Abstract—Artifact is one of the most important factors in degrading the CT image quality and plays an important role in diagnostic accuracy. In this paper, some artifacts typically appear in Spiral CT are introduced. The different factors such as patient, equipment and interpolation algorithm which cau the artifacts are discusd and new developments and image processing algorithms to prevent or reduce them are prented.
Keywords—CT artifacts, Spiral CT, Artifact removal.
I.I NTRODUCTION
-RAY Computed Tomography (CT) has been successfully ud as an important medical image modality to reveal the interior organs of human body for many years. Many generations of CT scanners have been designed to improve their geometrical aspect and conquently to reduce the scanning time. In conventional CT scanners, the gantry rotates around stationary patient and all views in a slice are at same table position. It takes around 3-4 conds between slice scannings. Nowadays, with u of spiral (helical) and cone beam CT [1], the new generations of CT scanners, we are able to save the time by rapid examine of patient during a single breath-hold, as well as to demonstrate a true 3D imaging capability. Spiral CT provides a continuous gantry rotation and a continuous table motion as ga
ntry rotates. So each view is at different table position and no interscan delay is needed. The time between each slice is less than 1 cond. In spite of the new technologies, the physical principle of CT scanners is remaining the same and artifacts still persist in spiral CT as in conventional CT. For a x-ray CT, artifacts are the difference between the Hounsfield numbers (or CT numbers or HU) in resulting CT image and the expected attenuation coefficient of objects. Unfortunately, it is not always possible to say if it exists an artifact in CT images becau it is difficult to determine the expected values which depend on the viewer (such as physicians) judgment. Here, we focus on the typical artifacts which appear in x-ray CT images. Artifacts degrade enormously the CT image quality so that the physicians are not able to give a reliable Manuscript received September, 2007.
M. Yazdi is with the Department of Electrical Engineering, School of Engineering, Shiraz University, Shiraz, Iran (e-mail: yazdi@ shirazu.ac.ir).
L. Beaulieu is with Département de Radio-Oncologie et Centre de Recherche en Cancérologie, Hôtel-Dieu de Québec, 11 Côte du Palais, Québec G1R 2J6, Canada and Département de Physique, Génie Physique et d’Optique, Université Laval, Québec, Canada. diagnostic becau the anatomies are hidden or completely distorted.
We can classify the artifacts in four categories:
•Physics bad: include beam hardening, photon starvation
and undersampling artifacts. •Patient bad: include metallic and motion artifacts. •Scanner bad: artifacts caud by detector nsitivity and
mechanical instability.
•Spiral bad: artifacts ari due to spiral interpolation. Most artifacts appear as streak effects in CT images (Fig. 1 shows an example of streak artifacts). Metallic objects, beam hardening, photon starvation and object motion can cau the streak artifact. Other important artifacts ari from interpolation aspect of spiral CT. Careful patient positioning and optimum lection of scanning parameters are important factors in avoiding CT artifacts. However, some should be corrected by the scanner software. We discuss common artifacts in CT and give some recent solutions to prevent or reduce them.
Fig. 1 Example of artifacts produced by scanning a patient with two hip prosthes using a Siemens Somatom scanner, Hotel-Dieu
Hospital Center, Quebec, Canada
II.M ETAL A RTIFACT
A common problem in CT images is streak artifacts caud by the prence of high-attenuation objects in the field of view of scanner device. Metallic implants such as hip prosthes (Fig. 1), surgical clips and dental fillings cau this type of the artifact. The results of scanning a metal object are distinct regions in the projection matrix, i.e. the data exited directly from CT-scanner before CT image reconstruction, with high values. The reconstruction of this matrix using standard CT reconstruction method, i.e. filtered backprojection (FBP), caus the effect of bright and dark streaks in CT images (e
Artifacts in Spiral X-ray CT Scanners: Problems
and Solutions
Mehran Yazdi, and Luc Beaulieu
X
Fig. 1). As a matter of fact, the problem comes from an inaccurate beam hardening correction in FBP [2-3]. Although the new CT scanners are equipped with correction techniques for body organs, the high attenuation objects are still excluded. Metallic artifacts significantly degrade the image quality so that an effective radiation treatment planning cannot be applied.
Different techniques for metallic artifact reduction (MAR) have been propod [4-6]. The most efficient methods work on projection matrix. Two different methods have been introduced. In iterative reconstruction methods, the projection data associated with metal objects in projection matrix are disregarded and reconstruction is applied only for non-corrupted data [7-10]. Although the algorithms are reliable for incomplete/noisy projection data, they must deal with convergence problems and they are computationally expensive for clinical CT scanners (even with their fast implementation [11]).
