Method for extracting airways and pulmonary lobes and apparatus therefor

Patent metadata

Field	Value
Publication number	US-9996919-B2
Application number	US-201414909304-A
Country	US
Kind code	B2
Filing date	Jun 5, 2014
Priority date	Aug 1, 2013
Publication date	Jun 12, 2018
Grant date	Jun 12, 2018

Abstract

Provided is a method and apparatus for segmenting airways and pulmonary lobes. An image processing apparatus obtains a first candidate region of an airway from a three-dimensional (3D) human body image by using a region growing method, obtains a second candidate region of the airway based on a directionality of a change in signal intensity of voxels belonging to a lung region segmented from the 3D human body image, segments an airway region by removing noise based on similarity of a directionality of a change in signal intensity of voxels belonging to a third candidate region acquired by combining together the first and second candidate regions. Furthermore, the image processing apparatus segments a lung region from a 3D human body image by using a region growing method, obtains a fissure candidate group between pulmonary lobes based on a directionality of a change in signal intensity of voxels belonging to the lung region, reconstructs an image of the lung region including the fissure candidate group into an image viewed from a front side of a human body and generates a virtual fissure based on a fissure candidate group shown in the reconstructed image, and segments the pulmonary lobes by using the virtual fissure.

First claim

The invention claimed is:

1. A method of segmenting a human tissue, the method comprising:
acquiring a three-dimensional (3D) human body image;
obtaining a first candidate region of a human tissue from the 3D human body image by using a region growing method;
obtaining a second candidate region of the human tissue based on a directionality of a change in signal intensities of voxels;
obtaining a third candidate region of the human tissue by using the region growing method to form a single image obtained by combining together the first and second candidate regions;
segmenting a human tissue region by removing noise based on similarity of a directionality of a change in signal intensity of voxels belonging to the third candidate region; and
displaying the human tissue region in a output device;
wherein the acquiring of the 3D human body image comprises obtaining the 3D human body image using computed tomography (CT) scanning or magnetic resonance imaging (MRI).

2. The method of claim 1 , wherein the obtaining of the first candidate region comprises obtaining the first candidate region by using the region growing method that grows a region based on an initial upper limit value of signal intensity of a voxel belonging to the human tissue in the 3D human body image.

3. The method of claim 2 , wherein the obtaining of the first candidate region comprises:
decreasing, when a volume of a candidate region obtained by using the region growing method is less than or equal to a reference volume, the initial upper limit value and then obtaining a new candidate region by using again the region growing method based on the decreased upper limit value;
repeating obtaining of a new candidate region by decreasing the initial upper limit value until a volume of the new candidate region is greater than or equal to the reference volume; and
determining, when the volume of the new candidate region is greater than or equal to the reference volume, a previously obtained candidate region as the first candidate region.

4. The method of claim 2 , wherein the obtaining of the first candidate region comprises:
decreasing the initial upper limit value;
calculating a ratio between the number of voxels belonging to a candidate region newly obtained by using the region growing method based on the decreased upper limit value and the number of voxels belonging to a previously obtained candidate region; and
determining, if the ratio exceeds a preset threshold value, a previously obtained candidate region as the first candidate region.

5. The method of claim 1 , wherein the obtaining of the first candidate region further comprises performing a morphological closing operation on the first candidate region by using a structuring element.

6. The method of claim 1 , further comprising, before the obtaining of the first candidate region, performing anisotropic diffusion filtering.

7. The method of claim 1 , wherein the obtaining of the second candidate region comprises:
segmenting the human tissue from the 3D human body image by using the region growing method;
calculating a first eigenvector and a first eigenvalue for a change in signal intensity of voxels belonging to the human tissue in each direction; and
obtaining the second candidate region by comparing the first eigenvector and the first eigenvalue of each of the voxels with a reference eigenvector and a referenced eigenvalue preset for a change in signal intensity of a human tissue structure.

8. The method of claim 7 , wherein, in the obtaining of the second candidate region, eigenvalues λ1, λ2, and λ3 corresponding to three directions of each of the voxels satisfy a condition in which λ1 is less than the reference eigenvalue and λ2 and λ3 are greater than the reference eigenvalue.

9. The method of claim 1 , wherein the removing of the noise comprises calculating similarity between a vector representing a change in signal intensity of a voxel in the third candidate region and an average vector for a change in signal intensity in a neighboring space including the voxel and considering the voxel having similarity less than or equal to a preset value as noise and removing the voxel.

10. The method of claim 9 , wherein the removing of the noise comprises:
calculating first similarity between a vector of a first voxel in a difference (−) image between the second and first candidate regions and an average vector for a neighboring region of a voxel in the first candidate region that is at the same position as the first voxel and obtaining a fourth candidate region by considering the first voxel having the first similarity less than or equal to a specific value as noise and removing the first voxel;
calculating second similarity between a vector of a second voxel in a difference (−) image between the third and first candidate regions and an average vector for a neighboring region of a voxel that is at the same position in the difference (−) image between the second and first candidate regions as the second voxel and obtaining a fifth candidate region by considering the second voxel having the second similarity less than or equal to a specific value as noise and removing the second voxel;
calculating third similarity between a vector of a third voxel in an intersection region composed of voxels belonging to the third candidate region, voxels not belonging to the fourth candidate region, and voxels not belonging to the fifth candidate region and an average vector for a neighboring region of a voxel that is at the same position in the difference (−) image between the fifth and first candidate regions as the third voxel and segmenting the airway region by restoring voxels having the third similarity greater than or equal to a specific value.

11. The method of claim 10 , wherein the segmenting of the human tissue region further comprises removing an island having less than a preset number of voxels among islands obtained by performing 3D labeling on the segmented human tissue and connecting and restoring an island having a number of voxels that are greater than or equal to the preset number of voxels based on the third candidate region.

Assignees

• Seoul Nat Univ R&Db Foundation

Inventors

• Park Sang Joon
• Goo Jin Mo
• Chung Jin Wook
• Lee Doo Hee
• Kim Yang Wook

Classifications

• A61B6/50

  specially adapted for specific body parts; specially adapted for specific clinical applications · CPC title

• A61B6/5211

  involving processing of medical diagnostic data · CPC title

• G06T7/155

  involving morphological operators · CPC title

• G06T2207/10081

  Computed x-ray tomography [CT] · CPC title

• A61B6/466

  adapted to display 3D data · CPC title