11th European Conference on Non-Destructive Testing (ECNDT 2014), October 6-10, 2014, Prague, Czech Republic Beam Attenuation Grid Based Scatter Correction Algorithm for More Info at Open Access Database www.ndt.net/?id=16542 Cone Beam Volume CT Dongcai Hu, Hao Chen, Yuan Wang, Yunbin Chen, Zhou Xu Institute of Applied Electronics, China Academy of Engineering Physics, Mianyang 621900, China Phone: +86 0816 2493340, Fax: +86 0816 2493340: E-mail: dongcaihu@163.com Abstract In the flat panel detector-based cone beam CT system, the scatter photons which reach the flat panel detector increase the detected signal, produce the image cupping artifact and reduce the image contrast. To conquer the X-Ray scatter problem, this study proposes the beam attenuation grid -based scatter correction algorithm that needs to add a beam attenuation grid (BAG) between the X-Ray source and the test object. This paper mainly describes the beam attenuation grid device and the beam attenuation grid based scatter correction algorithm. We also introduce some image quality assessments which can evaluate the effect of the scatter correction method. The cupping degree and the image contrast degree are evaluated with the cylinder sample and the lantern ring sample, respectively. The results indicate that this method work well for the cone beam CT system. It can effectively reduce the cupping artifact and increase the image contrast of the reconstruction slice image. Key words:cone beam CT,Compton scatter,scatter correction,beam attenuation grid 1 Introduction Flat panel detector-based cone beam CT(CBCT) system acquires the projection images of the test object with the cone beam X-Ray source and the flat panel detector, which recently have made a significant progress in the reconstruction algorithms, system integrations and its applications [1-3]. However, scatter is found to be one of the most important factors which affects the quality of images for the cone beam CT. Nowadays, developing and optimizing an X-Ray scatter control and reduction technique remains a significant challenge for the cone beam CT system because CBCT is much less immune to scatter than fan beam CT [4]. Scatter in the CBCT system, without effective control and reduction, causes the error of the linear attenuation coefficient, and produces the image cupping artifact, it also reduces the image contrast and introduces notable construction errors [5-6]. All of these indicate that scatter is very disadvantageous for the testing of the small defection. Nowadays, the scatter correction technique in the three-dimension CT system is still a great challenge. Because of the use of the flat panel detector, some conventional scatter correction methods which can be effectively used in the two-dimension CT system can not be used in the three-dimension CT system, for example the collimator method. Recently, as the development of the three-dimension CT technique, many research have been launched for the scatter correction, and a series of scatter reduction methods have been reported [7-9].The most used scatter correction method is the beam stop array (BSA) algorithm [4,10]. This method demands that the
lead ball absolutely stop the beam and the projection area of the lead ball is smaller as possible as can. In the low energy cone beam CT system, BSA scatter correction method can get a good result, but if in the high energy CBCT system, BSA method is very hard to meet the size demand of the lead ball. In the application of industrial CT (ICT), for many high density objects which only high energy can pass through, it is hardly to get the appropriate result. We have also developed a beam attenuation grid scatter correction algorithm. This method can overcome the high demand of the small lead ball size in the beam stop array correction method for the high energy CBCT. In the remainder of this paper, we describe the beam attenuation grid-base scatter correction method for the CBCT and evaluate its feasibility with the cylinder sample and the lantern ring sample. 2 Materials and methods 2.1 The Scatter Mechanisms and Effects It is well known that the X-Ray scatter mainly originates from the Compton effect and the photoelectric effect. During the interaction between the X-Ray photons and the object they pass through, as shown in figure 1(a), part of X-Ray photons depart from their original linear trajectories, which cause the lower computed tomography (CT) numbers inside the material and bring the reconstruction slice image cupping artifact. Because the deflection angle of the X-Ray photons is stochastic, the distribution of the scatter photons received by the detector is the low frequent signal, as shown in figure 1(b), described by the dash line. The scatter photons mixing with the primary photons constitute the mixed signal, which reduce the contrast of the projection image, as shown in figure 1(b), described by the dot line. Consequently, scatter not only produces the image cupping artifact but also reduces the image contrast. object Intensit primary+scatter primary a Figure 1.The interaction between the X-Ray photons and the object (a) and the effect of the Compton scatter (b). 