IMAGE RECONSTRUCTION AND THE EFFECT ON DOSE CALCULATION FOR HIP PROSTHESES

Transcription

1 Medical Dosimetry, Vol. 28, No. 2, pp , 2003 Copyright 2003 American Association of Medical Dosimetrists Printed in the USA. All rights reserved /03/$ see front matter doi: /s (02) IMAGE RECONSTRUCTION AND THE EFFECT ON DOSE CALCULATION FOR HIP PROSTHESES PAUL J. KEALL, LEAH B. CHOCK, ROBERT JERAJ, JEFFREY V. SIEBERS, and RADHE MOHAN Department of Radiation Oncology, Medical College of Virginia, Richmond, VA; Joz ef Stefan Institute, University of Ljubljana, Slovenia; and University of Wisconsin, Madison, WI ( Accepted 12 June 2003) Abstract High atomic number inserts, such as hip prostheses and dental fillings, cause streak artifacts on computed tomography (CT) images when filtered back-projection (FBP) methods are used. These streak artifacts severely degrade our ability to differentiate the tumor volume. Also, incorrect Hounsfield numbers yield incorrect electron density information that may lead to erroneous dose calculations, and, as a result, compromise clinical outcomes. The aim of this research was to evaluate the dosimetric consequences of artifacts during radiotherapy planning of a prostate patient containing a hip prosthesis. The CT numbers corresponding to an iron prosthesis were inserted into the right femoral head of an existing CT image set. This artifact-free image was used as the standard image set. CT projections through the image set formed the sinogram, from which filtered back projection and iterative deblurring methods were used to create reconstructed image sets. These reconstructed image sets contained artifacts. Prostate treatment plans were then calculated using a Monte Carlo system for the standard and reconstructed CT image sets. Close to the prosthesis, the CT numbers between the reconstructed and standard image sets differed substantially. However, because the CT number differences covered only a small area, the dose distributions on the reconstructed and standard image sets were not significantly different. The dose-volume histograms for the prostate, rectum, and bladder were virtually identical. Our results indicate that even though CT image artifacts restrict our ability to differentiate tumors and critical structures, the dose distributions for a prostate plan containing a hip prosthesis, calculated on both artifact-free image sets and image sets containing artifacts, are not significantly different American Association of Medical Dosimetrists. Key Words: Image reconstruction, Dose calculation, Hip prosthesis, Monte Carlo. Reprint requests to: Paul J. Keall, Ph.D., Department of Radiation Oncology, PO Box , Medical College of Virginia, Richmond, VA pjkeall@vcu.edu INTRODUCTION The standard method of computed tomography (CT) image reconstruction, filtered back-projection (FBP), gives characteristic streak artifacts due to aliasing when attempting to create images from patients containing high atomic number prostheses. Streak artifacts reduce the diagnostic information needed for tumor delineation and introduce incorrect density values used for the dose calculation. Streak artifacts are most common for headand-neck patients, due to fillings, and patients with artificial hip prostheses. Additionally, surgical clips and brachytherapy seeds can also cause artifacts. Several attempts to overcome this limitation have been made in the past. 1 4 However, these methods, though promising, have yet to be incorporated into clinical practice by CT manufacturers. Wang et al. 1 were able to obtain images nearly free of streaking by using the iterative deblurring (IDB) technique for phantom studies. For patient geometries, however, the method was not as successful, particularly in the case of complex phantoms. This difference could be related to the sharp contrast between the structures that were being reconstructed. According to Wang, the performance of iterative deblurring depends on the contrast of structures being reconstructed (G. Wang, private communication, 1999). A recently proposed multidimensional adaptive filtering 5 method shows potential for widespread application. The aim of this research was both to investigate improved methods of image reconstruction and to determine the effect of the image reconstruction artifacts on the dose distribution for patients with hip prostheses undergoing pelvic radiotherapy. METHOD Image reconstruction Due to the CT image reconstruction limitations for high density and atomic number prosthesis materials described above, it was not possible to use an artifactfree CT image set of a patient with a hip prosthesis. Thus, a 28-mm-diameter sphere with a CT number corresponding to a density of 8.0 g cm 3, representing a typical hip prosthesis, 6 was inserted into the right femoral head of an existing CT image set of a patient. Hip prostheses generally consist of cobalt-chrome alloy, stainless steel, or titanium. The calculations performed here assumed that the prosthesis was iron (stainless steel). The calculations with the iron approximate those of cobalt-chrome pros- 113

