Ⅰ. INTRODUCTION
Computed tomography (CT) images were obtained by measuring and reconstructing the attenuation values of the X-rays transmitted in the detector. The Hounsfield unit (HU) values obtained through CT images have become a reference value for radiation therapy because of the difference in tissue electron density, which can be applied to dose calculation in treatment planning[1,2]. The average energy of the transmitted beam increases when the X-ray penetrates the object, which affects the beam hardening effect and absorbs low photon energy[3]. Metal materials inserted into the body, such as dental fillings, spinal stabilization implants, and hip replacements, generate artifacts, form streak-type starbursts, and cause serious medical imaging errors with no diagnosis or use[4,5].
Conventional methods to reduce metal artifacts have been proposed to increase the tube voltage of reduced slice thickness in CT scans, but have the disadvantage of increasing the exposure dose to the patient. The metal artifact reduction (MAR) algorithm is a method for reducing metal artifacts, which obtains a corrected image through the iterative reconstruction process by reconstructing the sinogram of the CT raw data of attenuation information from the image detector and separating the sinogram of the metal material inserts from the original sinogram[6].
Representative MAR algorithms include metal artifact reduction for orthopedic implants (O-MAR) in Philips, single-energy projection for metallic artifact reduction in Toshiba, smart metal artifact reduction (SMART MAR) in GE, and iterative-metal artifact reduction (iMAR) in Siemens[7]. iMAR is a post-processing software for the syngo platform provided by Siemens that combines the normalized MAR (NMAR) and frequency split MAR algorithms (FSMAR) in an iterative loop[8-10]. The effect of metal artifacts using the iMAR algorithm can be reduced without increasing the radiation exposure dose.
In radiation therapy, factors such as CT numbers (i.e., HU values), contouring of the region of interest, and dose calculation are important for an accurate workflow of radiation treatment. CT images are distorted, and accurate contouring cannot be performed because of metal artifacts existing in the body[10]. Metal artifacts cause errors not only in the inaccuracy of anatomical information in delineating target or normal organs but also in unreliable electron density information in dose calculations[11,12]. Therefore, the accuracy of CT numbers must be evaluated along with the effects of dose calculation during radiation treatment planning[13,14]. External beam radiation therapy, which excludes electron beams, is mostly used in three-dimensional conformal radiation therapy (3D-CRT) and intensity-modulated radiation therapy (IMRT). Although many studies have assessed the effect of MAR in radiation therapy, limited studies have assessed special treatment units, such as Tomotherapy. Questions remain regarding the effect of the iMAR algorithm in Tomotherapy that can be treated with IMRT with gantry rotation and TomoDiect using a fixed gantry[15].
This study aimed to evaluate the effect of three metal inserts with high-density materials, aluminum, titanium, and steel, on CT numbers and radiation treatment plans for Tomotherapy.
Ⅱ. MATERIAL AND METHODS
1. Phantom and CT scan
A cylindrical TomoPhantom {Virtual Water phantom (Gammex RMI, Middleton, WI)} and CT simulator (SOMATOM Confidence, Siemens, Munich, Germany) were used to obtain CT images to measure the CT number and radiation treatment plans. As shown in Fig. 1, the phantom was 22 mm in diameter and 70 mm in length and contained a total of 20 plugs. Materials of various densities can be inserted into rods, replaced in the cylinder hole on the phantom, and replaced with other rods. Seven cylindrical rods represented various materials including brain (1.053 g/cm3), water (1.0 g/cm3), breast (0.983 g/cm3), solid water (1.019 g/cm3), air (0.001 g/cm3), liver (1.094 g/cm3), and cortical bone (1.821 g/cm3) as shown in Fig. 1. Five rods were replaced, and one (6 o’ clock) was removed to represent the air material, as shown in Fig. 1(b).
Fig. 1. (a) Computed tomography (CT) scanner with solid water phantom and metal insert rods on the couch and (b) a total of seven rods in a cylindrical solid water phantom.
