CT image-based lung lobe segmentation method and device

A technology of CT images and lung lobes, which is applied in the field of image processing, can solve problems such as no lung lobes segmentation, and achieve the effect of saving time

Active Publication Date: 2017-11-24
SHENYANG NEUSOFT MEDICAL SYST CO LTD
3 Cites 19 Cited by

AI-Extracted Technical Summary

Problems solved by technology

[0004] At present, chest solutions in the prior art mainly focus on whole lung segmentation, tracheal segme...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Method used

By screening the lobe fissure points extracted, some points that cannot highlight the lobe fissure surface, and some lobe fissure points that include less voxels are removed, so that the fissure surface of the lobe lung that can be constructed can be ensured High accuracy.
With the method for step 401, the lung lobe crack point on all planes is screened, the lung lobe crack point with the maximum z value on all planes screened out forms the lung lobe crack point set; by this method, a part of the crack point is removed, so that The accurac...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Abstract

The present invention provides a CT image-based lung lobe segmentation method and device. The method includes the following steps that: a lung region is extracted from a CT image, the lung region is divided into a left lung region and a right lung region; lung lobe crack points are extracted from the left lung region and the right lung region; and a lung lobe crack surface is constructed on the basis of the lung lobe crack points. According to the CT image-based lung lobe segmentation method and device of the invention, the lung region is extracted from the CT image, and the lung region is divided into the left lung region and the right lung region, wherein the left lung region contains two lung lobes, and the right lung region contains three lung lobes; the lung lobe crack points are extracted from the left lung region and the right lung region; and the lung lobe crack surface is constructed on the basis of the lung lobe crack points. With the method and device of the invention adopted, lung lobe segmentation can be automatically performed; after the lung lobe segmentation is performed through the method, if five clear lung lobes cannot be obtained, it can be indicated that physiological abnormalities exist in lungs, or tumors exists around the in the lungs. Since the method can automatically segment the lungs into five lung lobes, if the method fails to obtain five lung lobes through segmentation, it is indicated that the abnormalities exist in the lungs, and therefore, a lot of time can be saved for doctors; and the doctors do not have to spend a lot of time manually analyzing CT images to find suspicious areas.

Application Domain

Technology Topic

Wilms' tumorLung lobe +4

Image

  • CT image-based lung lobe segmentation method and device
  • CT image-based lung lobe segmentation method and device
  • CT image-based lung lobe segmentation method and device

Examples

  • Experimental program(4)
  • Effect test(1)

Example Embodiment

[0066] Method embodiment one:
[0067] See figure 1 This figure is a flowchart of Embodiment 1 of a method for segmentation of lung lobes based on CT images provided by the present invention.
[0068] The method for segmentation of lung lobes based on CT images provided in this embodiment includes the following steps:
[0069] Step 101: Extract a lung area from a CT image, and divide the lung area into a left lung area and a right lung area.
[0070] It should be noted that when taking a CT image of the lungs, the image of the chest is usually directly taken. Therefore, it is necessary to extract the lung area from the chest CT image and determine the boundary between the left lung area and the right lung area. After determining the area of ​​the left and right lungs, the lung lobes can be segmented in the left and right lung areas. Generally, the lung area image extracted from the CT image, such as figure 2 Shown. The left lung area includes 2 lung lobes, and the right lung area includes 3 lung lobes. Knowing the number of lung lobes in the left and right lung regions helps in the next step of segmentation.
[0071] Step 102: Extracting lung lobe crack points from the left lung area and the right lung area.
[0072] The CT images corresponding to the left lung area and the right lung area contain many points, some of which are lung lobes crack points. The purpose of extracting the lung lobes crack points is to find the separation between the lung lobes and the lung lobes, which can also be called cracks. Lung lobe fissure points can be used to construct lung lobe fissure surfaces. However, in CT images, some points are interference points caused by external interference during the process of acquiring CT images, and some points are points on other tissues and organs in the human body obtained during the process of capturing CT images, such as cracks. Local bright spots, blood vessel wall signal points, etc., these points can not be used as lung lobe crack points to construct the lung lobe fissure surface, which will cause certain interference to the extraction of lung lobe fissure points. Therefore, it is necessary to extract the lobular fissure points that can represent the septum of the lobes from the corresponding CT images in the left lung area and the right lung area, and no interference points are needed.
[0073] Step 103: Construct a lung lobe fissure surface from the extracted lung lobe fissure points.
[0074] After extracting the crack points of the lung lobes, the surface fitting algorithm can be used to fit the extracted crack points of the lung lobes to the surface, and then the lung lobes fissure surface is obtained, that is, the surface is constructed by the points.
[0075] First, extract the lung area from the CT image, and divide the lung area into a left lung area and a right lung area; because the left lung area includes 2 lung lobes, and the right lung area includes 3 lung lobes. Lung lobe fissure points are extracted from the left lung area and the right lung area; the lung lobe fissure surface is constructed from the extracted lung lobe fissure points. Such as image 3 with Figure 4 As shown, the two-dimensional image and the three-dimensional image obtained after lung lobes segmentation are respectively. It can be seen from the images that this method can automatically achieve lung lobes segmentation, such as Figure 5 The image shown is the image obtained after lung lobes segmentation using the method provided by the present invention, from Figure 5 It can be seen that the five lung lobes cannot be clearly segmented from the CT image, which indicates that the lung itself has physiological abnormalities or there are tumors around the lungs. With this method, five lung lobes can be segmented automatically based on CT images. If five lung lobes cannot be segmented, the lungs are abnormal. Doctors don’t have to spend a lot of time manually analyzing CT images to find suspicious areas, saving doctors a lot of money. time.