2 114 Medical Dosimetry Volume 28, Number 2, 2003 Fig. 1. A schematic diagram showing how the CT datasets were obtained. Fig. 2. CT images of (a) the original geometry with the prosthesis inserted in the right femoral head, (b) the FBP reconstruction of the original CT with the prosthesis, and (c) the IDB reconstruction of the original CT with the prosthesis. theses, because of their comparable densities and effective atomic numbers, and yield similar dosimetric results. Because iron has a higher atomic number and density than that of titanium, both the streak artifacts and dose differences will be greater for iron than for titanium. The CT image set with the added prosthesis was used as the standard for the additional reconstructions; an ideal image reconstruction algorithm would reproduce these image sets. The reconstructed geometries were obtained by taking the standard image set (on a slice-byslice basis) to calculate the image sinogram representing the detector signals at various angles taken during a CT scan. The image reconstruction from this sinogram formed the reconstructed geometries shown in Fig. 1. Two image reconstruction methods, FBP and IDB, are presented here. 1,3 In the calculation of the image sinogram, a mono-energetic photon beam with no photon scatter was assumed, with a perfect detector (noisefree). An actual CT image reconstruction of the geometry may exhibit different artifacts from this idealized case. Commercial CT reconstruction is almost exclusively performed via the direct solution of a matrix equation using the image sinogram, or signal from the detectors at each imaging angle. 7 This process is called FBP. A good description of this method is given by Webb. 8 In the iterative deblurring method, an initial guess image is chosen. Arrays store information about which rays pass through metal, so that the sinogram data affected by metal is ignored. A weighting function compensates for the variable number of nonmetal rays that may pass through each pixel. For each iteration, a sinogram is calculated from the guess image and compared with the actual sinogram via division. The quotients are used to construct a correction factor for each pixel, which is multiplied into the guess image to form the next guess. The iteration process is stopped when further iterations produce little change in the reconstructed image. Another image reconstruction method implemented

3 Image reconstruction and dose calculation P.J. KEALL et al. 115 Fig. 3. Differences in CT number between the reconstructed and standard images as a function of position for a profile through the center of the prosthesis. The prosthesis region is shaded. was the rubout method of Glover and Pelc. 4 This method attempts to subtract average metal effects from the sinogram while retaining small variations in that region. These small ripples should result from nonmetal structures in the phantom. Glover and Pelc successfully applied the rubout method to remedy artifacts caused by small metal surgical clips. Although the rubout method produced images with fewer artifacts than those obtained through FBP, due to the larger size of the prosthesis relative to surgical clips, streaks were produced. Because the rubout method produced images inferior to those obtained through IDB, they will not be shown here. Fig. 5. Dose as a function of position for a profile through the center of the prosthesis. The prosthesis region is shaded. Dose calculation Once the CT image set is obtained, a method is needed to transport particles accurately and to take into account the density 2 and composition of the prosthesis. Currently used commercial algorithms, both correctionbased and model-based, will not accurately predict the dose near high atomic number inserts. However, the Monte Carlo method 9 11 is well suited for this, because individual particles are transported through a medium, from which relevant interaction cross-sections can be taken. The Monte Carlo code EGS4 12 was used with usercodes BEAM 13 and DOSXYZ 14 and was interfaced with our treatment planning computer using the MCV system. 15 The energy cutoffs were AE ECUT 700 kev and AP PCUT 10 kev. The voxel size for the calculations was cm 3, which is the standard voxel size used for treatment planning at our institution. The dose calculations used an existing treatment plan for the patient. The analysis of the dose distributions obtained for the different image sets was performed via isodose distribution comparison and dose-volume histogram (DVH) comparison. RESULTS Fig. 4. Isodose plans for the treatment calculated on (a) the patient geometry with the prosthesis, (b) the FBP reconstructed geometry, and (c) the IDB reconstructed geometry. The 40-, 60-, and 70-Gy isodose lines are shown. Image reconstruction CT images of the standard image set and the FBP and IDB geometries are shown in Fig. 2. The characteristic streaking artifacts of the FBP geometry can be observed. The IDB image shows improvement for some parts of the image; however, artifacts remain. Note that these artifacts originate from the edge of the prosthesis to the edges of the structures with contrast, such as the rectum. Such artifacts were not seen for phantom studies, like those performed by Robertson et al. 3 with less complex images. The differences in CT numbers between the reconstructed and standard geometries through the center of the hip prosthesis are shown in Fig. 3. Differences in the