In projection-interpolation bad methods [12-16], the projection data corresponding to rays through the metal objects are considered as missing data. Kalender et al. [12-13] identified manually the missing projections and replaced them by interpolation of non-missing neighbor projections. Rajgopa
l et al. [14] ud a linear prediction method to replace the missing projections. In other work [15], a polynomial interpolation technique is ud to bridge the missing projections. A wavelet multiresolution analysis of projection data is also propod to detect the missing data and interpolate them [16].
(a)
(b)
Fig. 2 Result of applying the method propod in [18] for reduction
of metallic artifacts; a) original CT image, b) modified CT image Recently, Mahenken et al. [17] ud another strategy for computing the interpolation value by the sum of weighted nearest not-affected projection values within a window centred by the missing projection. The weights are modeled only bad on the distance. Although they exploit the contribution of not-affected projections in all directions to determine the replacement values, they do not prerve the continuity of the structure of the projections. Furthermore, becau there is no continuity between resulting replacement values, the risk of noi production is also high. In my previous work [18], an optimization scheme is propod by exploiting both the distance and the value of not affected projections to determine the interpolation values and by using still an interpolation scheme to prerve the continuity of replacement values. This new scheme computed more effectively the interpolation values bad on the structure of nearest not affected projections and resulted an excellent performance in the ca of hip prosthesis. Fig. 2 shows the result of applying this propod method on two hip prosthes in Fig. 1.
III.P HOTON S TARVATION A RTIFACT
Photon starvation can cau streak artifacts, especially near to the heart, hip and shoulder where the patient’s tissue volume increas. This can be particularly en in patients with mass body. Artifacts ari becau some parts of individual projection can be very noisy due to insufficient photons passing through widest part of patient. Fig. 3 (a) shows the projects in the projection matrix for a patient. When the projects are reconstructed by standard algorithm of scanner, the noi is magnified, resulting in streaks in the direction of widest part (Fig. 3 (b)).
(a)
(b)
Fig. 3 Example of photon starvation artifact; a) matrix of projections, the circles show the noisy regions, b) resulting CT image (provided by Siemens Somatom scanner, Hotel-Dieu Hospital Center, Quebec,
Canada)
(a)
(b)
Fig. 4 Result of photon starvation removal by applying an optimal adaptive filtering; a) modified matrix of projections, b) resulting
modified CT image
Some scanners u a mA modulation allowing an increa of photon flux (by increasing current (mA) through the scanner tube) through widest parts without changing the photon flux through narrower parts. In this way, the number
of photons received by all detectors will be balanced. We can
also u an adaptive filtration of the projections to reduce this
effect [19]. In this approach, the areas on projection matrix with high values are smoothed, resulting in reducing the noi. An extension of this is multi-dimensional adaptive filtration, where further steps are taken to reduce noi levels in certain projections [20]. The success of this approach depends on choosing the best filter parameters and detecting correctly the areas where the filter should apply. Fig. 4 shows the result of applying an adaptive filter experimentally optimized for the patient in Fig. 3. As we can e, the streak artifacts are mostly removed.
IV. M OTION A RTIFACT
Patient movement during CT scanning results in image artifacts, which appear as streaks or blurring effects across an image (e Fig. 5 (a) as an example). The movement can be voluntary, such as th
e movement of the chest during inspiration and expiration, or involuntary such as cardiac motion, both cau motion artifacts. Severely injured patients or children frequently move during scanning, causing motion artifacts. In spiral CT scanners, the scan is usually short enough for patients to hold their breath, thus removing the possibility of breathing artifact. Besides, some techniques can be ud during scanning to reduce the effect of motion artifacts [21].
However, the cardiac motion is still a problem. Some new CT
(a)
(b)
Fig. 5 Example of patient motion artifact; a) original image with artifact, b) modified image with removal artifact (images are
provided by St George's Hospital, Tooting, London)
scanners are equipped with the technique of ECG gating which allows synchronizing the data acquisition with the rhythmic beating of the heart. The key elements of the new technology include acquiring an image of the heart by triggering an image acquisition scan starting at the point of the
cardiac cycle having minimized motion.
