The present application relates to the diagnostic imaging arts. It finds particular application in computed tomography imaging of a subject that includes high density regions such as metal implants, dental fillings, and the like, and will be described with particular reference thereto. However, it can also find application in other types of tomographic imaging such as single photon emission computed tomography (SPECT), positron emission tomography (PET), three-dimensional x-ray imaging, and the like.
In CT imaging, high absorbing objects such as metal bodies may cause significant artifacts, which may compromise the diagnostic value of the image. Metal artifacts arise when the imaged region of interest contains metal implants, dental fillings, bullets, or other articles of high radiation absorption which prevent the x-rays from fully penetrating the subject. Projection line integrals passing through the regions of high density are so highly attenuated by the high density regions that data about other regions along the line integral are lost or overshadowed. This leads to substantial measurement errors. The filtered backprojection or other reconstruction process translates these measurement errors into image artifacts, e.g. streaks which typically emanate from the high intensity region. The streaks deteriorate image quality and can obliterate structure of the region.
The current correction algorithms substantially reduce the artifacts in highly corrupted regions. One such method for correcting metal artifacts includes performing filtered backprojection to generate an uncorrected reconstructed image, identifying a region of high density in the uncorrected reconstructed image, and replacing rays of projections that pass through the high density region with synthetic projection data having reduced absorption attenuation values. The corrected projection data again undergoes filtered backprojection to produce a corrected reconstructed image.
Another method, the sinogram completion method, replaces corrupted regions in the sinogram data of a corrected model. The original, uncorrected tomogram image is segmented into different material classes such as bone, tissue, air. Pixels of the metal are identified and assigned a Hounsfield number of the surrounding material. The sinogram model data is generated from the classified tomogram data by a forward projection adapted to the scanner geometry. The segments of the original tomogram that have been identified as metal are replaced by the respective segments from the model sinogram. A conventional backprojection is then used to reconstruct a metal reduced 3D image.
The above and other known correction algorithms work well in certain applications, particularly in images with a severe artifact. However, the known correction techniques cause a significant reduction of the contrast resolution.
There is a need for an automated correction technique that compensates for mild metal artifacts in the image. The present invention contemplates a method and apparatus that overcomes the aforementioned limitations and others.
According to one aspect of the present application, a diagnostic imaging system which corrects metal artifact streaks emanating from a high attenuating object in an uncorrected tomographic image is disclosed. A first vector means determines a direction vector for each pixel, which direction vector points in the direction of one of the high attenuating objects. A second vector means determines an orientation vector for each pixel, which orientation vector coincides with a steepest gradient direction. An adaptive filter means adaptively filters the uncorrected tomographic image based at least on one of the determined orientation size of the steepest gradient vector, and a size of the direction vector.
According to another aspect of the present application, a method of diagnostic imaging which corrects metal artifact streaks emanating from a high attenuating object in an uncorrected tomographic image is disclosed. A direction vector for each pixel is determined, which direction vector points in the direction of one of the high attenuating objects. An orientation vector for each pixel is determined, which orientation vector coincides with a direction of steepest gradient. The uncorrected tomographic image is adaptively filtered based at least on one of the determined orientation of the steepest gradient vector, a size of the steepest gradient vector, and a size of the direction vector.
One advantage of the present application resides in dynamic reduction of mild artifacts.
Another advantage resides in local dynamic adjustment of the noise smoothing.
Another advantage resides in reduction of degradation of image quality caused by other correction techniques.
Numerous additional advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments.
The invention may take form in various components and arrangements of components, and in various process operations and arrangements of process operations. The drawings are only for the purpose of illustrating preferred embodiments and are not to be construed as limiting the invention.
With reference to
Preferably, the radiation source 14 produces a fan-beam or cone-beam of x-rays. The radiation source 14 and the detector 18 are preferably mounted in oppositely facing fashion on a rotating gantry 20 so that the detector 18 continuously receives x-rays from the radiation source 14. As the source 14 and the detector 18 rotate about the examination region 16 on the rotating gantry 20, views are acquired over an angular range of preferably about 360° or more. Optionally, a reduced scan of between about 180° and 360° is used. In one embodiment, the detector 18 is replaced by a stationary detector ring mounted on a stationary gantry 22. Typically, a subject support 24 is linearly movable in an axial or z-direction by a motor means 26.
Multiple-slice computed tomography projection data are acquired by performing successive axial scans with the subject support 24 being stationary during each axial scan and stepped linearly between axial scans. In this arrangement, the detector 18 can have either a single row of detector elements (that is, a one-dimensional detector) or a two-dimensional array of detector elements. Alternatively, helical computed tomography projection data are acquired during continuous linear movement of the subject support 24 and simultaneous rotation of the gantry 20.
The outputs of detector elements of the radiation detector 18 are converted to electric acquired integrated attenuation projection values μd0 that are stored in a data memory 28. Each projection datum μd0 corresponds to a line integral of attenuation along a line from the radiation source 14 to a corresponding one of the detector elements of the detector 18. The projection data can be represented in a sinogram format in which each two-dimensional slice of the imaged region of interest is represented by a projection data array having coordinates of viewing angle (φ) and line integral index (n).
For typical fan-beam and cone-beam geometries, the line integral index n typically corresponds to a detector index indicating a detector element used to measure the projection of index n. It is contemplated, however, that the line integral index n may lack a direct correspondence with detector element number. Such a lack of direct correspondence can result, for example, from interpolation between rebinned projections.
