Patent Application
20090202034
This application claims priority of German application No. 10 2008 008 611.8 filed Feb. 12, 2008, which is incorporated by reference herein in its entirety.
The invention relates to a method for three-dimensional image reconstruction of a dynamically moved object from projection data of an imaging device as well as to an associated imaging device, especially a conventional x-ray device, a C-arm x-ray device or a computer tomography device.
For image reconstruction of x-ray projection data, primarily in computer tomography, the primary method employed is Filtered Back Projection (FBP). FBP is based on the assumption that there is a stationary object in the projection data. In filtered back projection for static objects each individual projection of the projection data is filtered after corresponding preprocessing with a convolution core and projected back into the object space. The sum of these back projections produces the reconstruction of the sought object. For non-stationary objects this assumption leads to movement artifacts such as duplicated objects, stripes or to a low contrast for example.
Especially in the reconstruction of an if necessary contrasted part of a coronary vessel system—the coronary sinus and the adjacent branch-off vessels, there are two main sources of movement in the form of the heart movement and breathing movement.
Different approaches to the reconstruction of dynamically moved objects such as a coronary vessel system for example are possible with the aid of the filtered back projection.
One method is what is referred to as “gating” of the projection data. In this case an identification of the movement state of the object to be reconstructed is known (movement phase). Then, depending on the gating strategy, only the projections which are most similar to a specific reference movement phase are selected for the reconstruction. Basically two approaches exist for this. The first option of a reconstruction with a subset of the projection data (projection gaps) leads to a reduced image quality. The second option is the recording of further projections until all gaps are almost filled. A further method is movement correction. Mostly the change in the object over time is not known. This will be estimated from the projection data (estimation of the movement). With filtered back projection a target movement phase is then reconstructed in which the movement difference is compensated for for each projection image.
The methods described at the outset are disadvantageous in that a number of consecutive processing steps must be applied in a time-intensive way.
The underlying object of the invention is to specify an improved method for three-dimensional image reconstruction as well as an associated imaging device by comparison with these previous methods.
This object is achieved by the features specified in the independent claims. Advantageous developments of the invention are specified in the dependent claims.
The subject matter of the invention is a method for three-dimensional image reconstruction of a dynamically moved object from projection data of an imaging device as well as an associated imaging device, especially a conventional x-ray device, a C-arm x-ray device or a computer tomography device. The inventive method comprises the following steps:
Depending on the above-mentioned comparison result, step c) can be at least partly repeated with a corrective transformation.
A further aspect of the invention is an imaging device, embodied with modules for three-dimensional image reconstruction from projection data in accordance with the above-mentioned method.
In an advantageous manner the invention describes a method for reconstruction of dynamically moved objects or objects able to change over time from projection data which is integrated into a method for filtered back projection if necessary. This enables the extra time spent in undertaking the reconstruction to be reduced.
A further advantage obtained from the invention is that movement artifacts are reduced in the reconstruction of dynamically moved objects.
Advantageously the object can represent a coronary vessel system which is contrasted if necessary.
Expediently the projection data can be filtered beforehand.
A further advantageous development of the invention makes provision for the comparison in step d) to be undertaken with the aid of maximum values from a quality function which is applied to the volumes compared in step d).
For determining the reference volume a tomographic and/or symbolic reconstruction from at least a part of the projection data can be applied.
The disjoint consistent subsets from step b) can be represented by single-element subsets.
Advantageously a transformation can be applied in step c) with the aid of checkpoints distributed over the object.
An exemplary embodiment of the invention is described in greater detail with reference to a drawing.
The drawing shows the following figures:
As a concrete application the invention is represented by an example of the reconstruction of a contrasted part of a heart vessel system—the coronary sinus and the adjacent branch-off vessels.
a) A reference volume R, which contains information about the object to be reconstructed and the movement phase to be reconstructed, is created from the projection data P. This can be done in diverse ways and will be chosen individually. Examples are the tomographic or symbolic reconstruction from all projection data or from a subset. Also conceivable are postprocessing steps such as windowing, threshold value formation, segmentation or similar. If the object movement is divided up into phases, this information can then be included in the selection of the projection data.
Movement phases in the form of relative heart phases which can be created from an EKG signal are available for the reconstruction of the coronary sinus. The reference volume is reconstructed from a subset of the projection data in the heart rest phase with the FBP. As a rule projections from the area between 60-85% are selected. These can be represented with the visualization software, if necessary windowed and cut to shape.
Using an iterative application of the algorithm the reference volume can be improved. This means that the reference volume of the i+1th iteration is created from the result of the ith reconstruction. This can be applied in the example of the coronary sinus.
b) The projection data is divided into disjoint and consistent subsets P={P1, . . . , Pn}. Let F(Pi) be the part reconstruction of the projection data in Pi. This can in its turn be determined dynamically or statically. The aim of this step is to find consistent subsets which are as large as possible.
For the coronary sinus reconstruction there is a breathing and a heart movement. In order to obtain consistent subsets in relation to the heart phases, the heart phases will be divided up into K equal-size intervals and the projections assigned in accordance with their heart phase. The part reconstructions are obtained by a recursive call of the algorithm specifically for correction of the breathing movement. In such cases the projections are divided up into single-element subsets. The transformation (see next step) is a strict translation which is restricted to one movement within the image plane. The target function and the reference volume remain unchanged. The part reconstruction obtained is thus breath-corrected and restricted to one heart phase area.
c) The definition of a transformation T(Θ), which is able to describe the movement of the object to be reconstructed using parameters Θ. Applied to a part reconstruction F(Pi), this is modified in accordance with the parameterization. Examples are a strict movement in the coordinate directions, a rotation, a scaling or a free deformation. When selecting T it should be ensured that where possible this describes a movement which is unique. This mostly depends on the size of the subsets Pi. If for example |Pi|=1, no downwards movement along the direction of projection can be determined, which is why this degree of freedom should remain unconsidered where possible.
The heart vessel movement is characterized by a global up and down movement of the vessel tree over the heart cycle. This is described by a translation in the coordinate system. To take account of local deformations of the vessel tree, checkpoints distributed evenly over the volume are used, to which displacement vectors are assigned. For positions between the checkpoints displacement vectors from the surrounding neighbors are interpolated. This is done for example using a linear or B-spline interpolation.
As already explained above, only single-element subsets are used for the breathing movement. The breathing movement is assumed to be strict. Only a translation within the image plane and not in the camera direction (downwards) is possible.
d) The definition of a quality function λ(R,T(Θ){F(Pi)}), which describes the match between the reference volume R and the transformed filtered back projection T(Θ){F(Pi)}. An optimally transformed part reconstruction of the search area of the parameter set Θ maximizes λ. Let this optimally transformed part reconstruction be T({circumflex over (Θ)}){F(Pi)}. For the coronary sinus it is useful for the intensity of the back projections to be maximized at bright points of the reference volume R. This is achieved by the following target function:
e) A complete reconstruction is provided by summing the transformed part reconstructions:
| Number | Date | Country | Kind |
|---|---|---|---|
| 10 2008 008 611.8 | Feb 2008 | DE | national |
