System and method for coronary digital subtraction angiography

1. A method for extracting coronary vessels from a contrast image, comprising:

calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field;

generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images;

combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer; and

extracting a coronary vessel layer from the contrast image by subtracting the final estimated background layer from the contrast image.

2. The method of claim 1 , wherein said step of calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field comprises:

detecting vessel regions in the contrast image using learning-based vessel segment detection;

determining the background region of the contrast image based on the detected vessel regions; and

estimating background motion between each of the plurality of mask images and the background region of the contrast image by estimating the motion field between each of the plurality of mask images and the background image and performing covariance-based filtering over the estimated motion field.

3. The method of claim 1 , further comprising:

receiving a sequence of contrast images of a coronary region, wherein said contrast image is one of the sequence of contrast images; and

repeating said steps of receiving a contrast image, calculating a motion field, generating a plurality of background layer predictions, combining the plurality of background layer predictions, and extracting a coronary vessel layer from the contrast image for each of the sequence of contrast images.

4. The method of claim 1 , wherein the plurality of mask images comprises a sequence of mask images of the coronary region taken over at least one full cardiac cycle.

5. The method of claim 1 , wherein said step of detecting vessel regions in the contrast image comprises:

applying a set of steerable filters to the contrast image to detect ridge features in the contrast image; and

detecting vessel segments at the detected ridge features using a learned binary classifier.

6. The method of claim 1 , wherein said step of generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images comprises:

determining a predication probability density function for each mask image to predict pixel values of the background layer based on the motion field and covariances using uncertainty propagation.

7. The method of claim 1 , wherein said step of combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer comprises:

fusing multiple prediction probability density functions of the background layer by calculating a minimum-mean-square error (MMSE) estimate.

8. The method of claim 7 , wherein the multiple prediction probability density functions of the background layer comprise the plurality of background layer predictions generated from the plurality of mask images and one or more predictions generated from background layer estimates of one or more previous contrast images in a sequence of contrast images.

9. The method of claim 1 , further comprising:

generating an enhanced coronary image by enhancing coronary vessels in the contrast image based on the extracted coronary vessel layer.

10. The method of claim 1 , further comprising:

generating an enhanced coronary image by fading the final estimated background layer in the contrast image.

11. An apparatus for extracting coronary vessels from a contrast image, comprising:

means for calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field;

means for generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images;

means for combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer; and

means for extracting a coronary vessel layer from the contrast image by subtracting the final estimated background layer from the contrast image.

12. The apparatus of claim 11 , wherein said means for calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field comprises:

means for detecting vessel regions in the contrast image using learning-based vessel segment detection;

means for determining the background region of the contrast image based on the detected vessel regions; and

means for estimating background motion between each of the plurality of mask images and the background region of the contrast image by estimating the motion field between each of the plurality of mask images and the background image and performing covariance-based filtering over the estimated motion field.

13. The apparatus of claim 11 , wherein the plurality of mask images comprises a sequence of mask images of the coronary region taken over at least one full cardiac cycle.

14. The apparatus of claim 11 , wherein said means for generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images comprises:

means for determining a predication probability density function for each mask image to predict pixel values of the background layer based on the motion field and covariances using uncertainty propagation.

15. The apparatus of claim 11 , wherein said means for combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer comprises:

means for fusing multiple prediction probability density functions of the background layer by calculating a minimum-mean-square error (MMSE) estimate.

16. The apparatus of claim 15 , wherein the multiple prediction probability density functions of the background layer comprise the plurality of background layer predictions generated from the plurality of mask images and one or more predictions generated from background layer estimates of one or more previous contrast images in a sequence of contrast images.

17. The apparatus of claim 11 , further comprising:

means for generating an enhanced coronary image by enhancing coronary vessels in the contrast image based on the extracted coronary vessel layer.

18. The apparatus of claim 11 , further comprising:

means for generating an enhanced coronary image by fading the final estimated background layer in the contrast image.

19. The apparatus of claim 11 , further comprising:

means for receiving the plurality of mask images; and

means for receiving a sequence of contrast images.

20. A non-transitory computer readable medium encoded with computer executable instructions for extracting coronary vessels from a contrast image, the computer executable instructions defining steps comprising:

calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field;

generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images;

combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer; and

extracting a coronary vessel layer from the contrast image by subtracting the final estimated background layer from the contrast image.

21. The computer readable medium of claim 20 , wherein the computer executable instructions defining the step of calculating a motion field between each of a plurality of mask images and a background region of the contrast image and calculating covariances of motion vectors of each motion field comprise computer executable instructions defining the steps of:

detecting vessel regions in the contrast image using learning-based vessel segment detection;

determining the background region of the contrast image based on the detected vessel regions; and

estimating background motion between each of the plurality of mask images and the background region of the contrast image by estimating the motion field between each of the plurality of mask images and the background image and performing covariance-based filtering over the estimated motion field.

22. The computer readable medium of claim 19 , further comprising computer executable instructions defining the steps of:

receiving a sequence of contrast images of a coronary region, wherein said contrast image is one of the sequence of contrast images; and

repeating said steps of receiving a contrast image, calculating a motion field, generating a plurality of background layer predictions, combining the plurality of background layer predictions, and extracting a coronary vessel layer from the contrast image for each of the sequence of contrast images.

23. The computer readable medium of claim 20 , wherein the plurality of mask images comprises a sequence of mask images of the coronary region taken over at least one full cardiac cycle.

24. The computer readable medium of claim 19 , wherein the computer executable instructions defining the step of generating a plurality of background layer predictions by generating a background layer prediction from each of the plurality of mask images based on the motion field and the covariances of the motion field from each of the plurality of mask images comprise computer executable instructions defining the step of:

determining a predication probability density function for each mask image to predict pixel values of the background layer based on the motion field and covariances using uncertainty propagation.

25. The computer readable medium of claim 19 , wherein the computer executable instructions defining the step of combining the plurality of background layer predictions using statistical fusion to generate a final estimated background layer comprise computer executable instructions defining the step of:

fusing multiple prediction probability density functions of the background layer by calculating a minimum-mean-square error (MMSE) estimate.

26. The computer readable medium of claim 25 , wherein the multiple prediction probability density functions of the background layer comprise the plurality of background layer predictions generated from the plurality of mask images and one or more predictions generated from background layer estimates of one or more previous contrast images in a sequence of contrast images.