Example Embodiment

[0076] Method embodiment two:
[0077] The initial position of each lobe is connected to a bronchus. Following this physiological characteristic, this embodiment divides the airway into two levels. The first level is the trachea, and the second level is the bronchus. The end of the bronchi is used to find the initial position of each lung lobe. The following describes the lung trachea in the CT image Perform segmentation to determine the initial position of the lung lobe.
[0078] See Image 6 This figure is a flowchart of Embodiment 2 of the method for segmentation of lung lobes based on CT images provided by the present invention.
[0079] The lung lobe segmentation method provided in this embodiment includes:
[0080] Step 201: Extract the lung area from the CT image.
[0081] Step 202: Divide the lung area into a left lung area and a right lung area.
[0082] Step 203: segment the trachea and bronchus from the CT image of the lung area.
[0083] Step 202 and step 203 have no sequence relationship, and are two independent steps, which can be executed in parallel.
[0084] Using the lung area extracted from the CT image in step 201, the boundary of the lung area is determined. And according to the boundary of the lung area to find the starting point of the trachea. Then, the lung trachea is further segmented. In this embodiment, the lung airway is divided into two levels, and the end of the bronchus can be used to accurately locate the initial position of the lung lobe.
[0085] The following briefly introduces the method of trachea and bronchus segmentation:
[0086] The region growth algorithm based on "leakage" control is selected to perform preliminary segmentation of the lung trachea; according to the knowledge of human anatomy, the part of the trachea whose width is greater than a preset threshold is removed. In this example, the preset threshold is set It is 80mm, of course, it can also be set to other values, and there is no limitation here. Further identify the largest connected component and correct the morphological closure. Perform further branch processing on the trachea after the preliminary segmentation, use the wavefront propagation algorithm to detect the bifurcation, and filter out the part that exceeds the variable threshold. In this embodiment, the variable threshold can be set to 5 times the current tracheal cross-sectional area . Similarly, the variable threshold can also be set to other values ​​as required, and this embodiment does not make any limitation on this. After filtering out the part that exceeds the variable threshold, the preliminary segmentation of the lung trachea is basically completed. Since the trachea of ​​the lungs are distributed in a tree shape, it is necessary to extract the center line of the tracheal tree using a skeletonization algorithm, and mark its branch structure, that is, to segment the bronchus.
[0087] Step 204: Determine the initial position of the lung lobes from the ends of the segmented bronchi, each of which is connected to a bronchus.
[0088] The initial position of the lung lobes can be determined according to the trachea and bronchi segmented in step 203. The physiological structure of the trachea is that the main trachea connects to the bronchus, and the five bronchus connects to the five lung lobes. That is, each lobe is connected to a bronchus, and the end position of each bronchus is the beginning position of each lobe. Therefore, after segmenting the bronchus, the starting position of the lung lobe can be determined.
[0089] If there is a lung lobe that is not connected to the end of the segmented bronchi, it can be determined that the lung lobe may be abnormal.
[0090] Step 205: In the left lung area and the right lung area separated in step 202, extract the crack points of the lung lobes.
[0091] Step 206: Construct a lung lobe fissure surface from the extracted lung lobe fissure points and the initial position of the lung lobe.
[0092] That is, using the initial position of the lung lobes can assist in constructing the lobular fissure surface, and also help to screen the abnormalities of the lung lobes.
[0093] On the basis of extracting the crack points of the lung lobes, and further constructing the lung lobes fissure surface according to the initial position of the lung lobes, the edge position of the lung lobes fissure surface can be determined, and the constructed lung lobes fissure surface can be more accurate.

