Methods for the Segmentation of Lungs, Lung Vasculature and Lung Lobes from CT Data and Clinical Applications
AssigneeVeran Medical Technologies, Inc.
InventorsAndrew Dougherty, Mark Hunter
Life of this patent
Estimating the term…
Abstract
Method and processes for segmentation of lungs lobes from CT image data are disclosed. Lobe segmentation relies on other segmentations, including the lungs, the lung airways and vasculature. Segmentation methods and processes begin with the acquisition of 3-D image data such as from a high resolution CT scan of a patient's lungs or at least a region thereof. The image is then processed to provide a segmentation mask from which the components are extracted to identify the lungs and airway. Lung vasculature and lobes are identified by similar processes.
Claims16
A method for providing an anatomical segmentation of a displayed patient anatomy for visualization by a user, comprising the following steps:
- loading a 3D image of at least a region of a patient's lungs with specified reference coordinates system (RCS) coordinates into a memory component of an image analysis system;
- applying histogram thresholding to the 3D image to segment air from background in the 3D image to provide a computed threshold;
- thresholding voxels of a specified range into a binary mask based on the computed threshold, wherein connected components are extracted from the binary mask, resulting in a histogram threshold segmentation comprised of a plurality of components;
- determining which of the plurality of components correspond to the lungs by morphologically closing and then eroding the histogram threshold segmentation to provide a closed and eroded histogram threshold segmentation comprised of lung seed voxels that are best left or best right of the connected components,
- identifying a plurality of voxels nearest individual components in the closed and eroded histogram threshold segmentation, lung seed voxels that are best left or best right of the connected components and also in the histogram threshold segmentation being used to region grow all other voxels that are 6-connected to the lung seed voxels in the histogram threshold segmentation, resulting in a binary 3D image;
- extracting contours for individual successive slices of the 3D binary image in a decreasing axial direction;
- computing centroids for a subset of the contours that lie a predetermined percentage of a way along a sagittal extent of the binary 3D image and that have an excepted area;
- determining a trachea seed point by comparing the centroids on a predetermined number of the individual successive slices that are within a predetermined distance of one another, and having areas within a predetermined acceptable tolerance of one another, the trachea seed point being the centroid for a middle of the successive slices;
- and
- extracting, from the plurality of components, a first largest component and a second largest component by removing individual voxels in the first extracted component that are 6-connected to those of the second extracted component, the first extracted component representing a first lung and the second extracted component representing a second lung.
A method for providing an anatomical segmentation of a displayed patient anatomy for visualization by a user, comprising the following steps:
- receiving a 3D binary image that has been generated based on a non-binary 3D image of at least a region of a patient's lungs with specified reference coordinates system (RCS);
- extracting contours for individual successive slices of the 3D binary image in a decreasing axial direction;
- computing centroids for a subset of the contours that lie a predetermined percentage of a way along a sagittal extent of the binary 3D image and that have an excepted area;
- determining a trachea seed point by comparing the centroids on a predetermined number of the individual successive slices that are within a predetermined distance of one another, and having areas within a predetermined acceptable tolerance of one another, the trachea seed point being the centroid for a middle of the successive slices;
- and
- extracting, from the plurality of components, a first largest component and a second largest component by removing individual voxels in the first extracted component that are 6-connected to those of the second extracted component, the first extracted component representing a first lung and the second extracted component representing a second lung.
Dependencies are inferred from the claim text.