METHOD, SYSTEM AND SOFTWARE PRODUCT FOR MOTION CORRECTION OF TOMOGRAPHIC IMAGES
The invention relates to the motion correc- tion of tomographic images.
BACKGROUND OF THE INVENTION
Tomography is a method which enables one to get information about the inner structure of an' article without damaging the article itself. The most common application area of tomography is medicine, in which a patient is scanned e.g. when conducting pharmacological researches. Since the scanning takes a long time, the motion of the patient causes inaccuracy of imaging. In addition, the measurement data, which is called sinograms, typically contains a lot of noise, making it difficult to generate the final image. A sinogram is not a complete image of an object to be scanned or measured, but its projections at regular intervals, ranging between 0°-180°. Mathematically, a tomography is usually modelled using a radon transformation. A sinogram is reconstructed to form an image using some approximation of a reverse Radon transformation. One such generally known reconstruction method is FBP (filtered backprojection) , in which singorams are filtered in the frequency plane e.g. with a ramp filter prior to the backprojection. A Ra- don transformation and an FBP reconstruction are known per se e.g. from the book "A. Jain, Fundamentals of Digital Image Processing, Englewood Cliffs, NJ, Prentice Hall International, 1989", on pages 434-448. In tomographic imaging, the determination of the motion of a patient is usually made based on reconstructed images, and the motion correction is directed to a two- or three dimensional series of im-
ages. The reconstruction of images is not a reverse operation in the same manner as a stackgram transformation, instead when reconstructing a sinogram to form an image and projecting the image back into a sino- gram, the final result does not completely correspond to the original sinogram. Motion correction cannot be made directly by means of sinograms either, because the geometry of sinograms is based on the measurement data of the cross-sectional plane. One common recon- struction method is filtered back-projection, which is known per se e.g. from the above-mentioned book of A. Jain. One solution for eliminating the noise caused by the sinogram information is to filter out the noise from the sinogram. This is, however, not simple because direct filtering of a sinogram can distort the sinusoids of the sinogram, and efficient filtering can cause a considerable inaccuracy in the image. The most common method at present is radial filtering which is performed in parallel to the lines of the sinogram.
The method is known e.g. from the reference publication "P.J. La Riviere, X. Pan, Nonparametric regression Sinogram Smoothing Using a Roughness-Penalized Poisson Likelihood Objective Function, IEEE Transac- tions on Medical Imaging, 19(8), 2000, pages 773- 786.". It is possible to filter a sinogram also cor- nerwise perpendicularly to the lines of the sinogram. It is, however, not a commonly used method due to the inaccuracy caused. Further, the filtering methods of sinograms usually are dependant on the target of application. As the motion of a research object in tomographic imaging is a common problem, several alternative solutions have been developed to address the problem. Patent publication US 6535570 relates to a method for eliminating the motion in traditional tomographic imaging. In the method, the motion is cor-
rected by means of a separate correlation coefficient. Also patent publication US 2002/0163994 discloses a method which uses a separate correction coefficient. Patent publication US 6026142 discloses a method in which the edges of images blurred by noise are retrieved from sinograms. The method in accordance with the invention enables one to search for the edges also from the final constructed tomographic image. Patent publication US 2002/0172321 calculates specific devia- tion signals for each projection angle. The problem with the prior-art solutions is the difficulty of filtering as well as losing information in reconstructions. Due to this there is an obvious need for an efficient filtering method.
OBJECTIVE OF THE INVENTION
The objective of the invention is to disclose a new type of method for eliminating the interference caused by motion in tomographic imaging. One specific objective of the invention is to disclose a method in which when filtering, information is lost as little as possible.
SUMMARY OF THE INVENTION
In the method and system in accordance with the invention, a patient or research subject is scanned using tomographic equipment to generate sino- grams. The invention is characterised by the fact that the sinograms are transformed to stackgrams when eliminating the inaccuracies caused by the motion of the research subject. The locus signals generated from the sine waves of the stackgrams are compared to a reference signal. The locus signal best corresponding to the reference is transferred into place. Once the comparison has been performed for all the signals, the
stackgrams are transformed back to sinograms for the construction of the final image, or are summed directly to form an image. The advantage of the method and system of the invention compared to conventional methods and systems is the improvement of the quality of images, which, in turn, enables use of bigger resolutions and obtaining more accurate research results. According to the invention, in filtering images, information is lost con- siderably less than by conventional methods. Further, the images need not be reconstructed for the motion correction, so necessary information is not lost in unnecessary reconstructions.
