IMAGE FUSION FOR RADIATION THERAPY

• US 20080025638A1
• Filed: 07/31/2006
• Published: 01/31/2008
• Est. Priority Date: 07/31/2006
• Status: Active Grant

• Current Assignee: CareStream Health Incorporated
• Original Assignee: Eastman Kodak Company
• Inventors: Chen, Shoupu, Schildkraut, Jay S.
• Granted Patent: US 7,848,592 B2
• Time in Patent Office: Days
• US Class Current: 382/284
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/35 using statistical methods

Abstract

An image fusion method for medical applications, comprising:

• • a. acquiring a first image with a planned radiation region;
  • b. acquiring a second image with actual radiation region;
  • c. determining if user defined landmarks have been placed on the first and second images, if user defined landmarks are present go to step (d), if not go to step (e);
  • d. pre-transforming the first image or second image or both images;
  • e. performing a first delineation step on the actual radiation region;
  • f. determining if the delineation is correct, if yes go to step (g), if not go to step (h);
  • g. fusing the first and second image and exit process; and
  • h. selecting multiple contour points around the actual radiation region in the second image;
  • i. performing a second delineation step on the actual radiation region and go to step (f).

Claims

1. An image fusion method for medical applications, comprising:

a. acquiring a first image with a planned radiation region;

b. acquiring a second image with actual radiation region;

c. determining if user defined landmarks have been placed on said first and second images, if user defined landmarks are present go to step (d), if not go to step (e);

d. pre-transforming said first image or second image or both images;

e. performing a first delineation step on said actual radiation region;

f. determining if said delineation is correct, if yes go to step (g), if not go to step (h);

g. fusing said first and second image and exit process; and

h. selecting multiple contour points around said actual radiation region in said second image;

i. performing a second delineation step on said actual radiation region and go to step (f).

2. A method as in claim 1 wherein the landmarks in said first image correspond to the landmarks in said second image.

3. A method as in claim 1 wherein said first delineation step comprises:

e.1. applying a level set contour finding algorithm with an adaptive weight to said second image to delineate said second region; and

e.2. forming a region contour vector of said second region.

4. A method as in claim 1 wherein:

h.1. said multiple contour points are selected around said second region as seed-target pairs; and

wherein said second delineation step comprises;

i.1. applying an intelligent scissors algorithm with adaptive banding method to said second image to delineate said second region using said see-target pairs; and

i.2. forming a region contour vector of said second region.

5. A method as in claim 1 wherein fusion comprises the steps of:

g.1. forming a mask that encloses said second region using said region contour vector with distinct values inside and outside of said mask;

g.2. defining a mask feature point;

g.3. defining a feature point for said first region in said first image; and

g.4. searching maximum similarity of said first and second images in multi-resolution, multi-dimensional spaces using information extracted from the inside or outside or both of said mask region.

6. A method as in claim 5 wherein said mask feature point is derived from mask geometry or selected landmarks of said second region.

7. A method as in claim 5 comprising the steps of:

g.4.1. decomposing said mask, said first and second images into a plurality of masks and images with different resolutions levels;

g.4.2. transforming said decomposed said first image with multi-dimensional parameters at a resolution level;

g.4.3. extracting a portion of said transformed decomposed first image using said decomposed mask and said feature point of said first image at said resolution level;

g.4.4. extracting a portion of said decomposed second image using said decomposed mask at said resolution level;

g.4.5. searching for maximum similarity of said decomposed second and first images using extracted information in g4.3 and g4.4;

g.4.6. transforming said decomposed first image at a higher resolution level using the transformation parameters of the current level that produce the maximum similarity;

g.4.7. repeating previous steps for all resolution levels and accumulating transformation parameters; and

g.4.8. transforming said first image using said cumulated transformation parameters.

8. A method as in claim 7 wherein the step of searching for maximum similarity further comprises the steps of:

g.4.5.1. computing entropy of the extracted portion of said second image with adaptive binning;

g.4.5.2. computing entropy of the extracted portion of said first image;

g.4.5.3. computing joint entropy of the extracted portions of said first and second images;

g.4.5.4. computing mutual information of said first and second images; and

g.4.5.5. saving the multi-dimensional parameters associated with the maximum mutual information.

9. An image fusion method for medical applications, comprising:

a. acquiring a first image with a planned radiation region;

b. acquiring a second image with actual radiation region;

c. automatically delineating said actual radiation region;

d. determining if said automatic delineation is correct, if yes go to step (e), if not go to step (f);

e. fusing said first and second image and exit process;

f. selecting multiple contour points around said actual radiation region in said second image;

g. semi automatically delineating said actual radiation region; and

h. determining if said semi automatic delineation is correct if yes go to step (e), if not go to step (f).

10. An image fusion method, using user defined landmarks for medical applications, comprising:

a. acquiring a first image with a planned radiation region;

b. acquiring a second image with actual radiation region;

c. pre-transforming said first image or second image or both images using user defined landmarks;

d. automatically delineating said actual radiation region;

e. determining if said automatic delineation is correct, if yes go to step (f), if not go to step (g);

f. fusing said first and second image and exit process;

g. selecting multiple contour points around said actual radiation region in said second image;

h. semi automatically delineating said actual radiation region; and

i. determining if said semi automatic delineation is correct if yes go to step (f), if not go to step (g).

11. An image fusion method for medical applications, comprising:

a. acquiring a first image with a first region;

b. acquiring a second image with a second region;

c. determining if user defined landmarks have been placed on said first and second images, if user defined landmarks are present go to step (d), if not go to step (e);

d. pre-transforming geometrically said first image or second image or both images;

e. performing a first delineation step using active contour model with adaptive weighting on said second region;

f. determining if said delineation is correct, if yes go to step (g), if not go to step (h);

g. fusing said first and second image using information extracted from inside or outside or both of said second region in said second image and information extracted accordingly in said first image and exit process;

h. selecting multiple contour points around said second region in said second image; and

i. performing a second delineation step using interactive segmentation with adaptive banding on said second region with said multiple contour points and go to step (f).