2.2 Experiment Device The scatter correction method mentioned in this paper needs to add a beam attenuation grid (BAG) between the X-Ray source and the test object, as shown in figure 2(a). The beam attenuation grid is made of the organic glass filled with small steel balls, as shown in figure 2(b). We only require that the X-Ray photons can absolutely penetrate the steel balls and the test object and that the shadow area of each steel ball in the projection image is as small as possible. Obviously, it is very easy to meet the demand of the beam attenuation grid. In our beam attenuation grid, the diameter of each steel ball is 3mm and the ball distance between each other is 5mm, as shown in figure 2(b). scatter detector b primary scatter position
Detector Beam Attenuation Grid X-Ray Source Phantom a 5mm b 3mm Figure 2. The configuration of the scatter correction system (a) and the structure of the beam attenuation grid (b). 2.3 The Measurements of the Scatter Field When we scan the test object, as shown in figure 3(a), we suppose that the initial photons emitted by the X-Ray source are I 0. After they pass through the test object the primary photons are I 3, the scatter photons are S and the detector received photons are C 1 (C 1 = I 3 + S). Then, we add the beam attenuation grid between the test object and the X-Ray source. Keeping the same voltage and the same current, we scan the test object with the beam attenuation grid, as shown in figure 3(b). This time, the photons emitted by the x-ray are also I 0. After they pass through the steel ball the photons are I 1. Then, after they pass through the test object, the photons are I 2, the scatter photons are S and the detector received photons are C 2 (C 2 =I 2 +S). test object test object I 0 a S C 1 I 3 Figure 3, the illustration of the measurement of the scatter field. (a) the case of the test object only. (b) the case of test object with the beam attenuation grid. As shown in figure 3(a), according to the Beer theorem, when we scan the test object only, taking the test object as the research case, we write as 2 I3 C1 S I0e u l = = (1) As shown in figure 3(b), while we scan the test object with the beam attenuation grid, taking the steel ball as the research case, we also write 1 I1 I0e u d Taking the test object as the research case, we can obtain I 0 S steel ball b I 1 I 2 = (2) I = I e = I e (3) u2l ( u1d + u2l) 2 1 0 C 2 where d is the diameter of the steel ball, l is the thickness of test object, 1 is the attenuation coefficient of the steel ball and 2 is the attenuation coefficient of the test object.
When the X-Ray photons attenuate the steel ball and the test object, we know Putting (4) in (3), we obtain I2 = C2 S (4) = = 2 I2 C2 S I1e u l Dividing (5) by (1), we can obtain I1 C2 S = I0 C1 S Then, putting (6) in (2), we can get Hence, u d C S I = I e = I S 1 2 1 0 0 C1 (5) (6) (7) C2 C1e S = 1 e (8) Putting C k = C 2 1 in (8), we obtain ( C ) 1 k e S = 1 e (9) Hence, we can obtain the scatter-to-primary ratio (SPR) SPR = ( k e ) 1 e (10) According to the above formulas, we can calculate the scatter-to-primary (SPR) value of the projection position of each steel ball center. Then employing a spatial cubic spline interpolation on these values, we can estimate the scatter-to-primary distribution of the protection images at each different scanned angle. 2.4 Scatter Correction Method The flow chart of the scatter correction method this study proposed is shown in figure 4. The mainly experiment steps are: (1) correct the projection images that output of the detector, initial the X-Ray source, then acquire one projection image of the air; (2) put the beam attenuation grid between the X-Ray source and the test object, then acquire one projection image of the beam attenuation grid; (3) put the test object on the rotated worktable, then acquire 360 projection images (set I) of the test object with the beam attenuation grid at each different angles; (4) remove the beam attenuation grid, then acquire 360 projection images (set II) of the test object at each different angles.