Three materials, aluminum (Al: 2.7 g/cm3), titanium (Ti: 4.5 g/cm3), and steel (Fe: 7.8 g/cm3) were selected to verify the effect of high-density materials (metal inserts) which are custom-made with a 70-mm length and 20-mm hole diameter. Metal inserts were replaced with rods with a density of 1.0 g/cm3 in shown number 2 (yellow circle) of Fig. 2(a). The acquisition parameters were 120 kVp, 195 mAs, field of view of 500 mm, and slice thickness of 3 mm for all CT scans. A total of three CT image sets were acquired: reference CT images (without metal inserts) and two metal insert CT image sets to apply iMAR (uncorrected CT images=non-iMAR; corrected CT images=iMAR).
Fig. 2. (a) A reference computed tomography (CT) image wherein the numbers indicate various materials (1=brain, 2, 8=water, 3=solid water, 4=air, 5=cortical bone, 6=liver, 7=breast) and non-iterative-metal artifact reduction (non-iMAR) image of (b) Al, (c) Ti, and (d) Fe and iMAR image of (e) Al, (f) Ti, and (g) Fe, respectively (red circle).
2. CT number and standard deviation (SD)
Fig. 2 shows CT images of the reference materials (non-iMAR) and metal inserts (iMAR) with Al, Fe, and Ti materials according to the iMAR applied. To measure the mean value of the CT number and SD, a total of eight cylindrical regions of interest (ROI) with a 20-mm diameter within the 28-mm holes were delineated and calculated using MIM software (v6.6.14, MIM Software Inc., Cleveland, OH, USA), which was mostly used to contour the treatment target and normal tissues in the radiation treatment planning, as shown in Fig. 2 (a).
CT number and SD were obtained for each ROI as follows:
\(\begin{aligned}CTnumber=\frac{\mu_{\text {tissus }}-\mu_{\text {water }}}{\mu_{\text {water }}} \times 1000\end{aligned}\) (1)
where μtissue and μwater are the attenuation coefficients of measured tissue and water, respectively. The CT number is the HU value, which is a linear reduction coefficient of human tissue and is defined as the reference substance, water (HU:0), air (HU: -1000), bone (HU: +1000), and ranges from -1024 to +3071.
\(\begin{aligned}S D(\delta)=\sqrt{\frac{\sum_{1-I}^{N}(\Xi-X)^{2}}{N-1}}\end{aligned}\) (2)
where N, ∈, and X are the total number of pixels, the CT number value of pixels, and the CT number of each pixel, respectively. SD is defined as the CT number change between pixels constituting an image and pixels, which is an important factor in determining the quality of a CT image.
The percentage difference was calculated to evaluate the effect between the measured reference CT images (HUref) and two metal insert CT image sets (non-iMAR and iMAR) as HUmetal values as shown in Eq. (3).
\(\begin{aligned}\% C T=\frac{\left|H U_{r e f}-H U_{m e t a l}\right|}{H U_{r e f}}\end{aligned}\) (3)
3. Treatment planning (3D-CRT and IMRT)
The 3D-CRT and IMRT plans were generated for dosimetry influence through a comparison between non-iMAR and iMAR. Ti was selected as the metal insert to consider as a treatment for patients with cervical cancer, including hip prostheses. A total of six ROIs, including the planning target volume (PTV) as the target and the organ at risk (OARs: bowel, bladder, and rectum), are shown in Fig. 3(a). Metal implants (right and left) with Ti inserts were used for the treatment planning.
Fig. 3. (a) A treatment plan computed tomography (CT) image which number indicates ROIs [virtual target (number 5), normal organs (1=bowel, 2=bladder, and 6=rectum), and Ti metal inserts (number 3 and 4)]. three-dimensional conformal radiation therapy plan with four angles in (b) non-iterative-metal artifact reduction (non-iMAR) and (c) iMAR, and intensity-modulated radiation therapy plan in (d) non-iMAR and (e) iMAR applied, respectively.
