Single- and multi-modality alignment of medical images in the presence of non-rigid deformations using phase correlation

1. A system comprising:

at least one programmable processor; and

a non-transitory machine-readable medium storing instructions which, when executed by the at least one programmable processor, cause the at least one programmable processor to perform operations comprising:

receiving a first medical image of a patient taken at a first time and at a first position and a second medical image of the patient taken at a second time and at a second position, the first medical image and the second medical image being different from each other due to a deformation;

identifying a skin surface of the patient in the first medical image or the second medical image;

zeroing at least one voxel in the first medical image or the second medical image that is outside the skin surface;

comparing, after the zeroing, the first medical image and the second medical image using a phase correlation method to determine a displacement, the comparing comprising:

calculating an inverse Fourier transform of a cross-power spectrum of the first medical image and the second medical image, the inverse Fourier transform having a peak spread around neighboring voxels due to the deformation; and

calculating the displacement based at least partially on the intensities of the neighboring voxels;

determining, based on the displacement, a change to a physical location of the patient; and

correcting, based on the determined change, the second position of the patient to more closely conform to a first position of the patient.

2. The system of claim 1 , wherein the skin surface is identified in the first medical image and the second medical image.

3. The system of claim 2 , the identifying further comprising applying a marching squares algorithm to a plurality of slices of the first medical image and the second medical image.

4. The system of claim 3 , wherein the slices are transverse slices.

5. The system of claim 3 , wherein an isosurface for the marching squares algorithm is 0.5 times an average intensity of the voxels in the first medical image and the second medical image.

6. The system of claim 1 , the calculating of the displacement further comprising:

finding a maximum intensity of the inverse Fourier transform of the cross-power spectrum; and

selecting as the peak, from a plurality of voxels having intensities greater than a threshold determined based on the maximum intensity, a voxel for which a sum of voxel intensities of the neighboring voxels around the voxel is highest, wherein the neighboring voxels are in a window sized as a fraction of a number of voxels along each dimension of a common registration grid.

7. The system of claim 6 , further comprising determining the displacement of the peak by calculating a centroid of the voxel intensities of the neighboring voxels inside the window.

8. The system of claim 1 , wherein the first medical image and the second medical image are obtained using different imaging modalities.

9. A non-transitory machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising:

receiving a first medical image of a patient taken at a first time and at a first position and a second medical image of the patient taken at a second time and at a second position, the first medical image and the second medical image being different from each other due to a deformation;

identifying a skin surface of the patient in the first medical image or the second medical image;

zeroing at least one voxel in the first medical image or the second medical image that is outside the skin surface;

comparing, after the zeroing, the first medical image and the second medical image using a phase correlation method to determine a displacement, the comparing comprising:

calculating an inverse Fourier transform of a cross-power spectrum of the first medical image and the second medical image, the inverse Fourier transform having a peak spread around neighboring voxels due to the deformation; and

calculating the displacement based at least partially on the intensities of the neighboring voxels;

determining, based on the displacement, a change to a physical location of the patient; and

correcting, based on the determined change, the second position of the patient to more closely conform to a first position of the patient.

10. The machine-readable medium of claim 9 , wherein the skin surface is identified in the first medical image and the second medical image.

11. The machine-readable medium of claim 10 , the identifying further comprising applying a marching squares algorithm to a plurality of slices of the first medical image and the second medical image.

12. The machine-readable medium of claim 11 , wherein the slices are transverse slices.

13. The machine-readable medium of claim 11 , wherein an isosurface for the marching squares algorithm is 0.5 times an average intensity of the voxels in the first medical image and the second medical image.

14. The machine-readable medium of claim 9 , the calculating of the displacement further comprising:

finding a maximum intensity of the inverse Fourier transform of the cross-power spectrum; and

selecting as the peak, from a plurality of voxels having intensities greater than a threshold determined based on the maximum intensity, a voxel for which a sum of voxel intensities of the neighboring voxels around the voxel is highest, wherein the neighboring voxels are in a window sized as a fraction of a number of voxels along each dimension of a common registration grid.

15. The machine-readable medium of claim 14 , further comprising determining the displacement of the peak by calculating a centroid of the voxel intensities of the neighboring voxels inside the window.

16. The machine-readable medium of claim 9 , wherein the first medical image and the second medical image are obtained using different imaging modalities.

17. A system comprising:

at least one programmable processor; and

a non-transitory machine-readable medium storing instructions which, when executed by the at least one programmable processor, cause the at least one programmable processor to perform operations comprising:

receiving a first medical image of a patient taken at a first time and at a first position and a second medical image of the patient taken at a second time and at a second position, the first medical image and the second medical image being different from each other due to a deformation;

identifying a skin surface of the patient in the first medical image or the second medical image, wherein the skin surface is identified in the first medical image and the second medical image, the identifying further comprising applying a marching squares algorithm to a plurality of slices of the first medical image and the second medical image, wherein an isosurface for the marching squares algorithm is 0.5 times an average intensity of the voxels in the first medical image and the second medical image;

zeroing at least one voxel in the first medical image or the second medical image that is outside the skin surface;

comparing, after the zeroing, the first medical image and the second medical image using a phase correlation method to determine a displacement;

determining, based on the displacement, a change to a physical location of the patient; and

correcting, based on the determined change, the second position of the patient to more closely conform to a first position of the patient.

18. The system of claim 17 , wherein the skin surface is identified in the first medical image and the second medical image.

19. The system of claim 17 , wherein the slices are transverse slices.

20. The system of claim 17 , the comparing comprising:

calculating an inverse Fourier transform of a cross-power spectrum of the first medical image and the second medical image, the inverse Fourier transform having a peak spread around neighboring voxels due to the deformation; and

calculating the displacement based at least partially on the intensities of the neighboring voxels.