4 116 Medical Dosimetry Volume 28, Number 2, 2003 The isodose patterns on the different images are very similar. The dose profile through the center of the prosthesis is shown in Fig. 5. This figure shows that differences in dose calculated on the different image sets are within the statistical uncertainty of the Monte Carlo calculations. Here we may face another problem, as some treatment planning system manufacturers specify a maximum value for density considerably lower than that of titanium (4.54 g cm 3 ), and much lower than that of steel or gold. DVHs for the prostate, rectum, and bladder calculated on the 3 image sets are shown in Fig. 6. This figure shows indistinguishable differences between the DVHs calculated on the different image sets. To investigate effects seen with a different algorithm, DVHs were also calculated using a pencil beam algorithm on the 3 image sets. As with the Monte Carlo results, negligible differences were observed between the DVHs calculated on the different image sets. CONCLUSIONS Different reconstruction algorithms were used to obtain the CT image set for a prostate cancer radiotherapy patient with a hip prosthesis. None of the reconstruction methods investigated yielded an artifact-free image set. The artifacts in the reconstructed images degraded the ability to delineate tumor and critical structures in the images. Further research into image reconstruction methods for patients with high density and atomic number inserts is needed. Artifacts also yield erroneous density information on the image set, affecting radiation transport and dose calculations. However, the magnitude of these artifacts was not sufficient to perturb the radiation dose for the calculations performed. It should be noted that even though no significant differences were observed for the calculations studied here, one cannot generalize these results for other patient data and other high-z implants (such as dental fillings), where significant differences may exist. Fig. 6. Dose-volume histograms for (a) the prostate, (b) the rectum, and (c) the bladder as calculated on the actual geometry, FBP geometry, and the IDB geometry. Acknowledgments The authors acknowledge the financial support of NIH Grant CA and the Grant-in-Aid Program for Faculty of Virginia Commonwealth University. The authors thank Dr. Bruce Libby for assistance with the phase space calculations, and the radiation physics group at MCV for their helpful advice. The authors thank Devon Farnsworth for revising this manuscript. REFERENCES CT numbers of up to 400 Hounsfield units are observed for both the FBP and the IDB geometries. Dose calculation The isodose distribution for the same treatment plan calculated on the different image sets is shown in Fig Wang, G.; Snyder, D.L.; O Sullivan, A.O.; Vannier, M.W. Iterative deblurring for CT metal artifact removal. IEEE Trans. Med. Imaging 15:657 64; Morin, R.; Raeside, D.E. Removal of streaking artifact in computed tomography. J. Med. Syst. 6:387 97; Robertson, D.D.; Yuan, J.; Wang, G.; Vannier, M.W. Total hip prosthesis metal-artifact suppression using iterative deblurring reconstruction. J. Comp. Assis. Tomogr. 21:293 98; 1997.

5 Image reconstruction and dose calculation P.J. KEALL et al Glover, G.H.; Pelc, N.J. An algorithm for the reduction of metal clip artifacts in CT reconstructions. Med. Phys. 8: ; Kachelriess, M.; Watzke, O.; Kalender, W.A. Generalized multidimensional adaptive filtering for conventional and spiral singleslice, multi-slice, and cone-beam CT. Med. Phys. 28:475 90; Yamamuro, T. Recent advances in clinical use of artificial joint of the hip and knee. In: Rubin, L.R., editor. Biomaterials in Reconstructive Surgery. St. Louis: Mosby; Sprawls, P. Physical Priciples of Medical Imaging, 2nd ed. Madison, WI: Medical Physics Publishing; Webb, S. The Physics of Medical Imaging. Bristol: IOP Publishing; Raeside, D.E. Monte Carlo principles and applications. Phys. Med. Biol. 21:181 97; Andreo, P. Monte Carlo techniques in medical radiation physics. Phys. Med. Biol. 36: ; Jenkins, T.M.; Nelson, W.R.; Rindi, A. Monte Carlo Transport of Electrons and Photons. New York: Plenum; Nelson, W.R.; Hirayama, H.; Rogers, D.W.O. The EGS4 Code System. SLAC-265, Stanford Linear Accelorator Center; Rogers, D.W.O.; Faddegon, B.A.; Ding, G.X.; et al. BEAM: A Monte Carlo code to simulate radiotherapy units. Med. Phys. 22:503 24; Ma, C.-M.; Reckwerdt, M.; Holmes, M.; et al. DOSXYZ Users Manual. PIRS-0509b, National Research Council of Canada; Siebers, J.V.; Keall, P.J.; Kim, J.O.; Mohan, R. Performance Benchmarks of the MCV Monte Carlo System, Presented at the XIII International Conference on the Use of Computers in Radiation Therapy, Heidelberg, Germany; 2000.