All treatment plans were conducted using the Radiation Treatment System (RTP) in the Accuray Precision (ver.2.0.1.0, accuracy Inc., Sunnyvale, CA) program. The prescription dose was 50 Gy to the PTV of 95% volume in 25 fractions for all the plans. For the 3D-CRT plan, the irradiation angles were presented in four perpendicular directions (0°, 90°, 180°, and 270°), and a jaw width of 2.5 cm with a dynamic jaw mode, 0.25 pitch, compensation of low, and compensation resolution of low were used. The treatment plan parameters were a jaw width of 2.5 cm in dynamic jaw mode, a pitch of 0.287, a modulation factor of 2.4, and an optimization resolution of 100 times for the IMRT plan. The PTV coverage was calculated using parameters such as the percentage of coverage, homogeneity index (HI), and conformity index (CI) provided by the RTP system as follows:
The HI and CI were calculated for PTV.
\(\begin{aligned}H I=\frac{D_{95}}{D_{5}}\end{aligned}\) (4)
where HI is an indicator of homogeneity, and evaluates the uniformity of the delivered doses within the PTV. The smaller or 1 HI level indicates a more uniform dose distribution within the PTV.
\(\begin{aligned}C I=\frac{P T V_{95 \% P D}}{V_{P T V}}\end{aligned}\) (5)
where CI defines the proportion of PTVs receiving 95% of the prescription doses (PD) for PTV volume, indicating the conformity of the treatment plans. For the OAR doses of all plans, the mean OAR doses (bladder, rectum, bowel, and right and left metal implants), including the minimum and maximum doses, were calculated in the RTP system.
Ⅲ. RESULT
1. CT number and SD
1.1. Reference CT images
Table 1 shows that the mean CT number ± SD of the reference CT images were -975.2 ± 11.40 in air, -49.76 ± 13.14 in the breast, -10.27 ± 13.51 in water, -2.2 ± 13.51 in solid water, 20.53 ± 14.96 in the brain, 71.59 ± 13.34 in the liver, and 1,190 ± 18.23 in the bone.
Table 1. Mean CT number and SD for the seven materials
1.2. Comparison (reference vs. metal insert CT images)
Table 2 shows that the percentage deviations of mean CT numbers ranged from 0.2% (air) to 10.29% (breast) in Al, from 0.71% (liver) to 10.0% (solid water) in Ti, and from 1.91% (breast) to 17.27% (solid water) in Fe. In air, the Fe content was 3.09% higher than that of Al (0.2%) and Ti (0.9%). Fe was 8.96% higher in water than Al (0.39%) and Ti (1.75%). In the bone, Fe was 2.45% higher than Al (0.18%) and Ti (1.18). The percentage deviations for SD ranged from 1.7% (bone) to 13.34% (liver) for Al, 5.43% to 39.35% for Ti, and 13.97% (brain) to air (92.72%) for Fe.
Table 2. Percentage deviations of mean CT number and SD between reference and metal inserts CT images of three metal material inserts (Al, Ti, and Fe)
*Location of the water rod
In air, the Fe content was 92.72% higher than that of Al (3.6%) and Ti (22.72%). Fe was 51.59% higher in water than Al (8.52%) and Ti (39.35%). Fe was 26.33% higher in the bone than in Al (1.7%) and Ti (5.43%). Overall, Fe had the largest percentage deviation compared with the other samples.
1.3. Comparison (non-iMAR vs. iMAR)
The percentage deviations of mean CT numbers ranged from 0.0% (air) to 8.30% (breast) in Al, 3.26% (air) to 28.63% (water) in Fe, and from 1.19% (air) to 15.58% (water) in Ti. In air, the Fe content was 3.26% higher than that of 0% (Al) and 1.19% (Ti). In water, the Fe content was 28.63% higher than that of 0.31% (Al) and 15.58% (Ti). In the bone, Fe was 2.8% higher than that of 0.01% (Al) and 1.61% (Ti) as shown in Table 3.