Some correction algorithms are also propod for motion
artifact removal. Crawford et al. developed a pixel-specific
filtered backprojection algorithm for motion artifact reduction
[22]. In their algorithm, in-plane motion is corrected by pixel-specific reconstruction in the coordinate system associated with the in-plane motion. We can also overscan the heart area and average the repeated projections to remove the effect of cardiac motion. Fig. 5 (b) shows the result of applying this approach to remove motion artifacts. V. S PIRAL A RTIFACT In general the same artifacts are produced in spiral and conventional scanning. Meanwhile, becau the spiral
scanning requires an interpolation process to recover the
consistent projections of individual slices, additional artifacts may be produced. Appearance and verity of spiral artifacts depend on scanning pitch and the type of interpolation algorithm. In single CT spiral scanner, the pitch is the table movement per tube rotation/slice collimation. For a typical 1 cond rotation scanner a pitch of 2 means the table traveled 10 mm with a 5 mm slice width or colli
mation. In multi-slice CT spiral scanners, the definition is table movement per rotation/single slice collimation. With a 1 c scanner there is 1 rotation per cond. So if the table travels 4 mm in a cond and a 1 mm collimator is ud then the pitch would be 4. Fig. 6 shows a spiral scanning and the pitch for this scanning. If pitch is incread while holding kVp, mA, and beam
collimation constant, then the table speed increas, mAs
decreas, patient do decreas, and either the effective slice width increas or the image noi increas. So for reducing the artifacts due to spiral rotation, we should decrea pitch. Fig. 7 shows the effect of reducing pitch for a multi-slice spiral scanner [23].
pitch = 4 x single slice pitch
Direction patient movement
Fig. 6 Multi-slice spiral scanning
Becau during the gantry rotation, the table is moving, we need to u an interpolation to average data either side of the reconstruction position to estimate projection data at that point. There are two algorithms: 360 and 180 degree雅思是什么
algorithms. In spiral scanning the individual views that reprents the X-ray absorption describes a spiral movement
over the patient (e Fig. 6). This means that for image reconstruction only one view is in the same plane for being reconstructed plane. All the other views are before and after
(a)
(b)
(c)
Fig. 7 The effect of decreasing pitch on reducing image artifacts; a)
pitch=4.5, b) pitch=3.5, c) pitch=2.5 [23]
the reconstruction plane. Thus, the views required for a pure cylindrical data t and also required fo
r an appropriate image reconstruction are calculated through interpolation of views with the same view angle. The weighting factor of the individual views within this interpolation is computed by distance of this view to the reconstruction plane and has a linear relation. This interpolation technique is called 360-degree interpolation. 180 degree interpolator makes u of the fact that opposite views are equivalent. So, a spiral data t interpolation is applied over 180 degrees of data on either sides of the reconstruction plane. This data t is also called complementary data (e Fig. 8). Again the weighting factor is computed by distance of the view to the reconstruction plane and has a linear relation. This interpolation technique is called 180-degree interpolation. Fig. 8 shows the interpolation techniques.
360 Degree interpolation
180 Degree interpolation data ° direct data °
Fig. 8 360 and 180 degree interpolation algorithms
Since the 360 degree interpolation us two views, each reconstructed image is reprenting a width slice. The 180 degree interpolation does not suffer from this enlarged slice width since it us only one view. So the artifacts due to interpolation are less effective. Fig. 9 shows the result of recon
structing a CT image using two interpolation techniques. As we can e, the 180 degree interpolation produces fewer artifacts. In general practice, the 180 degree interpolation algorithm is ud to reconstruct CT images.
VI. C ONCLUSION
Many sources can be the origin of CT artifacts. Artifacts degrade the CT image quality and conquently reduce diagnostic quality. Most artifacts can be prevented by using new designs in scanner technology, by careful positioning of patients during scanning, and by optimum lecting of scanner parameters (pitch, filter kind, delivered energy). Some others can be reduced by addressing the problem in software developments.
(a)
(b)
Fig. 9 The effect of interpolation algorithms on artifacts produced in reconstructed CT images; a) reconstructed CT image with 360 degrees interpolation algorithm, b) reconstructed CT image with 180 degrees interpolation algorithm. Arrows show the obrved artifacts due to 360 interpolations which are reduced in 180 interpolation (images are provided by Siemens Somatom scanner, Hotel-Dieu
Hospital Center, Quebec, Canada)
R EFERENCES
[1] Kalender WA, “Computed tomography: fundamentals, system
technology, image quality, and applications” Wiley, John & Sons, 1999.