scatter field distribution images scatter corrected projection images Figure 4. The flow chart of the scatter correction method based on the beam attenuation grid. We can calculate the projection position of each steel ball center from the projection image of the beam attenuation grid. In this paper, firstly, we binary the projection image of the beam attenuation grid by the da-jin method. Secondly, we trace the counter of the binarization image to obtain the counter of each steel ball in the beam attenuation grid. Thirdly, according to each counter of the steel ball, using the circle fitting technique, we can get the center and the diameter of each steel ball. Lastly, regarding the cycle center as the square center and the cycle diameter as the square edge, we search the minimum value in the square area and regard it as the projection position of the ball center. In order to improve the veracity of this method, we do some simulated experiments and the results show that it fits well. According to formula (2), we can figure out the penetrate coefficient e 1 d of each steel ball center by the projection image of the air and the beam attenuation grid. Using the method mentioned in the part 2.3, we can obtain the scatter-to-primary distribution by the projection images (set I) of the test object with the beam attenuation grid, the projection images (set II) of the test object, the projection position of each steel ball center and the penetrate coefficient of each steel ball. Then, multiplying the SPR distribution by the projection images of the test object, we can get the scatter field distribution of each different projection angle. Finally, subtracting the scatter field images from the projection images (set II) of the test object, we can get the scatter corrected projection images of the test object at each different projection angle. Adopting the filtered backprojection algorithm (i.e., FDK), we can reconstruct the scatter corrected slice image by the scatter corrected projection images. 3 Application Examples 3.1 Tested Samples In order to validate the feasibility of the method we mentioned, two samples are used for this study: a cylinder sample (radius = 20mm, made of aluminum) and a lantern ring sample (inside cylinder: radius = 10mm, made of iron; outside ring: out radius = 20mm, inside radius = 10mm, made of organic glass). In the scan of the cylinder sample and the lantern ring sample, the operating voltage and the current of the X-Ray tube are set to be 320kv, 0.24mA, respectively.
3.2 Image Quality Assessment In this study, the cupping degree and the image contrast degree are calculated for the reconstruction slice image to assess the performance of the scatter correction method [11]. The cupping artifact is measured by the cylinder sample to evaluate the image uniformity of the reconstructed image. In order to quantify the cupping artifact, we calculate the cupping degree which defines as follow: D cup edge center = 100% (11) edge where, edge is the mean value at four regions of interest (ROI), which are defines as up, down, left, right edge regions of the image, and image. center is the mean value at the center region of the We evaluate the image contrast by the lantern ring sample using the image contrast degree which defines as follow: D ROI background contrast = (12) background where, ROI is the mean value of the insider iron cylinder and background is the mean value of the outside organic glass ring. 3.3 Results The scatter correction algorithm described above was tested by the cylinder sample and the lantern ring sample. For each case, we compared the linear gray distribution of the reconstructed image at the same position of the uncorrected image and the scatter corrected image. Figure 5(a), 5(b) show the uncorrected image and scatter corrected image of the cylinder sample, respectively. Figure 5(c) illustrates the horizontal linear gray distribution across the center of the image at the same position of the uncorrected image and the corrected image. Clearly, the image corrected by the Beam Attenuation Grid based scatter correction method represents the better image uniformity and the better image contrast. a b
c corrected The gray value uncorrected Figure 5, (a) the uncorrected image of the cylinder sample, (b) the scatter corrected image, (c) the linear gray distribution of the uncorrected image and the scatter corrected image at the same position Distance (pixel) In the cylinder sample, we calculate the mean value at five regions of interest, which are up, down, left, right and center regions of the image. And we regard the average of the up, down, left and right mean value as the edge value. From table 1, the cupping ratio of the uncorrected image and the corrected image are 11.89% and 5.51%, respectively. It revealed that the scatter correction method could reduce the cupping artifact. Table 1 Image cupping artifact assessment for the sour image and the scatter corrected image of the cylinder sample edge D center cup up down left right average uncorrected 0.036141 0.036408 0.036533 0.036149 0.036307 0.031991 11.89% corrected 0.051122 0.052352 0.050297 0.049099 0.050717 0.047921 5.51% Figure 6(a), 6(b) show the uncorrected image and scatter corrected image of the lantern ring sample, respectively. Figure 6(c) illustrates the horizontal linear gray distribution across the center of the image at the same position of the uncorrected image and the corrected image. Clearly, the image corrected by the Beam Attenuation Grid based scatter correction method represents the better image uniformity and the better image contrast. a b