Table 3. Percentage deviations of mean CT number and SD between non-iMAR and iMAR for three metal material inserts (Al, Ti, and Fe)
*Location of the water rod
The percentage deviations for SD ranged from 1.27% (brain) to 9.36% (breast) in Al, from 0.38% (breast) to 30.53% (air) in Fe, and from 1.67% (brain) to 15.45% (breast) in Ti. In air, the Fe content was 30.53% higher than that of 1.75% (Al) and 2.81% (Ti). In water, the Fe content was 26.79% higher than that of 8.44% (Al) and 6.74% (Ti). In the bone, the Al content was 8.12% higher than that of 2.41% (Fe) and 4.28% (Ti). Generally, Fe had the largest percentage deviation compared to the others, excluding Al in the SD as shown in Table 4.
Table 4. Comparison of CI and HI value of the PTV with Ti metal inserts along with used non-iMAR and iMAR for 3D-CRT and IMRT
3D-CRT, three-dimensional conformal radiation therapy; CT, computed tomography; HI, homogeneity index; iMAR, iterative-metal artifact reduction; PTV, planning target volume
2. Treatment plan evaluation (3D-CRT and IMRT)
2.1. Target coverage
The percentage deviations for SD ranged from 1.27% (brain) to 9.36% (breast) in Al, from 0.38% (breast) to 30.53% (air) in Fe, and from 1.67% (brain) to 15.45% (breast) in Ti. In air, the Fe content was 30.53% higher than that of 1.75% (Al) and 2.81% (Ti). In water, the Fe content was 26.79% higher than that of 8.44% (Al) and 6.74% (Ti). In the bone, the Al content was 8.12% higher than that of 2.41% (Fe) and 4.28% (Ti). Generally, Fe had the largest percentage deviation compared to the others, excluding Al in the SD as shown in Table 4.
2.2. OARs dose
The mean doses of OARs (difference percentage) in 3D-CRT were: bladder (0.28%), rectum (0.09%), bowel (0.41%), right metal implant (same), and left metal implant (0.03%). The range of difference for the minimum dose ranged from 0.26% (rectum) to 1.31% (right metal implant), and that of the maximum dose ranged from 0% (left metal implant) to 0.4% (bowel). The percentage differences in IMRT were: bladder (0.06%), rectum (0.10%), bowel (0.12%), right metal implant (0.68%), and left metal implant (0%).
The range of difference for the minimum dose ranged from 0% (bowel) to 0.31% (right metal implant), and that for the maximum dose was from 0.02% (bladder) to 0.23% (left metal implant) as shown in Table 5.
Table 5. Mean OAR dose between non-iMAR and iMAR with metal inserts in 3D-CRT and IMRT
Note: Min=minimum dose; Max=maximum dose.
3D-CRT, three-dimensional conformal radiation therapy; iMAR, iterative-metal artifact reduction; OAR, organ at risk
Ⅳ. DISCUSSION
Correction of beam hardening is a very important process for improving image quality and accurately analyzing the shape of the human body[3]. It is possible to overcome the current limitation of relying on clinical experience and predicting the contouring of normal and tumor tissues that have become unclear due to artifacts in radiation treatment plans[13]. Axente et al.[16] reported that the corrected image accurately implements the CT number regardless of the metal size and acquired energy, and can help contouring work. In their study, the mean difference in plastic water between the corrected and reference CT images was -1.3 HU in all experiments with a 90% confidence interval of HU. In our study, the mean difference of mean CT number (SD) in water between reference (without metal inserts) and metal inserts CT (corrected by iMAR) images were 0.1 HU (2.57 HU) in Al, 1.78 HU (2.0 HU) in Ti, and 2.02 HU (3.35 HU) in Fe, respectively. No significant difference was observed in the reference values between the iMAR CT image sets. On comparing each metal insert, the higher-density material generally had the largest difference in mean CT number and SD in our study.
