METHOD FOR DETERMINING A PROPERTY MAP OF AN OBJECT, PARTICULARLY OF A LIVING BEING, BASED ON AT LEAST A FIRST IMAGE, PARTICULARLY A MAGNETIC RESONANCE IMAGE

• US 20100049032A1
• Filed: 01/09/2009
• Published: 02/25/2010
• Est. Priority Date: 07/12/2006
• Status: Active Grant

• Current Assignee: Eberhard-Karls-Universitat Tubingen Universitatsklinikum, Max-Planck-Gesellschaft Zur Foerderung Der Wissenschaften e.V. (Max-Planck-Gesellschaft zur Frderung der Wissenschaften e)
• Original Assignee: Eberhard-Karls-Universitat Tubingen Universitatsklinikum, Max-Planck-Gesellschaft Zur Foerderung Der Wissenschaften e.V. (Max-Planck-Gesellschaft zur Frderung der Wissenschaften e)
• Inventors: Hofmann, Matthias, Steinke, Florian, Scholkopf, Bernhard, Pichler, Bernd
• Granted Patent: US 8,290,568 B2
• Time in Patent Office: Days
• US Class Current: 600/416
• CPC Class Codes: A61B 6/037 Emission tomography
• CPC Class Codes: G01R 33/481 MR combined with positron emission tomography [PET] or single photon emission computed tomography [SPECT]
• CPC Class Codes: G06T 2207/10088 Magnetic resonance imaging [MRI]
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 2207/30196 Human being; Person
• CPC Class Codes: G06T 7/33 using feature-based methods

Abstract

It is disclosed a system and method (12) for determining a property map (82) of an object, particularly a human being, based on at least a first image (84), particularly an magnetic resonance (MR) image, of the object. In the method (12), a structure of reference pairs is defined in a first step (96), wherein each reference pair (16-26) comprises at least two entries (62). The first entry represents a property value, particularly an attenuation value. The second entry (62) preferably represents a group of image points (67) belonging together, which is extracted particularly from MR images (28) and comprises an interesting image point corresponding to the property value. In another step (98) of the method (12) a plurality of training pairs (16-26) is provided. A structure of the training pairs (16-26) corresponds to the structure of reference pairs, and the entries of respective training pairs (16-26) are known. In another step (100) of the method (12), an assignment between the first entries and the other entries (62-66) of the training pairs (16-26) is determined by machine learning, thus allowing prediction of a property value (88) corresponding to an arbitrary point (90) of the first image (84).

Claims

1. A method for determining a property map of an object based on at least a first image of the object, comprising:

defining a structure of reference pairs, wherein each reference pair has at least two entries, wherein the first entry represents a property value, and wherein the second entry preferably represents a group of image points belonging together, which are particularly extracted from MR images, wherein the group includes at least one interesting image point which corresponds to the property values;

providing a plurality of training pairs, wherein a structure of the training pairs corresponds to the structure of reference pairs, and wherein the entries of the respective training pairs are known;

determining an assignment between the first entries and the other entries of the training pairs by machine learning for predicting in this manner a corresponding property value for an arbitrary image point of the first image.

2. The method of claim 1, wherein the first image is a magnetic resonance (MR) image, and wherein the first entry represents an attenuation value.

3. The method of claim 1, wherein the structure of reference pairs additionally comprises third or additional entries, which are taken into account during determination of the assignment to the first entry, together with a second entry.

4. The method of claim 3, wherein the third or other entries (64) is selected from a group consisting of:

a registered coordinate;

another group of image points belonging together of at least a second image, wherein the other group of image points of the at least second image shows at least partially an identical object patch like the group of image points of the first image; and

a class affiliation.

5. The method of claim 1, wherein the second entry represents a registered coordinate.

6. The method of claim 4, wherein the registered coordinate is defined by one of a coordinate of a point equivalent to the interesting point in a reference object, and one or more distances of the point being equivalent to the interesting point relative to the anatomically characteristic positions within the reference object.

7. The method of claim 4, wherein a similarity between the registered coordinate and a position-dependent distance measure is determined, which is dependent on properties of the reference object as well as a position of the equivalent point within the reference object.

8. The method of claim 4, wherein the at least second image is one of a positron emission tomography (PET) image and another MR image.

9. The method of claim 4, wherein the class affiliation is selected form a group consisting of:

a patient'"'"'s age, a patient'"'"'s sex, a patient'"'"'s weight, and patient'"'"'s habit.

10. The method the claim 1, wherein the group of image points belonging together defines a window.

11. The method of claim 10, wherein the window is rectangular.

12. The method of claim 1, wherein a subgroup is selected from the group of image points belonging together, which comprises the interesting image point.

13. The method of claim 1, wherein each image point has an intensity value.

14. The method of claim 1, wherein a texture is determined for each group of image points belonging together.

15. The method of claims 9 and 13, wherein an intensity histogram is established based on the subgroup, which can be added to the structure as another entry.

16. The method of claim 1, wherein an intensity average is formed based on the intensity values of at least the image points of the group of image points belonging together, the intensity average being added to the structure together with the intensity of the interesting image points as another entry.

17. The method of claim 1, wherein at least one of the at least first, the MR images, and the at least second image are pre-processed by the extraction of features.

18. The method of claim 17, wherein the pre-processing is performed a Fourier/Wavelet transformation, SIFT, edge detectors, or contrast amplification.

19. The method of claim 1, wherein a structure is provided for the arbitrary point of the object, which corresponds to the structure of the reference pairs, and wherein the structure of the arbitrary point comprises at least one entry which is known and different to the first entry, and wherein the first entry is missing.

20. The method of claim 18, wherein the prediction of the first entry of the arbitrary point is performed by applying the learnt assignment based on at least one or more of the known entries of the arbitrary point.

21. The method of claim 20, wherein the prediction, if any second entry is known for the arbitrary point is performed based on one or more known additional entries.

22. The method of claim 1, wherein additionally at least the second image is used for registering with the at least first image, and for extending the at least first image.

23. A system for determining a property map of an object based on at least a first image of the object, comprising:

means for defining a structure of reference pairs, wherein each reference pair comprises at least two entries, wherein the first entry represents a property value, and wherein the second entry represents a group of image points belonging together, which comprises an interesting point corresponding to the property value;

means for reading a plurality of training pairs; and

a learning machine for determining an assignment between the first entries and the other entries of the training pairs, allowing prediction of a property value corresponding to an arbitrary image point of the first image.