AUTOMATED VASCULAR REGION SEPARATION IN MEDICAL IMAGING

• US 20130202170A1
• Filed: 06/30/2010
• Published: 08/08/2013
• Est. Priority Date: 01/29/2010
• Status: Active Grant

• Current Assignee: International Business Machines Corporation, Mayo Foundation For Medical Education And Research (Mayo Clinic)
• Original Assignee: International Business Machines Corporation, Mayo Foundation For Medical Education And Research (Mayo Clinic)
• Inventors: Blezek, Daniel J., Erickson, Bradley J., Schommer, Peter J., Ramirez, Nelson, Limpert, David R.
• Granted Patent: US 9,042,611 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20128 Atlas-based segmentation
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/162 involving graph-based methods
• CPC Class Codes: G06V 20/64 Three-dimensional objects
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing

Abstract

A system and/or method automatically identifies one or more vascular regions in a medical image or set of medical images. For example, the system/method may automatically identify vascular structures as belonging to the left carotid, right carotid, and/or basilar vascular regions in the head. The system/method takes as input the medical image(s) and automatically identifies one or more vascular regions. The system/method may also automatically generate MIP renderings of the identified region or regions.

Claims

1. A method comprising:

determining a probability for each voxel in a patient-specific image data set that the voxel belongs to one or more vascular regions of interest;

segmenting patient-specific vasculature in the patient-specific image to generate a set of nodes and edges representative of the patient-specific vasculature;

classifying each node and edge based on one or more statistics associated with each node and edge; and

determining to which of the one or more vascular regions of interest each voxel in the patient-specific image data set belongs based on the probability and the classifications.

2. The method of claim 1 wherein determining to which of the one or more vascular regions of interest each voxel belongs comprises determining whether each voxel in the patient-specific image data set belongs to a left carotid vascular region, a right carotid vascular region or a basilar vascular region.

3. The method of claim 1 further comprising automatically generating a MIP rendering of the vascular region of interest based on the association.

4. The method of claim 3 further comprising displaying the MIP rendering on a user interface.

5. The method of claim 1 wherein determining a probability for each voxel in a patient-specific image data set that the voxel belongs to one or more vascular regions of interest comprises assigning each voxel in the patient-specific image data set a patient-specific probability based on a corresponding voxel in each of one or more probabilistic atlases, each probabilistic atlas corresponding to a different one of the vascular regions.

6. The method of claim 1 wherein determining to which of the one or more vascular regions of interest each voxel belongs comprises:

if the patient-specific probability for a voxel in the patient-specific image data set indicates that the voxel belongs to a first one of the vascular regions and the classification indicates that the voxel belongs to the first vascular region, determining that the voxel belongs to the first vascular region;

if the patient-specific probability for a voxel in the patient-specific image data set indicates that the voxel does not belong to the first one vascular region and the classification indicates that the voxel belongs to the first vascular region, determining that the voxel belongs to the first vascular region;

if the patient-specific probability for a voxel in the patient-specific image data set indicates that the voxel belongs to the first vascular region and the classification indicates that the voxel does not belong to the first vascular region, determining that the voxel does not belong to the first vascular region; and

if the patient-specific probability for a voxel in the patient-specific image data set indicates that the voxel does not belong to the first vascular region and the classification indicates that the voxel does not belong to the first vascular region, determining that the voxel does not belong to the first vascular region.

7. The method of claim 1 wherein classifying each node and edge based on one or more statistics comprises classifying each node and edge based on one or more of the average probabilities that the edge or node is in the left carotid, right carotid, or basilar arterial region;

a centroid of the vessel segment;

a minimum radius of the vessel segment;

a maximum radius of the vessel segment;

an average radius of the vessel segment;

an average cross sectional area of the vessel segment;

a direction of the vessel segment;

a length of the vessel segment;

a curvature of the vessel segment;

whether a node is a terminal node; and

a 3-dimensional location within the patient'"'"'s head of the vessel segment.

8. A method comprising:

receiving a plurality of manually generated maximum intensity projection (MIP) renderings of a vascular region, each corresponding to one of a plurality of image data sets obtained from subjects in a sample population;

generating an anatomic atlas comprised of an average image data set based on the plurality of image data sets obtained from the subjects in the sample population;

projecting each voxel in the anatomic atlas into each of the plurality of manually generated MIP renderings;

determining, for each voxel in the anatomic atlas, a probability that the voxel belongs to the vascular region based on the projection; and

storing the determined probabilities as a probabilistic atlas for the vascular region indicative of a likelihood that each voxel in a patient-specific image data set belongs to vascular region.

9. A system comprising:

a controller that receives a patient-specific image data set;

a probability module, executed by the controller, that determines a probability for each voxel in a patient-specific image data set that the voxel belongs to one or more vascular regions of interest;

a segmentation module, executed by the controller, that segments patient-specific vasculature in the patient-specific image to generate a set of nodes and edges representative of the patient-specific vasculature;

a classification module, executed by the controller, that classifies each node and edge based on one or more statistics associated with each node and edge; and

a vessel location analysis module, executed by the controller, that determines to which of the one or more vascular regions of interest each voxel in the patient-specific image data set belongs based on the probability and the classifications.

10. The system of claim 9 wherein the vascular region of interest comprises one or more of a left carotid vascular region, a right carotid vascular region, or a basilar vascular region.

11. The system of claim 9 further including a manual intensity projection (MIP) module, executed by the controller, that automatically generates a MIP rendering of the vascular region of interest based on the association.

12. The system of claim 9 further including a user interface that displays the MIP rendering.

13. The system of claim 9 wherein the probability module assigns each voxel in the patient-specific image data set a patient-specific probability based on a corresponding voxel in each of one or more probabilistic atlases, each probabilistic atlas corresponding to a different one of the vascular regions.