The iMAR algorithm could not only improve the image quality in metal artifact-affected CT images for radiotherapy planning, but also reduce the uncertainty of radiation delivery dose uncertainty. First, metal artifacts affect the target volume definition process. Hagen et al.[17] compared the delineation of gross tumor volumes and OARs in the pelvic, head, and neck regions using CT with and without iMAR in phantom and clinical cases. They reported that the use of iMAR improves the anatomical delineation of prostate cancer patients with bilateral hip implants because of quantitatively improved image reconstruction. Second, a change in the CT number from metal artifacts can cause dose errors in radiotherapy planning. Bär et al.[18] investigated the effect of dose distribution for dental implants on IMRT plans and reported that comparing patient plans from corrected and uncorrected CT images revealed dose differences of up to ±5% in the target volume and OARs depending on the treatment site. März et al.[14] evaluated the dosimetry accuracy of iMAR for IMRT and volumetric modulated arc therapy plans and reported that iMAR maintained a dose tolerance within 3%. Therefore, in our study, the dose evaluation of the IMRT plan according to iMAR applied was not different for the evaluation indices such as the CI and HI values, including no significant difference for OARs dose as shown in Fig. 3. However, we verified that 3D-CRT has a difference in target coverage compared with corrected iMAR CT images in our study. The CI at iMAR was approximately 20.0% better than that at non-iMAR, including no significant difference in the uniformity of delivered doses (i.e., HI).
Another method to reduce metal artifacts is megavoltage computed tomography (MVCT)[19]. This solves the problem of inaccuracy in the dose calculation caused by the metal implant, but it is difficult to apply to actual treatment because the image quality deteriorates and the electron density is different. Lim et al.[7] evaluated a method for acquiring CT images using a dual-energy device to transmit different X-ray sources and selectively reconstruct the attenuated energy. This can overcome the disadvantage of image distortion when using the MAR algorithm; however, the image reconstruction time is long and the exposure dose is increased by approximately four times[6]. In the image-guided radiation therapy process, corrected cone-beam computed tomography and two-dimensional kV images using the iMAR algorithm are important for the accuracy of dose delivery to the treatment target with surrounding metal inserts. In Tomotherapy, the reference planning CT images also need to respect the process of image registration with MVCT images.
The limitations of our study include the phantom experiment without the inclusion of clinical cases, simple treatment planning in Tomotherapy, and no consideration to affect measurement accuracy, such as measurement location, image blurring, and metal insert size. These topics must be addressed in future studies.
Ⅴ. CONCLUSION
We verified the effects of image quality and dose evaluation using iMAR for 3D-CRT and IMRT plans in Tomotherapy. The CT number (SD) differences increased as the density increased, in accordance with the presence of metal inserts. In terms of the treatment plans, the target coverage differed between 3D-CRT and IMRT. In 3D-CRT, the CI at iMAR was approximately 20.0% better than that at non-iMAR, including uniformity of delivered doses to treatment target and was small difference in IMRT. There was small difference for the OAR dose between the non-iMAR and iMAR groups. Therefore, iMAR is helpful for target or OARs delineation and for reducing uncertainty for 3D-CRT in Tomotherapy.
