Method and system for validating image registration

1. A method for validating non-rigid image registration, comprising:

acquiring a source image and a target image;

performing a first registration from source image to target image using a non-rigid registration technique to produce a forward transformation map;

performing a second registration from the target image back to the source image using the non-rigid registration technique to produce a backward transformation map;

measuring consistency registration error as an indication of a change in local volume of the source with respect to the target image using the produced forward transformation map and the produced backward transformation map; and

validating the non-rigid registration technique based on the measured consistency registration error,

wherein the measured consistency registration error is calculated for the source image and the target image, within a time series of images, to produce a sequence of values and a spectrum of the sequence of values are analyzed, in the frequency domain, to determine whether there is a peak frequency.

2. The method of claim 1 , wherein the registration error is measured using a Jacobian matrix of the forward transformation map or the backward transformation map.

3. The method of claim 1 , wherein the consistency registration error is measured as VCRE=log|D(φ∘ψ)|+log|Dψ| or VCRE=log|D(ψ∘φ)|+log|Dφ| wherein φ represents the forward transformation map, ψ represents the backward transformation map, D(x) represents the Jacobian matrix of x, and |D(x)| represents the determinant of the Jacobian matrix of x.

4. The method of claim 1 , wherein the source and target image are part of a time series of medical images.

5. The method of claim 4 , wherein the time series of medical images is a pulmonary time series.

6. The method of claim 4 , wherein the time series of medical images is a cardiac time series.

7. The method of claim 1 , wherein the measured consistency registration error is calculated for a plurality of image pairs within the source and target image and a single value ε is determined from the plurality of calculated consistency registration errors.

8. The method of claim 7 , wherein the measured consistency registration error is calculated for a plurality of source images and the target image within a time series of images to produce a sequence of values ε.

9. The method of claim 1 , wherein validating the non-rigid registration technique based on the measured consistency registration error includes determining whether the non-rigid registration technique is sensitive to periodic motion.

10. A method for validating non-rigid image registration, comprising:

acquiring a source image and a target image;

performing registration from source image to target image using a non-rigid registration technique to produce a forward transformation map;

performing registration from the target image back to the source image using the non-rigid registration technique to produce a backward transformation map;

measuring consistency registration error as an indication of a change in local volume of the source with respect to the target image using the produced forward transformation map and the produced backward transformation map; and

validating the non-rigid registration technique based on the measured consistency registration error,

wherein the measured consistency registration error is calculated for a plurality of image pairs within the source and target image and a single value ε is determined from the plurality of calculated consistency registration errors,

wherein the measured consistency registration error is calculated for a plurality of source images and the target image within a time series of images to produce a sequence of values ε, and

wherein the sequence of values ε is transformed into the frequency domain and the resulting spectrum is analyzed to determine whether there is a predominant peak frequency.

11. The method of claim 10 , wherein analysis of the spectrum to determine whether there is a predominant peak frequency is performed by calculating a peak-to-base frequency ratio.

12. The method of claim 11 , wherein the peak-to-base frequency ratio is considered in validating the non-rigid registration technique.

13. The method of claim 12 , wherein a relatively large peak-to-base frequency ratio is used to determine that the non-rigid registration technique can perform acceptable motion compensation while a relatively small peak-to-base frequency ratio is used to determine that the non-rigid registration technique has failed to perform acceptable motion compensation.

14. The method of claim 10 , wherein the determination as to whether there is a predominant peak frequency is considered in validating the non-rigid registration technique.

15. A method for validating non-rigid image registration, comprising:

acquiring a sequence of images;

identifying a source image and a target image within the sequence of images;

performing registration from the source image to the target image using a non-rigid registration technique to produce forward transformation map;

performing registration from the target image back to the source image using the non-rigid registration technique to produce a backward transformation map;

measuring point-wise consistency registration error for a plurality of image pairs using the produced forward transformation map and the produced backward transformation map;

representing the measured point-wise consistency registration error for the plurality of image pairs using a single value ε;

determining a value ε for a plurality of source images and the target image within the sequence of images to generate a series of ε values;

transforming the series of ε values into the frequency domain to produce a spectrum;

determining whether there is a predominant peak frequency or how predominant a peak frequency is within the spectrum; and

validating the non-rigid registration technique based on whether there is a predominant peak frequency or how predominant the peak frequency is within the spectrum.

16. The method of claim 15 , wherein the consistency registration error is measured using a Jacobian matrix of the forward transformation map or the backward transformation map.

17. The method of claim 15 , wherein the registration error is measured as VCRE=log|D(φ∘ψ)|+log|Dψ| or VCRE=log|D(ψ∘φ)|+log|Dφ| wherein φ represents the forward transformation map, ψ represents the backward transformation map, D(x) represents the Jacobian matrix of x, and |D(x)| represents the determinant of the Jacobian matrix of x.

18. The method of claim 15 , wherein determining whether there is a predominant peak frequency is performed by calculating a peak-to-base frequency ratio for the spectrum.

19. The method of claim 18 , wherein a relatively large peak-to-base frequency ratio is used to determine that the non-rigid registration technique has failed to adequately perform motion compensation while a relatively small peak-to-base frequency ratio is used to determine that the non-rigid registration technique has achieved a good motion compensation.

20. A computer system comprising:

a processor; and

a non-transitory, tangible, program storage medium, readable by the computer system, embodying a program of instructions executable by the processor to perform method steps for validating non-rigid image registration, the method comprising:

acquiring a source image and a target image;

performing a first registration from source image to target image using a non-rigid registration technique to produce forward transformation map;

performing a second registration from the target image back to the source image using the non-rigid registration technique to produce a backward transformation map;

measuring consistency registration error using the produced forward transformation map and the produced backward transformation map;

determining whether the non-rigid registration technique is sensitive to periodic motion using the measured consistency registration error; and

validating the non-rigid registration technique based on whether the non-rigid registration technique is sensitive to periodic motion,

wherein the measured consistency registration error is calculated for the source image and the target image, within a time series of images, to produce a sequence of values and a spectrum of the sequence of values are analyzed, in the frequency domain, to determine whether there is a peak frequency.