Image comparison tool tolerant to deformable image matching

Patent metadata

Field	Value
Publication number	US-10489678-B2
Application number	US-201715707555-A
Country	US
Kind code	B2
Filing date	Sep 18, 2017
Priority date	Apr 13, 2015
Publication date	Nov 26, 2019
Grant date	Nov 26, 2019

Abstract

An apparatus and method for determining an image similarity based on image features. In one aspect, the image similarity determination is based on an image comparison tool. The image comparison tool may be trained, by a machine-learning system, to estimate a similarity between images based on a subset of image data comprised by image features. The estimate may be an estimate of how similar structures found in the images would be following a geometric transformation of some of the structures. In one aspect, an atlas image for performing automatic segmentation of an image is determined according to a comparison made using the image comparison tool.

First claim

What is claimed is:

1. A method comprising:
performing, with a processor that executes instructions for training an image comparison tool a comparison of a medical image pair to generate a first similarity result, wherein said performing the comparison of the medical image pair includes deforming at least one medical image of the medical image pair;
performing a comparison of extracted features from the medical image pair with the image comparison tool to generate a second similarity result; and
tuning the image comparison tool to increase correspondence between the first similarity result and the second similarity result.

2. The method of claim 1 , further comprising:
repeating said performing the comparison of the medical image pair, said performing the second comparison of the extracted features from the medical image pair, and said tuning for a respective one of a plurality of medical image pairs.

3. The method of claim 2 , wherein the first similarity result is indicative of similarity between a first anatomical structure of a first medical image of the medical image pair and a second anatomical structure of a second medical image of the medical image pair.

4. The method of claim 3 , wherein the first anatomical structure and the second anatomical structure are related to at least one of livers or bones.

5. The method of claim 4 , wherein the first similarity result is indicative of similarity between a first combination of anatomical structures of a first medical image of the medical image pair and a second combination of anatomical structures of a second medical image of the medical image pair.

6. The method of claim 5 , wherein said deforming comprises:
transforming original image data of one medical image of the medical image pair from an original spatial reference into a spatial reference in common with original image data of other medical image of the medical image pair.

7. The method of claim 6 , wherein said performing the comparison of the medical image pair further comprises:
segmenting a first medical image of the medical image pair to generate a first segmented medical image;
segmenting a second medical image of the medical image pair to generate a second segmented medical image; and
using the first segmented medical image and the second segmented medical image to determine the first similarity result.

8. A method comprising:
performing, with a processor that executes instructions for training an image comparison tool a comparison of a medical image pair to generate a first similarity result by using non-deformed image data of a first medical image of the medical image pair and using deformed image data of a second medical image of the medical image pair that represents original image data transformed from an original spatial reference into a spatial reference in common with the nondeformed image data;
performing a comparison of extracted features from the medical image pair with the image comparison tool to generate a second similarity result; and
tuning the image comparison tool to increase correspondence between the first similarity result and the second similarity result.

9. The method of claim 8 , further comprising:
repeating said performing the comparison of the medical image pair, said performing the comparison of the extracted features from the medical image pair, and said tuning for a respective one of a plurality of medical image pairs.

10. The method of claim 8 , wherein the first similarity result is indicative of similarity between a first anatomical structure of the first medical image and a second anatomical structure of the second medical image.

11. The method of claim 10 , wherein the first anatomical structure and the second anatomical structure are related to at least one of livers or bones.

12. The method of claim 8 , wherein the first similarity result is indicative of similarity between a first combination of anatomical structures of the first medical image and a second combination of anatomical structures of the second medical image.

13. The method of claim 8 , wherein said performing the comparison of the medical image pair comprises:
segmenting the first medical image to generate a first segmented medical image;
segmenting the second medical image to generate a second segmented medical image; and
using the first segmented medical image and the second segmented medical image to determine the first similarity result.

14. A method comprising:
providing an image comparison tool that is tunable by correspondence between a first image similarity result, that involves deformation of a first medical image relative to a second medical image, and a second image similarity result that is generated with the image comparison tool;
comparing target extracted features from a target medical image to reference extracted features from a plurality of reference medical images with the image comparison tool to generate a plurality of similarity results, wherein the target extracted features and the reference extracted features are non-deformed data; and
selecting, with a processor that executes instructions for selecting an atlas, a reference medical image as the atlas for the target medical image based on the plurality of similarity results.

15. The method of claim 14 , further comprising:
using the atlas from performance of said selecting to segment the target medical image.

16. The method of claim 14 , wherein said comparing comprises:
arranging the target extracted features into a plurality of categories and the reference extracted features into the plurality of categories per reference medical image; and
per reference medical image, comparing a category of target extracted features with a corresponding category of reference extracted features for a respective one of the plurality of categories.

17. The method of claim 14 , wherein said comparing does not involve deformation of at least one medical image relative to another medical image.

18. The method of claim 14 , wherein a respective similarity result of the plurality of similarity results is indicative of similarity between the target medical image and a respective reference medical image.

19. The method of claim 14 , wherein the first image similarity result is indicative of similarity between a first anatomical structure of the first medical image and a second anatomical structure of the second medical image.

20. The method of claim 19 , wherein the first anatomical structure and the second anatomical structure are related to at least one of livers or bones.

Assignees

• Varian Medical Systems Int Ag

Inventors

• Haas Benjamin
• Coradi Thomas
• Waschbuesch Michael

Classifications

• G06V10/776 Primary

  Validation; Performance evaluation · CPC title

• G06F18/217

  Validation; Performance evaluation; Active pattern learning techniques · CPC title

• G06T7/35

  using statistical methods · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T2207/10072

  Tomographic images · CPC title