LIST OF FIGURES
Fig. 1 represents one functional block diagram of an embodiment in accordance with the invention, and Fig. 2 represents one system in accordance with the invention.
DETAILED DESCRIPTION OF THE INVENTION A sinogram is an image matrix whose lines contain projections about the measurement object. The idea of a stackgra is based on the fact that by means of a stackgram it is possible to find out in the sinogram all the sine waves constituting a sinogram, i.e. locus signals. This is achieved by transforming the sinogram to a three-dimensional stackgram consisting of a stack of overlapping back projected projections. The sine waves of a sinogram are in the stackgram parallel to the vertical axis of the stackgram. The stackgram is transformed back to the sinogram by applying a Radon transformation to every layer of the stackgram. In practice, the implementation of stack-
gram transformations means rotating back projected images, i.e. it is a question about an interpolation problem. A transformation from a sinogram to a stack- gram is a completely reverse transformation, so the outcome is congruent with the original one within the limits of the numerical accuracy of a computer. This is possible by means of discrete sine interpolation. A stackgram can be applied to motion correc- tion, which enables one to obtain a motion corrected sinogram or an FBP image by multiplying the planes of the stackgram by a 2D Ramp filter and by summing the stackgrams thus obtained to form an image. The motion correction is made prior to the reconstruction of the image. The transformation of a sinogram to a stackgram gives a possibility to perceive and correct the motion of a patient or testee when scanning and to return the motion correction to the sinograms without losing information in the process other than within the limits of the numerical accuracy of a computer. As it is possible in the stackgrams to locate a locus signal or a group of locus signals representing one point or a sharp edge, it is possible to find out character positions by the motion of which the motion of a patient or an object during the research is judged. The extent of motion is not very large in the situations in question because e.g. in brain examinations the patient's head is propped using a head support . The transformation from a sinogram to a stackgram (Sg, Stackgram) is performed by formula 1, wherein h(x,y,'φ) =Sg = g(x*∞s + y*s , ) (1) The difference of the transformation compared to the back projection is the absence of an integral operator as the sum of the elements has been substi-
tuted with a third dimension. Correspondingly, the transformation of the stackgrams back to sinograms succeeds by formula 2. g(l,φ) =S'1h = Rφh(x,y,φ) (2)
The aim for a very big image resolution in new emission tomographic devices requires the correction of even the smallest motions. It would be advan- tageous to make the corrections as near the measurement situation as possible, in other words on the sinogram level. This means that the research is collected e.g. as periods of one second, which are, after the motion correction, summed to form the desired time series sinograms, and are reconstructed to form the images . Fig. 1 represents the utilisation of tomographic images in motion correction. Fig. 1 represents the inventive part of the method because the scanning of an object and the generation of a sinogram are known per se and generally used. In motion correction, a sinogram, in a three-dimensional measurement, sinograms, are transformed to stackgrams, step 10. After this, the locus signals of stackgrams are compared to the selected locus signal that serves as a reference, step 11. The comparison is performed e.g. inside a 15*15*15 window. The locus signal best corresponding to the reference is transferred into its correct place in the stackgrams, step 12. Once all the locus signals have been compared, the stackgrams are transformed back to sinograms. The final images can be constructed from the motion-corrected sinograms. The idea is in principle of the same kind as in a three-dimensional case comparing a pack of images pixel by pixel, and making motion correction based on it. If locus signals of stackgrams are used for the comparison, then instead of information of one pixel,
information is obtained more to the tune of projection angles. In addition to this, each locus signal is slightly different, so the comparison of signals is successful, although there would be a lot of noise. At its simplest, the comparison can be performed based on a mean absolute error (MAE number, Mean Absolute Error) . Fig. 2 represents a system as shown in Fig. 2 comprising measurement equipment 20 and a processing system 22, which have been connected to one another via a telecommunication connection 21. The measurement equipment is preferably measurement equipment which is suitable for tomographic imaging and which is used to scan a patient or research subject. The processing system 22 is used to eliminate the measurement accuracy that is caused by the motion of the measurement object. The processing system further comprises means 23 for generating stackgrams based on the sinograms, means 24 for comparing the locus signals of the stack- grams to a reference, means 25 for transferring in place the locus signals best corresponding to the reference, as well as means 26 for transforming the stackgram back to a sinogram. The processing system 22 can be connected to other systems, which are not an object of this invention, such as e.g. to a medical system, for analysing a tomographic image. It must be noted that means 23-26 can also be implemented as software instead of hardware . The invention is not limited solely to the embodiment examples referred to above, instead many variations are possible within the scope of the inventive idea defined by the claims.