Contrast-dependent regularization strength in the iterative reconstruction of CT images
A method is disclosed for reconstructing image data of an examination object from measured data, the measured data being captured during a rotating movement of a radiation source of a computed tomography system around the examination object. In at least one embodiment, different iteration images of the examination object are determined successively from the measured data by way of an iterative algorithm, wherein with the iterative algorithm the current iteration image at any one time is high-pass-filtered and weighted as a function of contrast using a variable which contains contrast information relating to the current iteration image at that time.
1. A method for reconstructing an image of an examination object from measured data, the measured data being captured during a rotating movement of a radiation source of a computed tomography system around the examination object, the method comprising:
determining, by a control and computing unit, different iteration images of the examination object, successively, from the measured data by use of an iterative algorithm, during each iteration of the iterative algorithm,
high-pass-filtering, by the control and computing unit, a current one of the iteration images,
determining a variable based on a contrast to noise ratio of the current one of the iteration images, the variable providing location dependent contrast information of the current one of the iteration images, and
weighting, by the control and computing unit, the high-pass-filtered current one of the iteration images as a function of contrast among pixels of the current one of the iteration images using the variable; and
determining a next one of the iteration images based on the current one of the iteration images and the weighted current one of the iteration images, wherein
an amount of reduction in noise associated with the current one of the iteration images depends on the location dependent contrast information, and
a level of admixture of the current one of the iteration images and the weighted current one of the iteration images for determining the next one of the iteration images is determined using the variable.
2. The method as claimed in claim 1 , wherein the variable contains contrast information for each pixel of the current iteration image at a given time.
3. The method as claimed in claim 1 , wherein the variable is written as
β
=
β
0
ⅇ
-
(
df
σ
)
2
,
where β 0 is a parameter denoting a scalar regularization strength effective in the current iteration, df is a contrast value and σ is a noise value of the current one of the iteration images.
4. The method as claimed in claim 1 , wherein the determining determines the next one of the current iteration images by applying the weighted current one of the iteration images to the current one of the iteration images.
5. A Computer Tomography (CT) system for reconstructing an image of an examination object from measured data, the measured data being captured during a rotating movement of a radiation source of the CT system around the examination object, the CT system comprising:
a control and computing unit configured to,
determine different iteration images of the examination object, successively, from the measured data by use of an iterative algorithm, during each iteration of the iterative algorithm, the control and computing is configured to,
high-pass-filter a current one of the iteration images,
determine a variable based on a contrast to noise ratio of the current one of the iteration images, the variable providing location dependent contrast information of the current one of the iteration images, and
weight the high-pass-filtered current one of the iteration images as a function of contrast among pixels of the current one of the iteration images using the variable; and
determine a next one of the iteration images based on the current one of the iteration images and the weighted current one of the iteration images, wherein
an amount of reduction in noise associated with the current one of the iteration images depends on the location dependent contrast information, and
a level of admixture of the current one of the iteration images and the weighted current one of the iteration images for determining the next one of the iteration images is determined using the variable.
6. A non-transitory computer readable medium including a computer program product, the computer program product comprising instructions, which when executed on a computer, causes the computer to perform functions for reconstructing an image of an examination object from measured data, the measured data being captured during a rotating movement of a radiation source of a computed tomography (CT) system around the examination object including:
determining different iteration images of the examination object, successively, from the measured data by use of an iterative algorithm, during each iteration of the iterative algorithm,
high-pass-filtering a current one of the iteration images,
determining a variable based on a contrast to noise ratio of the current one of the iteration images, the variable providing location dependent contrast information of the current one of the iteration images, and
weighting the high-pass-filtered current one of the iteration images as a function of contrast among pixels of the current one of the iteration images using the variable; and
determining a next one of the iteration images based on the current one of the iteration images and the weighted current one of the iteration images, wherein
an amount of reduction in noise associated with the current one of the iteration images depends on the location dependent contrast information, and
a level of admixture of the current one of the iteration images and the weighted current one of the iteration images for determining the next one of the iteration images is determined using the variable.
7. The method as claimed in claim 1 , wherein the iterative algorithm enables a noise reduction based on the variable at low-contrast locations of each of the iteration images.
8. The CT system as claimed in claim 5 , wherein the iterative algorithm enables a noise reduction based on the variable at low-contrast locations of each of the iteration images.
9. The non-transitory computer readable medium as claimed in claim 6 , wherein the iterative algorithm enables a noise reduction based on the variable at low-contrast locations of each of the iteration images.
10. The method as claimed in claim 1 , wherein high-pass filtering includes applying a regularization term to the current one of the iteration images, the regularization term indicating a desired image characteristic of the reconstructed image.
11. The method as claimed in claim 1 , wherein the determining a next one of the iteration images determines the next one of the iteration images by subtracting the weighted current one of the iteration images from the current one of the iteration images.
12. The method as claimed in claim 11 , wherein the determining a next one of the iteration images multiplies the weighted current one of the iteration images by a convergence factor prior to subtracting the weighted current one of the iteration images from the current one of the iteration images.
13. The CT system as claimed in claim 5 , wherein the control and computing unit is configured to high-pass filter the current one of the iteration images by applying a regularization term to the current one of the iteration images, the regularization term indicating a desired image characteristic of the reconstructed image.
14. The CT system as claimed in claim 5 , wherein the control and computing unit is configured to determine the next one of the iteration images by subtracting the weighted current one of the iteration images from the current one of the iteration images.
15. The CT system as claimed in claim 14 , wherein the control and computing unit is configured to multiply the weighted current one of the iteration images by a convergence factor prior to subtracting the weighted current one of the iteration images from the current one of the iteration images.
16. The non-transitory computer readable medium as claimed in claim 6 , wherein the high-pass filtering includes applying a regularization term to the current one of the iteration images, the regularization term indicating a desired image characteristic of the reconstructed image.
17. The non-transitory computer readable medium as claimed in claim 6 , wherein the determining a next one of the iteration images determines the next one of the iteration images by subtracting the weighted current one of the iteration images from the current one of the iteration images.
18. The non-transitory computer readable medium as claimed in claim 17 , wherein the determining a next one of the iteration images multiplies the weighted current one of the iteration images by a convergence factor prior to subtracting the weighted current one of the iteration images from the current one of the iteration images.