Image Registration for Image-Guided Surgery

• US 20130016889A1
• Filed: 07/14/2011
• Published: 01/17/2013
• Est. Priority Date: 07/14/2011
• Status: Active Grant

• Current Assignee: Accuray Incorporated
• Original Assignee: Accuray Incorporated
• Inventors: Jordan, Petr, Myronenko, Andriy, West, Jay Bailey
• Granted Patent: US 8,818,105 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10124 Digitally reconstructed radiograph [DRR]
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/337 involving reference images or patches

Abstract

An image registration process for detecting a change in position of a surgical target, such as a tumor, within a patient is disclosed. A pre-operative model of the target and surrounding area is generated, then registered to live patient images to determine or confirm a location of the target during the course of surgery. Image registration is based on a non-iterative image processing logic that compares a similarity measure for a template of the target (generated from the pre-operative model) at one location within the live image to other locations within the live image.

Claims

1. A method for identifying a structure within a patient image, comprising:

generating a reference image from an image of the patient;

generating a template from at least a portion of the reference image, the template comprising a set of patches;

for each patch, generating a patch similarity map by computing a local similarity measure between the patient image and the patch for the template located at each of several candidate locations for the structure in the patient image;

combining the patch similarity maps according to spatial relationships among the set of patches in the template to obtain a global similarity map; and

identifying the structure from a template location in the patient image that corresponds to a maximum value for the global similarity map.

2. The method of claim 1, wherein each patch similarity map in the combining step is weighted in proportion to the relative significance of the image information in the patch.

3. The method of claim 1, wherein the local similarity measure is a maximum similarity value for a neighborhood of the patch'"'"'s location in the patient image.

4. The method of claim 2, wherein a patch comprises pixels having values and the weighting is a standard deviation of the pixel values.

5. The method of claim 1, wherein the reference image is a Digitally Reconstructed Radiograph (DRR) and the generating a reference image step further includes generating a first DRR using all or part of the image of the patient, generating a second DRR using a smaller subset of the image than used for the first DRR;

generating a first template from all or part of the first DRR, the first template comprising a set of candidate patches;

generating a second template from all or part of the second DRR, the second template comprising a set of reference patches;

calculating a similarity measure between each candidate patch and a corresponding reference patch; and

using only those candidate patches satisfying a predetermined criteria for a similarity measure.

6. The method of claim 5, wherein the reference patches correspond to an immediate neighborhood of the structure in the image of the patient.

7. The method of claim 1, wherein the local similarity measure is normalized cross-correlation (NCC).

8. The method of claim 1, wherein each of the generating, combining and identifying steps is non-iterative.

9. The method of claim 1, further including generating a global similarity map for each of a plurality of patient images, where there is a known spatial relationship between the coordinates of at least a first image and a second image of the plurality of images,

using the known spatial relationship to combine the global similarity maps to thereby produce a combined similarity map, and

identifying the structure from a template location in the patient image that corresponds to a maximum value for the combined similarity map.

10. The method of claim 1, wherein the reference image is a Digitally Reconstructed Radiograph (DRR) generated from all or a part of a 3D image of the patient.

11. The method of claim 10, wherein the patient image is an x-ray and the DRR is generated from all or a part of a 3D image of the patient that was rendered from CT imaging data.

12. The method of claim 1, wherein the local similarity measure is a statistic of similarity values for a neighborhood of the patch'"'"'s location in the patient image.

13. The method of claim 12, wherein the statistic is a function of a mean, average or higher order statistic of the similarity values for a neighborhood of the patch'"'"'s location in the patient image.

14. The method of claim 1, further including a first image and a second image of a plurality of patient images that are spatially related such that a first coordinate location for the structure in at least the first and second images is the same,

generating the global similarity map for the first image,

extracting the first coordinate location for the structure from the global similarity map, and

locating a second coordinate for the structure in the second image including computing similarity values in the second image for only those candidate locations that include the first coordinate location.

15. The method of claim 2, wherein the patch similarity maps are weighted in proportion to the relative significance of the image information in the patch then summed together according to the spatial relationships among the set of patches.

16. The method of claim 1, wherein the structure is a treatment target and/or a tracking target.

17. A method for identifying a structure within a patient image, comprising:

generating a reference image from an image of the patient;

generating a template from all or part of the reference image, the template comprising a set of patches;

computing similarity values between the patient image and the template for the template located at each of several candidate locations for the structure in the patient image, wherein at each template location the similarity value is computed according to steps (a) and (b), as follows;

(a) for each patch, computing a local similarity value for a neighborhood of the patch'"'"'s location in the patient image, and

(b) combining the local similarity values to obtain the similarity value at the template location; and

identifying the structure in the patient image from a template location corresponding to a maximum of the similarity values.

18. The method of claim 17, further including the step of generating a global similarity map from the weighted similarity measures and then identifying the structure location from the global similarity map.

19. The method of claim 17, wherein the combining of local similarity values is a weighted sum of the local similarity values, wherein each local similarity value is weighted in proportion to the relative significance of the image information contained in the corresponding patch.

20. The method of claim 17, wherein the structure is a treatment target and/or a tracking target.

21. A computer-readable medium having stored therein computer-readable instructions for a processor, wherein the instructions, when read and implemented by the processor, cause the processor to:

generate a patch similarity map for each of a set of patches that form a template for a structure within a patient, wherein the template was generated from a reference image of the patient, and wherein the patch similarity map is generated by computing a local similarity measure between a patient image and the patch for the template located at each of several candidate locations for the structure in the patient image;

combine the patch similarity maps according to spatial relationships among the set of patches in the template to obtain a global similarity map; and

identify the structure from a template location in the patient image that corresponds to a maximum value for the global similarity map.

22. The method of claim 21, wherein the patient image is an x-ray and the reference image was generated from all or a part of a 3D image of the patient that was rendered from CT imaging data.

23. The method of claim 21, wherein the structure is a treatment target and/or a tracking target.

24. A method for verifying an expected location for a target structure in a patient image, comprising:

computing similarity measures between the patient image and patches for the structure for each of a plurality of candidate locations for the structure in the patient image;

generating a global similarity map from the computed similarity measures, wherein a maximum of the computed similarity measures is an expected location for the structure in the patient image;

displaying the global similarity map with the patient image such that a similarity measure for each of the candidate locations is displayed with the corresponding location in the patient image; and

visually comparing structure in the patient image against the similarity measures to verify that the expected location is the actual location of the target structure in the patient image.

25. The method of claim 24, wherein the computing step includes, for each patch, generating a patch-level similarity measure for each of the candidate locations and combining the patch-level similarity measures to form the global similarity map.

26. The method of claim 24, wherein the patient image is an x-ray and the global similarity map is a 2D color image distinguishing magnitudes of the similarity measures by color.