Automatic segmentation of image frames in anatomical scan based on non-rigid registration with manually segmented frame

Patent metadata

Field	Value
Publication number	US-9824456-B2
Application number	US-201514752105-A
Country	US
Kind code	B2
Filing date	Jun 26, 2015
Priority date	Jun 26, 2014
Publication date	Nov 21, 2017
Grant date	Nov 21, 2017

Abstract

A time series of image frames in an anatomical scan is segmented. An image frame k in the time series of image frames is manually segmented. A non-rigid registration is then performed between the image frame k and a next image frame k+1 in the time series of image frames. A segmentation on the image frame k+1 is computed based on the non-rigid registration. Each subsequent image frame k+n in the time series of image frames is iteratively segmented using non-rigid registration with the segmented previous image frame k+(n−1) in the time series of image frames.

First claim

We claim:

1. A system for automatically segmenting a plurality of image frames in an anatomical scan, the system comprising:
a storage device storing the image frames, wherein the image frames comprise a series of time-separated image frames taken during the anatomical scan;
a user interface device allowing a user to manually segment a first image frame of the image frames in order to form a manually segmented first image frame, the manually segmented first image frame including a first contour manually created by the user for delineating an anatomical structure in the first image frame;
an image processor coupled to the storage device and configured to perform a non-rigid registration between the first image frame and a second image frame of the image frames, wherein the second image frame is adjacent the first image frame in the series and the non-rigid registration involves determining a point correspondence between pixel locations in the first image frame and pixel locations in the second image frame;
the image processor further configured to automatically segment the second image frame based on the non-rigid registration in order to form an automatically segmented second image frame, the automatically segmented second image frame including a second contour delineating the anatomical structure in the second image frame, wherein the second contour is automatically created by the image processor transforming prior pixel locations of the first contour on the first image frame to new pixel locations for the second contour on the second image frame according to the non-rigid registration between the first image frame and the second image frame; and
the image processor further configured to iteratively segment a plurality of subsequent image frames in the series of time-separated image frames by transforming a previously automatically generated contour for an already segmented image frame to form a newly automatically generated contour for an adjacent image frame according to a non-rigid registration performed by the image processor between the already segmented image frame and the adjacent image frame;
whereby the first image frame is manually segmented by the user to delineate the anatomical structure, and the second image frame and the plurality of subsequent image frames in the series of time-separated image frames are thereafter automatically segmented by the image processor to delineate the anatomical structure.

2. The system of claim 1 , wherein the image processor is configured to generate a moving mesh representative of a deformation field to perform non-rigid registration between image frames.

3. The system of claim 1 , wherein:
the user interface device further allows the user to manually update a segmentation in a particular one of the plurality of subsequent image frames in order to form a second manually segmented image frame; and
the image processor is further configured to automatically update an automatic segmentation in an image frame adjacent to the particular image frame in the series by transforming a manual contour of the second manually segmented image frame to form an updated automatically generated contour for the image frame adjacent to the particular image frame according to a non-rigid registration performed by the image processor between the particular image frame and the image frame adjacent to the particular image frame;
whereby, when the user decides to change an initially performed automatic segmentation of the particular image frame by the image processor, the user can manually edit the segmentation of the particular image frame and the image processor automatically corrects the automatic segmentation in one or more image frames adjacent to the particular image frame.

4. The system of claim 3 , wherein the image processor is configured to automatically correct the automatic segmentation of a center image frame based on a weighted average of updated segmentations of an immediately subsequent image frame and an immediately previous image frame in the series.

5. The system of claim 1 , wherein the image processor is configured to perform non-rigid registration between a plurality of the image frames in the series in parallel.

6. The system of claim 5 , wherein the image processor includes a graphics processing unit (GPU).

7. A method for automatically segmenting a plurality of image frames in an anatomical scan, the method comprising:
receiving the image frames by an image processor, wherein the image frames comprise a series of time-separated image frames taken during the anatomical scan;
allowing a user to manually segment a first image frame of the image frames utilizing a user interface device in order to form a manually segmented first image frame, the manually segmented first image frame including a first contour manually created by the user for delineating an anatomical structure in the first image frame;
performing a non-rigid registration between the first image frame and a second image frame of the image frames by the image processor, wherein the second image frame is adjacent the first image frame in the series and the non-rigid registration involves determining a point correspondence between pixel locations in the first image frame and pixel locations in the second image frame;
automatically segmenting the second image frame based on the non-rigid registration by the image processor in order to form an automatically segmented second image frame, the automatically segmented second image frame including a second contour delineating the anatomical structure in the second image frame, wherein the second contour is automatically created by the image processor transforming prior pixel locations of the first contour on the first image frame to new pixel locations for the second contour on the second image frame according to the non-rigid registration between the first image frame and the second image frame; and
iteratively segmenting a plurality of subsequent image frames in the series of time-separated image frames by the image processor transforming a previously automatically generated contour for an already segmented image frame to form a newly automatically generated contour for an adjacent image frame according to a non-rigid registration performed by the image processor between the already segmented image frame and the adjacent image frame;
whereby the first image frame is manually segmented by the user to delineate the anatomical structure, and the second image frame and the plurality of subsequent image frames in the series of time-separated image frames are thereafter automatically segmented by the image processor to delineate the anatomical structure.

8. The method of claim 7 , wherein performing non-rigid registration comprises:
generating a moving mesh representative of a deformation field to perform non-rigid registration between image frames.

9. The method of claim 7 , further comprising:
allowing the user to manually update a segmentation in a particular one of the plurality of subsequent image frames in order to form a second manually segmented image frame; and
automatically updating by the image processor an automatic segmentation in an image frame adjacent to the particular image frame in the series by transforming a manual contour of the second manually segmented image frame to form an updated automatically generated contour for the image frame adjacent to the particular image frame according to a non-rigid registration performed by the image processor between the particular image frame and the image frame adjacent to the particular image frame;
whereby, when the user decides to change an initially performed automatic segmentation of the particular image frame by the image proce

Assignees

• Univ Alberta

Inventors

• Noga Michelle
• Punithakumar Kumaradevan

Classifications

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06V10/754

  involving a deformation of the sample pattern or of the reference pattern; Elastic matching · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title

• G06T7/12

  Edge-based segmentation · CPC title

• G06T7/174 Primary

  involving the use of two or more images · CPC title