Method and system for bone suppression based on a single x-ray image

1. A method for suppressing bone structures in an x-ray image, comprising:

receiving an x-ray image having a plurality of pixels;

extracting a set of wavelet features for each pixel in the x-ray image; and

generating a soft-tissue image without bone structures from the x-ray image by using a trained regression function to determine an intensity value for the soft-tissue image corresponding to each pixel in the x-ray image based on the set of wavelet features extracted for each pixel in the x-ray image.

2. The method of claim 1 , further comprising:

normalizing the x-ray image prior to extracting the set of features for each pixel in the x-ray image.

3. The method of claim 2 , wherein said step of normalizing the x-ray image comprises:

normalizing the x-ray image using multiple iterations of Gaussian blurring.

4. The method of claim 1 , wherein said step of extracting a set of wavelet features for each pixel in the x-ray image comprises:

calculating a vector of wavelet values for each pixel in the x-ray image using multiple wavelet functions corresponding to multiple scales of the x-ray image.

5. The method claim 1 , further comprising:

reducing a dimensionality of the set of wavelet features for each pixel of the x-ray image prior to generating the soft-tissue image.

6. The method of claim 5 , wherein said step of reducing a dimensionality of the set of wavelet features for each pixel of the x-ray image comprises:

linearly mapping the set of wavelet features for each pixel to a lower dimensional space.

7. The method of claim 1 , wherein said trained regression function is trained using Bayesian Committee Machine (BCM) to approximate Gaussian Process Regression (GPR).

8. The method of claim 1 , wherein said trained regression function is trained using one of k-Nearest-Neighbor regression (kNNR), support vector regression (SVR), and Gaussian process regression (GPR).

9. A method for suppressing bone structures in an x-ray image, comprising:

receiving a plurality of sets of training images, each set of training images comprising a training x-ray image and a corresponding training soft-tissue image;

extracting a set of wavelet features for each pixel in each of the training x-ray images; and

training a regression function to suppress bone structures in x-ray images based on the extracted wavelet features for the training x-ray images and the corresponding training soft-tissue images.

10. The method of claim 9 , further comprising:

normalizing the training x-ray images and corresponding training soft-tissue images prior to extracting the set of wavelet features for each pixel in each of the training x-ray images.

11. The method of claim 9 , further comprising:

reducing a dimensionality of the set of wavelet features for each pixel of each of the training x-ray images prior to training the regression function.

12. The method of claim 9 , wherein said step of training a regression function to suppress bone structures in x-ray images comprises:

training the regression function to predict intensity values of pixels of the training soft-tissue images from the sets of wavelet features extracted for corresponding pixels of the training x-ray images.

13. The method of claim 9 , wherein said step of training a regression function to suppress bone structures in x-ray images comprises:

training the regression function using Bayesian Committee Machine (BCM) to approximate Gaussian process regression (GPR).

14. The method of claim 9 , further comprising:

receiving an input x-ray image without a corresponding soft-tissue image; and

generating a soft-tissue image from the input x-ray image using the trained regression function.

15. An apparatus for suppressing bone structures in an x-ray image, comprising:

means for receiving an x-ray image having a plurality of pixels;

means for extracting a set of wavelet features for each pixel in the x-ray image; and

means for generating a soft-tissue image without bone structures from the x-ray image by using a trained regression function to determine an intensity value for the soft-tissue image corresponding to each pixel in the x-ray image based on the set of wavelet features extracted for each pixel in the x-ray image.

16. The apparatus of claim 15 , further comprising:

means for normalizing the x-ray image prior to extracting the set of features for each pixel in the x-ray image.

17. The apparatus claim 15 , further comprising:

means for reducing a dimensionality of the set of wavelet features for each pixel of the x-ray image prior to generating the soft-tissue image.

18. The apparatus of claim 15 , wherein said trained regression function is trained using Bayesian Committee Machine (BCM) to approximate Gaussian Process Regression (GPR).

19. An apparatus for suppressing bone structures in an x-ray image, comprising:

means for receiving a plurality of sets of training images, each set of training images comprising a training x-ray image and a corresponding training soft-tissue image;

means for extracting a set of wavelet features for each pixel in each of the training x-ray images; and

means for training a regression function to suppress bone structures in x-ray images based on the extracted wavelet features for the training x-ray images and the corresponding training soft-tissue images.

20. The apparatus of claim 19 , further comprising:

means for normalizing the training x-ray images and corresponding training soft-tissue images prior to extracting the set of wavelet features for each pixel in each of the training x-ray images.

21. The apparatus of claim 19 , further comprising:

means for reducing a dimensionality of the set of wavelet features for each pixel of each of the training x-ray images prior to training the regression function.

22. The apparatus of claim 19 , wherein said means for training a regression function to suppress bone structures in x-ray images comprises:

means for training the regression function using Bayesian Committee Machine (BCM) to approximate Gaussian process regression (GPR).

23. A non-transitory computer readable medium encoded with computer executable instructions for suppressing bone structures in x-ray images, the computer executable instructions defining steps comprising:

receiving an x-ray image having a plurality of pixels;

extracting a set of wavelet features for each pixel in the x-ray image; and

generating a soft-tissue image without bone structures from the x-ray image by using a trained regression function to determine an intensity value for the soft-tissue image corresponding to each pixel in the x-ray image based on the set of wavelet features extracted for each pixel in the x-ray image.

24. The non-transitory computer readable medium of claim 23 , further comprising computer executable instructions defining the step of:

normalizing the x-ray image prior to extracting the set of features for each pixel in the x-ray image.

25. The non-transitory computer readable medium claim 23 , further comprising computer executable instructions defining the step of:

reducing a dimensionality of the set of wavelet features for each pixel of the x-ray image prior to generating the soft-tissue image.

26. The non-transitory computer readable medium of claim 23 , wherein said trained regression function is trained using Bayesian Committee Machine (BCM) to approximate Gaussian Process Regression (GPR).

27. A non-transitory computer readable medium encode with computer executable instructions for suppressing bone structures in an x-ray image, the computer executable instructions defining steps comprising:

receiving a plurality of sets of training images, each set of training images comprising a training x-ray image and a corresponding training soft-tissue image;

extracting a set of wavelet features for each pixel in each of the training x-ray images; and

training a regression function to suppress bone structures in x-ray images based on the extracted wavelet features for the training x-ray images and the corresponding training soft-tissue images.

28. The non-transitory computer readable medium of claim 27 , further comprising computer executable instructions defining the step of:

normalizing the training x-ray images and corresponding training soft-tissue images prior to extracting the set of wavelet features for each pixel in each of the training x-ray images.

29. The non-transitory computer readable medium of claim 27 , further comprising computer executable instructions defining the step of:

reducing a dimensionality of the set of wavelet features for each pixel of each of the training x-ray images prior to training the regression function.

30. The non-transitory computer readable medium of claim 27 , wherein the computer executable instructions defining the step of training a regression function to suppress bone structures in x-ray images comprise computer executable instructions defining the step of:

training the regression function using Bayesian Committee Machine (BCM) to approximate Gaussian process regression (GPR).