[2] Joph PM, Ruth C. A method for simultaneous correction of spectrum
hardening artifacts in CT images containing both bone and iodine. Med Phys. 1997; 24: 1629-1634.
[3]Herman GT. Correction for beam hardening in computed tomography.
Phys Med Biol. 1979;24:81-106.新年快乐英文歌曲
[4] Robertson DD, Weiss PJ, Fishman EK, et al. Evaluation of CT
techniques for reducing artifacts in the prence of metallic orthopedic implants. Jour. Comput. Assist. Tomogr., 1988; 12(2): 236-241.
[5] Ebraheim NA, Coombs R, Rusin JJ, et al. Reduction of postoperative CT
might是什么意思artefacts of pelvic fractures by u of titanium implants. Orthopedics, 1990; 13: 1357-1358.
[6] Ling CC, Schell MC, Working KR, et al. CT-assisted asssment of
bladder and rectum do in gynecological implants. Int. Jour. Radiat.
subjectmatter
Oncol. Biol. Phys., 1987; 13: 1577-1582.
[7] Wang G, Snyder DL, O’Sullivan JA, et al. Iterative deblurring for CT
metal artifact reduction.IEEE Transactions in Medical Imaging, 1996;
15(5): 657-664.
[8] Robertson DD, Yuan J, Wang G, et al. Total hip prosthesis metal-artifact
suppression using iterative deblurring reconstruction.Journal of Comput. Assist. Tomogr., 1997; 21(2): 293-298.
[9] Nuyts J, De Man B, Dupont P, et al. Iterative reconstruction for helical
CT: a simulation study.Phys. Med. Biol., 1998; 43(4): 729-737.
[10]Wang G, Vannier MW, Cheng PC, et al. Iterative x-ray cone-beam
tomography for metal artifact reduction and local region reconstruction.
Microscopy and Microanalysis, 1999; 5(1): 58-65.
[11]Toft P. A very fast implementation of 2D iterative reconstruction
algorithms.IEEE proceeding of Medical Imaging Conference, 1996; 3: 1742-1746.burglary
[12]Kalender WA, Hebel R, Ebersberger JA. Reduction of CT artifactsuid
caud by metallic implants.Radiology, 1987; 164(2): 576-577. [13]Klotz E, Kalender WA, et al. Algorithms for reduction of CT artifacts
caud by metallic implants.SPIE Medical Imaging IV, 1990; 1234: 642-650.
[14]Srinivasa N, Ramakrishnan KR, Rajgopal K. Image reconstruction from
incomplete projection. Journal of Medical and Life Sciences Engg, 1997; 14: 1-19.新东方 王强
[15]Lewitt RM, Bates RHT. Image reconstruction from projections: III:
projection completion methods.Optik, 1978; 50(3): 189-204.
[16]Zhao S, Robertson DD, Wang G, et al. X-ray CT metal artifact reduction
奥巴马经典演讲稿
using wavelets: an application for imaging total hip prosthes.IEEE Transactions on Medical Imaging, 2000; 19(12): 1238-1247.
[17]Mahnken AH et al. A new algorithm for metal artifact reduction in
computed tomography: in vitro and in vivo evaluation after total hip replacement.Invest. Radiol., 2003 Dec; 38 (12): 769-775.
[18]Yazdi M, Gingras L, Beaulieu L. An adaptive approach of metal artifact
reduction in helical CT for radiation therapy treatment planning: experimental and clinical studies. Int. Jour. Radiat. Oncol. Biol. Phys., under publish.
[19]Hsieh J, Adaptive streak artifact reduction in computed tomography
resulting from excessive x-ray photon noi. Medical Physics 1998; 25,
(11): 2139-2147.
[20]Kachelriess M, Watzke O, Kalender WA. Generalized multi-dimensional
adaptive filtering for conventional and spiral single-slice, multi-slice, and cone-beam CT. Med Phys 2001; 28(4): 475-90.
[21]Goerres GW, Burger C, Kamel E et al. Respiration-induced attenuation
artifact at PET / CT: technical considerations. Radiology 2003; 226: 906-910.
[22] C. R. Crawford, K. F. King, C. J. Ritchie, and J. D. Godwin. Respiratory在线学习网
2020年9月四级真题
compensation in projection imaging using a magnification and displacement model. IEEE Trans. on Med. Imag., 1996; 15:327-332. [23]Taguchi K, Aradate H, Algorithm for image reconstruction in multi-slice
helical CT, Medical Physics, 1998; 25(4):550-561.