c corrected The gray value uncorrected Distance (pixel) Figure 6, (a) the uncorrected image of the lantern ring sample, (b) the scatter corrected image, (c) the linear gray distribution of the uncorrected image and the scatter corrected image at the same position In the lantern ring sample, we regarded the outside ring of the organic glass as the region of the background and regarded the inside iron cylinder as the region of the interest. As show in table 2, the contrast degree of the uncorrected image and the corrected image were 3.85 and 4.80, respectively. It represented that this scatter correction method could improve the image contrast. Table 2 Image contrast assessment for the sour image and the scatter corrected image of the lantern ring sample ROI background D contrast uncorrected 0.014885 0.072152 3.85 corrected 0.018020 0.104615 4.80 4 Discussion and conclusion This study describes the Beam Attenuation Grid -based scatter correction method for the cone beam CT system. The proposed method possesses the same properties in the scatter correction as the Beam Stop Array method. The new method makes it easier to meet the demand of the experiment devices. Furthermore, the BAG method can be easily used in the domain of industrial computed tomography. The feasibility of the BAG algorithm is proved to use the cylinder sample and the lantern ring sample. In order to quantify the performance of the proposed scatter correction method, aiming at the cupping artifact and the image contrast, we define the cupping degree and the image contrast degree. In the cylinder sample, the cupping degree decreases from 11.89% to 5.51%. Obviously, the BAG algorithm can reduce the cupping artifact. In the lantern ring sample, the image contrast degree increases from 3.85 to 4.80. It represents that this scatter correction method could improve the image contrast. All the above results indicate that Beam Attenuation Grid-based scatter correction method can effectively reduce the image cupping artifacts, increase the image contrast and improve the image quality. In conclusion, the Beam Attenuation Grid-based scatter correction method is a novel and effective approach to solve the X-ray scatter problem in the CBCT system. However, in the beam attenuation grid, the distance between each steel ball can not be small enough. This will get the difficult to the SPR surface fitting. Because of the fitting error, it will
bring a little noise. Further work will focus on the fitting technique of the SPR surface. References 1. L.A.Feldkamp, L.C.Davis, and J.W.Kress, Practical cone-beam algorithm, J.Opt.Soc.Am., 1(6),pp.612-619(1984). 2. X.Wang and R.Ning, A cone-beam reconstruction algorithm for circle-plus-arc data acquisition geometry, IEEE Trans.Med.Imag.vol.18, pp.815-824, 1999. 3. Ning R, Chen B, Yu R, Conover D, Tang X, Ning Y, Flat panel detector-based cone-beam volume CT angiography imaging: system evaluation, IEE Trans.Med.Imag.2000, 19: 946-963. 4. Ruola Ning, Xiangyang Tang, D.L. Conover, X-Ray scatter suppression algorithm for cone beam volume CT. Proc. SPIE vol. 4682, 1605-7422, 2002. 5. M.Endo, T.Tsunoo, N.Nakamori, K.Yoshida, Effect of scattered radiation on image noise in cone beam CT, Med.Phys., 28(4), pp.469-474(2001). 6. J.H.Siewerdsen, D.A.Jeffrey, Cone-beam computed tomography with a flat- panel imager: Magnitude and effects of X-ray scatter, Med.Phys., 28(2), pp.220-231(2001). 7. James A, Sorenson, Jacqueline Floch, Scatter rejection by air gaps: an empirical model, Med.Phys, 12,308-316(1985). 8. J.A.Seibert, J.M.Boone, X-Ray scatter removal by deconvolution, Med.Phys, 15, 567-575(1988). 9. Ogawa K, Harata Y, Ichihara Y, Kubo A, Hashimoto S, A practical method for position-dependent Compton-scatter correction in single photon emission CT, IEEE Trans Med Imag 1991; 10: 408 412. 10. Weixing Cai, Ruola Ning, David Conover, Scatter correction using beam stop array algorithm for cone-beam CT breast imaging. Proc. SPIE vol.6142, 61423E-1. 11. Y.C. Ni, M.L. Jan, K.W. Chen, Y.D. Cheng, K.S. Chuang, Y.K. Fu, Magnitude and effects of X-ray scatter of a cone-beam micro-ct for small animal imaging, Nuclear Instruments and Methods in Physics Research, A 569 (2006) 245 249.