With continuing reference to
With reference to
With reference again to
A combiner 50 fuses the sinogram completed image and the adaptively filtered image into a final corrected image in which highly corrupted tomogram regions are replaced by the result of the sinogram completed image and the remainder is replaced by the adaptively filtered image. Merging of the metal artifact reduced images is preferably performed by an appropriate fusion function such as the fusion function based on the artifact expectation map from the metal sinogram that is described in detail in W
Spatially successive artifact-corrected reconstructed image slices, slabs or volumes are accumulated in a corrected 3D image memory 52 to define a three-dimensional artifact-corrected reconstructed volume image. If, however, the acquired projection data is limited to a single slice of the region of interest, then the acquired projection data corresponding to the single slice is processed by the reconstruction processor 32 and the corrected 3D image memory 52 stores a two-dimensional artifact-corrected reconstructed image. Optionally, projection data corresponding to one or more image slices are acquired over a selected time interval to provide a temporal series of artifact-corrected reconstructed image slices or image volumes representative of a temporal evolution of the region of interest.
A video processor 54 processes some or all of the contents of the corrected 3D image memory 52 or, optionally, of the uncorrected 3D image memory 34 to create a human-viewable image representation such as a three-dimensional rendering, a selected image slice, a maximum intensity projection, a CINE animation, or the like. The human-viewable image representation is displayed on a display 56 of a user interface 58, which is preferably a personal computer, a workstation, a laptop computer, or the like. Optionally, selected contents of image memories 34, 52 are printed on paper, stored in a non-volatile electronic or magnetic storage medium, transmitted over a local area network or the Internet, or otherwise processed. Preferably, a radiologist or other operator controls the computed tomography imaging scanner 12 via an input means 60 to program a CT controller 62 to set up an imaging session, modify an imaging session, execute an imaging session, monitor an imaging session, or otherwise operate the scanner 12.
With continuing reference to
However, if it is determined that a pixel value is not degraded by a streak, a relatively narrow filter is applied in all directions. Preferably, for a streak, the filter means 80 applies an asymmetric Gaussian filter which is oriented toward the metal object 36. More specifically, the filter means 80 applies the local smoothing filter 82 with a function ƒ of characteristic width Δ which is a function of the local structure tensor and the direction vector {right arrow over (d)} at the pixel that is currently being smoothed. A filter adjusting means 84 adjusts the filter 82 to produce the biggest adjustment in the direction orthogonal to the direction vector {right arrow over (d)}, e.g. the direction perpendicular to the metal. The standard deviation σ⊥ of the Gaussian filter 82 perpendicular to the direction vector {right arrow over (d)} is a function of a size of the orientation vector {right arrow over (o)} or the maximum eigenvalue λmax of the structure tensor, and the dot product between the (normalized) direction vector {right arrow over (d)} and the (normalized) orientation vector {right arrow over (o)}:
where σ0, p1 and p2 are constants.
The filter adjusting means 84 adjusts the filter 82 to produce smaller adjustments in the direction parallel to the direction vector {right arrow over (d)}, e.g. the direction parallel to the metal. The standard deviation σ∥ of the Gaussian filter 82 parallel to the direction vector {right arrow over (d)} {right arrow over (d)} has a smaller value. For example:
σ∥=σ⊥/2.
As a result of such adjustments, the noise is adjusted locally at each pixel without substantially impairing the spatial resolution. The filter width in each pixel is adapted to a presence of the metal.
One example of a narrow filter for non-streak areas is one, which replaces the pixel to be filtered with the sum of 84% of its original grayscale value plus 2% of the value of each of its eight surrounding nearest neighbors. An example of a streak filter is one, which replaces the pixel to be filtered with the sum of 26% of its original grayscale value plus 30% of its nearest neighbor orthogonal to the streak direction in the direction of steepest gradient plus 20% of its next nearest neighbor in the same direction plus 10% of the next, next nearest neighbor in the same direction plus 2% of the seven other nearest neighbors in other directions. The actual magnitudes and number of pixels that the filter extends in each direction vary with the degree and nature of streak (or other) artifact. Various other filters are also contemplated.
The invention has been described with reference to the preferred embodiments. Obviously, modifications and alterations will occur to others upon reading and understanding the preceding detailed description. It is intended that the invention be construed as including all such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
This application claims the benefit of U.S. provisional application Ser. No. 60/649,676 filed Feb. 3, 2005, which is incorporated herein by reference.
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/IB2006/050346 | 2/1/2006 | WO | 00 | 8/1/2007 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2006/082563 | 8/10/2006 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5416815 | Hsieh | May 1995 | A |
5473656 | Hsieh et al. | Dec 1995 | A |
6507633 | Elbakri et al. | Jan 2003 | B1 |
6754298 | Fessler | Jun 2004 | B2 |
6940286 | Wang et al. | Sep 2005 | B2 |
6944264 | Bijjani et al. | Sep 2005 | B2 |
7356174 | Leue et al. | Apr 2008 | B2 |
7567693 | deCharms | Jul 2009 | B2 |
7660453 | Lang | Feb 2010 | B2 |
20030071220 | Bruder et al. | Apr 2003 | A1 |
20040257381 | Brunner et al. | Dec 2004 | A1 |
Number | Date | Country |
---|---|---|
2005076221 | Aug 2005 | WO |
Number | Date | Country | |
---|---|---|---|
20080152203 A1 | Jun 2008 | US |
Number | Date | Country | |
---|---|---|---|
60649676 | Feb 2005 | US |