Example Embodiment

[0094] Method embodiment three:
[0095] Since there are many image points in the acquired CT image, some of the image points may be interference points caused by certain factors in the process of acquiring the CT image, and some points may be points on other organs of the human body. Such as local bright spots in cracks, signal points on blood vessel wall, etc. The above-mentioned image points will all cause interference in the process of extracting the crack points of the lung lobes. Therefore, in order to extract the crack points of the lung lobes more accurately, these interference points need to be suppressed.
[0096] See Figure 7 This figure is a flowchart of Embodiment 3 of the method for segmentation of lung lobes based on CT images provided by the present invention.
[0097] The CT image-based lung lobe segmentation method provided in this embodiment includes:
[0098] Step 301: Extract a lung area from a CT image, and divide the lung area into a left lung area and a right lung area.
[0099] Step 302: Gaussian filtering is performed on the left lung area and the right lung area through a Gaussian filter to filter out noise in the image.
[0100] In this embodiment, a Gaussian filter is selected to filter out the noise in the image, and the Gaussian filter will react strongly near the crack. Since the size of the crack in the lung lobe is around 1.5mm, interference signals that are not around 1.5mm are filtered out. In this embodiment, the predetermined size of the Gaussian filter is selected from 1 mm to 2 mm, and the purpose is to filter out independent points with a size smaller than 1 mm and a size larger than 2 mm. The independent points are too far from the curve corresponding to the lung lobe cracks, so the plane characteristics of the cracks are captured by filtering these independent points.
[0101] Step 303: Inhibit the local bright spots of the cracks in the left lung area and the right lung area, inhibit the signal points of the blood vessel wall, inhibit the crack points near the blood vessel, and inhibit the non-planar and non-curved crack points.
[0102] Since the blood vessel wall is similar to the septum of the lung lobes in the CT image, in order to avoid the influence of the blood vessels on the segmentation of the lung lobes, it is necessary to suppress the blood vessels in the CT image.
[0103] In the CT image, there will be many crack local bright spots, blood vessel wall signal points, crack points near the blood vessel wall, non-planar and non-curved crack points, and the existence of these points will cause interference in the process of extracting the crack points of the lung lobes. Affect the extraction of crack points in lung lobes. Therefore, when extracting the crack points of the lung lobe, the local bright spots of the cracks, the signal points of the blood vessel wall, the crack points near the blood vessel wall, and the non-planar and non-curved crack points can be suppressed.
[0104] After Gaussian filtering, a Hessian matrix is ​​constructed for each pixel in the left lung area and the right lung area, and three eigenvalues ​​λ of the Hessian matrix are obtained. 1 , Λ 2 And λ 3 , And |λ 1 |≤|λ 2 |≤|λ 3 |;
[0105] Extract lung lobe crack points according to the following formula, and extract pixels corresponding to F greater than a predetermined threshold as the lung lobe crack points;
[0106] F=ΓF plane F wall F vessels
[0107] Among them, the first factor Γ is used to suppress the local bright spots of the crack, that is, to suppress the maximum eigenvalue λ 3 Positive points:
[0108]
[0109] Among them, the second factor F plane By searching for two obviously different maximum eigenvalue positions to detect flat or curved structures, to suppress non-planar and non-curved cracks, F plane The closer the value of is to 1, the more likely the point is on the plane or curve, F plane The closer the value of is to 0, the farther away the point is from the plane or the curve. Such points should be eliminated:
[0110]
[0111] p Is a preset value, in this embodiment, it is set to 0.5, used as The soft threshold.
[0112] Among them, the third factor F wall For suppressing the signal of the blood vessel wall, the blood vessel wall may have a relatively larger second characteristic value and may also have a third characteristic value compared with the fracture surface. In this embodiment, a soft threshold parameter w with a value of 3 is used. It should be noted that F wall Is to inhibit large blood vessels, F vessels It is to inhibit capillaries.
[0113]
[0114] The last item F vessels Used to inhibit the crack points near blood vessels, namely capillaries:
[0115]
[0116] Among them, DT{vessels} 2 It uses the watershed algorithm to transform the distance of blood vessels. Distance change is very sensitive to small blood vessels, and small blood vessels can approach or even pass through the boundaries of lung lobes, so small blood vessels may cause lung lobes segmentation errors. When suppressing the blood vessel wall signal points, these points are close to the high blood vessel pixels inside the blood vessel and are not detected. Therefore, the crack points near the blood vessel need to be suppressed here. In this embodiment, v is set to 5 mm. Similarly, different values ​​can be selected according to different needs, which is not limited here.
[0117] By suppressing the local bright spots of the cracks, the signal points of the blood vessel wall, the crack points near the blood vessel wall, and the non-planar and non-curved crack points, the pixel points corresponding to F greater than the predetermined threshold are extracted as the lung lobe crack points. If F is less than the predetermined threshold, the pixel will be rejected and will not participate in the fitting of the fracture surface.
[0118] Step 304: Extracting the crack points of the lung lobes from the local bright spots of the crack, the signal points of the blood vessel wall, the crack points near the blood vessel, and the left lung area and the right lung area after the non-planar and non-curved crack points are suppressed, and the extracted Lung lobe fissure points construct the lung lobe fissure surface.
[0119] Using the method provided in this embodiment, the interference noise in the CT image is filtered out, the lung lobe crack points are extracted from the pixels after the noise is filtered, and when the lung lobe crack points are extracted, the local bright spots of the cracks and the signal points of the blood vessel wall , The crack points near the blood vessel and the interference points such as non-planar and non-curved crack points are suppressed, and the resulting lung lobe crack points are more accurate, which prevents the structure of the lung lobe fissure surface due to the existence of interference points. The out-of-lobe fissure surface is not accurate enough.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Similar technology patents

Classification and recommendation of technical efficacy words

People also interested in

Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products