation method, extending the line along the two ends of the broken main trachea and fine trachea according to the slope deviation angle.

[0092] Furthermore, after splicing, there may still be some unconnected bronchioles on the outside. The reason they are not connected may be that they do not meet the distance criterion or the angle criterion, and therefore do not have matching main tracheal endpoints.

[0093] In this case, the present invention identifies it as a pseudo-tracheal region. The pseudo-tracheal region is similar to the lung trachea in that it is a low-brightness area in the middle and a high-brightness area around it. The pseudo-trachea can be removed by region growing, which takes advantage of the fact that the entire trachea is a connected structure. Alternatively, all voxel points other than the tracheal region can be set as non-tracheal points. After processing, the final result of lung tracheal tree segmentation is obtained.

[0094] In summary, the present invention has the following advantages:

[0095] (1) An adaptive high threshold is used to determine the growth conditions in the extraction of the main trachea, thereby realizing the extraction of the main trachea tree and successfully segmenting the main trachea under low leakage conditions.

[0096] (2) The skeleton and branch points of the main tracheal tree are extracted in advance. Based on the branch points of the skeleton, morphological and pattern recognition methods are used to segment the trachea. This not only reduces the shortcomings of simple morphological and pattern recognition extraction and improves the success rate, but also reduces the amount of computation and improves the extraction efficiency.

[0097] (3) Based on the distance criterion and the included angle criterion, qualified tracheal ends and main tracheal ends are determined, and the main trachea and trachea are spliced ​​together. Compared with simply performing an OR operation to merge the two, the presentation quality of the tracheal tree extraction is improved.

[0098] (4) Use region growth to remove pseudo-trachea or set all voxel points except trachea regions as non-trachea points to remove pseudo-trachea regions and improve the overall quality of trachea tree extraction.

[0099] The present invention further discloses a storage medium for storing computer-executable instructions, which, when executed by a processor, perform the above-described method for extracting tracheal trees by combining region growth and morphological analysis.

[0100] Obviously, those skilled in the art will understand that the various units or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device, or alternatively, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by the computing device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0101] The above description is a further detailed explanation of the present invention in conjunction with specific preferred embodiments. It should not be considered that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered to fall within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for extracting a tracheal tree by combining region growing and morphological analysis, characterized by, The method comprises the following steps: The CT image preprocessing step S110: The CT image is preprocessed to reduce image noise, reduce interfering tissues, and extract and segment the lung parenchyma region; The main bronchial tree segmentation step S120: The seed point is selected by using the center of mass method as the seed point of region growing, and the surrounding adjacent pixel points are judged by a three-dimensional 26-neighborhood with the seed point as the center and a high threshold value , wherein , the region growing is gradually completed in an iterative manner, and the leakage detection is first completed each time of iteration, if the leakage does not occur, the values of and are respectively increased by , otherwise the growing is stopped; The skeleton and branch point extraction step S130: extracting a three-dimensional skeleton map of the tracheal tree, converting the three-dimensional skeleton map into a two-dimensional image, defining a threshold length , removing branches with a length lower than the threshold length , re-converting the processed two-dimensional image into a three-dimensional image for pruning, re-converting the pruned three-dimensional image into a two-dimensional image, and displaying the branch points (Nodes) and branches (Branches) of the trachea on the two-dimensional image, and saving the coordinate positions corresponding to each branch point; The fine bronchial tree extraction step S140: The coordinate position of the branch point in step S123 is taken as a center point, morphological processing is performed on the position range of the branch in layers, the single-layer image fine bronchial region is segmented, and finally the segmentation result is reconstructed in three dimensions to synthesize a three-dimensional lung fine bronchial region; The main bronchus and fine bronchus judgment and splicing step S150: The extracted main bronchus and fine bronchus are synthesized into a lung bronchial structure, the end points of the main bronchus and the end points of the fine bronchus are respectively subjected to distance judgment and angle judgment in each layer of the lung CT image, and the end points of the main bronchus and the end points of the fine bronchus that pass the distance judgment and the angle judgment are connected.

2. The bronchial tree extraction method according to claim 1, wherein: In step S110, The threshold value is selected as -600 HU for CT image binarization processing; then a 7*7*7 spherical structure element is used for morphological closing operation on the binary image to fill the blood vessel and bronchial wall cavity; and finally, the lung parenchyma region is isolated from the original image by using the mask.

3. The bronchial tree extraction method according to claim 1, wherein: The step S120 specifically comprises: The automatic seed point determination step S121: The 10th axial slice at the upper end of the lung parenchyma is selected, a rectangular search region is demarcated in the 2D slice, the length and width of which are half of the length and width of the slice respectively, the search region is subjected to binarization with a threshold value of -750 HU, the largest connected domain is found, the largest connected domain corresponds to the main bronchus, and the center of the bronchial cross section is taken as a seed point for region growing; The adaptive high threshold iteration growing step S122: The 3D 26-neighborhood of the selected seed point is used to judge the neighboring pixel points: the neighboring pixel points with density value are marked as Confirm, i.e. the confirmed trachea region; the neighboring pixel points with density value are marked as Ready, which are regarded as candidate trachea regions; the pixel points with density value are marked as Away, which belong to the unconfirmed region; the region growing is completed step by step in an iterative manner, and the low threshold value and the high threshold value are set, wherein , in each iteration, the leakage detection is performed first, if no leakage occurs, the growing is continued, if leakage occurs, the next iteration is terminated, the values of and are increased respectively, and the last Ready region is converted into the Confirm region.

4. The bronchial tree extraction method according to claim 3, wherein: In step S120, A local region mutation index is introduced: （1） In formula (1) and The first k Next and the ( k + 1) The volume of the tracheal region in the next iteration, The ratio coefficient is defined as the change in volume of the adjacent iteration tracheal region, i.e., the number of Ready pixel points in the second iteration k ， ​ （2） After each iteration, the ratio coefficient is calculated A leak occurs when ​ 5. The bronchial tree extraction method according to claim 4, wherein: In step S120, the initial value = -980 HU, = 5 HU.

6. The bronchial tree extraction method according to claim 1, wherein: In step S140, a disc structure element with rotational invariance is used for morphological processing.

7. The bronchial tree extraction method according to claim 1, wherein: In step S150, The distance determination is that the distance between the set distal bronchiole end point and the main bronchus end point is , if within the distance threshold , it is considered that the two end points may exist a broken case; The angle is determined as: determining the included angle a between the slope at the end point of the main bronchus and the slope at the end point of the bronchiole, is the angle threshold value, when it is considered that there is a possibility of connection between the bronchiole and the main bronchus.

8. The bronchial tree extraction method according to claim 7, wherein: In the angle judgment in step S150, The calculation method of the included angle α between the slopes is as follows: taking the gravity center of the cross section of the main bronchus and the broken thin bronchus as points and , taking the gravity center of the cross section of the bronchus of the layer above the broken part of the main bronchus as , and taking the gravity center of the cross section of the bronchus of the layer below the broken part of the thin bronchus as , a straight line and the related slopes are determined through the two points, wherein the slope from the point to the point is , the slope from the point to the point is , formula (3) is obtained through the slope point multiplication theory, and the included angle α between the two slopes and is calculated through formula (3). Official (3).

9. The bronchial tree extraction method according to claim 7, wherein: In step S150, for those that do not satisfy the distance judgment and the angle judgment, they are identified as pseudo bronchial regions, the pseudo bronchial regions are removed by using region growing, or all the voxel points except the bronchial regions are set as non-bronchial points, and finally the final result of the lung bronchial tree segmentation is obtained.

10. A storage medium for storing computer executable instructions, wherein: The computer executable instructions, when executed by the processor, perform the method of tracheal tree extraction combining region growing and morphological analysis of any one of claims 1-9.

Images