References
- Y. K. Kim, S. Yang, T. U. Wang, "Evaluation of Artifact and Noise in the Standard and MAR Algorithms with Variation of Examination Conditions of CT", Journal of radiological science and technology, Vol. 43, No. 2, pp. 79-85, 2020. https://doi.org/10.17946/JRST.2020.43.2.79
- S. Mutic, J. R. Palta, E. K. Butker, I. J. Das, M. S. Huq, L. N. D. Loo, B. J. Salter, C. H. McCollough, J. V. Dyk, "Quality assurance for computed-tomography simulators and the computed-tomography-simulation process: report of the AAPM Radiation Therapy Committee Task Group No. 66", Medical Physics, Vol. 30, No. 10, pp. 2762-2792, 2003. https://doi.org/10.1118/1.1609271
- G. Van Gompel, K. Van Slambrouck, M. Defrise, K. J. Batenburg, J. De Mey, J. Sijbers, "Iterative Correction of Beam Hardening Artifacts in CT", Medical physics, Vol. 38, No. 1, pp. 36-49, 2011. http://dx.doi.org/10.1118/1.3577758
- F. E. Boas, D. Fleischmann, "CT Artifacts: Causes and Reduction Techniques", Imaging Med, Vol. 4, No. 2, pp. 229-240, 2012. http://dx.doi.org/10.2217/iim.12.13
- D. Giantsoudi, B. D. Man, J. Verburg, A. Trofimov, Y. Jin, G. Wang, L. Gjesteby, H. Paganetti, "Metal artifacts in computed tomography for radiation therapy planning: dosimetric effects and impact of metal artifact reduction", Physics in Medicine & Biology, Vol. 62, No. 8, pp. 49-80, 2017. http://dx.doi.org/10.1088/1361-6560/aa5293
- K. S. Park, W. J. Choi, D. H. Kim, "A Study of the Metal Artifact Reduction using Dual Energy CT Clinical Applications of Dual Energy and MAR Algorithm", Journal of the Korean Society of Radiology, Vol. 15, No. 3, pp. 273-279, 2021. https://doi.org/10.7742/jksr.2021.15.3.273
- E. Meyer, R. Raupach, M. Lell, B. Schmidt, M. Kachelriess, "Frequency Split Metal Artifact Reduction (FSMAR) in Computed Tomography", Medical physics, Vol. 39, No. 4, pp. 1904-1916, 2012. http://dx.doi.org/10.1118/1.3691902
- P. Lim, J. Barber, J. Sykes, "Evaluation of Dual Energy CT and Iterative Metal Artefact Reduction (iMAR) for Artifact Reduction in Radiation Therapy", Australasian Physical & Engineering Sciences in Medicine, Vol. 42, No. 4, pp. 1025-1032, 2019. http://dx.doi.org/10.1007/s13246-019-00801-1
- E. Meyer, R Raupach, M. Lell, B. Schmidt, M. Kachelriess, "Normalized Metal Artifact Reduction (NMAR) in Computed Tomography", Medical Physics, Vol. 37, No. 10, pp. 5482-5493, 2010. http://dx.doi.org/10.1118/1.3484090
- Y. G. Han, Y. J. Jang, S. Y. Kim, D. H. Kang, D. H. Lee, "Evaluation of Usefulness of Iterative Metal Artifact Rduction (iMAR) Algorithm in Proton Therapy Planning", The Journal of Korean Society for Radiation Therapy, Vol. 29, No. 1, pp. 49-56, 2017.
- L. Bedos, N. Ailleres N, D. Azria, P. Fenoglietto, "CT Number Accuracy Assessment of a New Metal Artifact Reduction Algorithm for CT Simulations in Radiation Therapy", International Journal of Radiation Oncology, Biology, Vol. 90, No. 1, pp, 865-866, 2014. http://dx.doi.org/10.1016/j.ijrobp.2014.05.2473
- G. Hilgers, T. Nuver, A. Minken, "The CT number accuracy of a novel commercial metal artifact reduction algorithm for large orthopedic implants", Journal of Applied Clinical Medical Physics, Vol. 15, No. 1, pp. 274-278, 2014. http://dx.doi.org/10.1120/jacmp.v15i1.4597
- J. R. Park, M. S. Kim, J. M. Kim, H. S. Chung, C. H. Lee, "A Study of Usefulness of Clinical Application of Metal Artifact Reduction Algorithm in Radiotherapy", The Journal of Korean Society for Radiation Therapy, Vol. 29, No. 1, pp. 9-17, 2017.
- M. Maerz, P. Mittermair, A. Krauss, O. Koelbl, B. Dobler, "Iterative Metal Artifact Reduction Improves Dose Calculation Accuracy", Strahlentherapie und Onkologie, Vol. 192, No. 6, pp. 403-413, 2016. http://dx.doi.org/10.1007/s00066-016-0958-z
- P. Franco, P. Catuzzo, D. Cante, M. R. La Porta, P. Sciacero, G Girelli, "TomoDirect: an Efficient Means to Deliver Radiation at Static Angles with Tomotherapy", Tumori Journal, Vol. 97, No. 4, pp 498-502, 2011. http://dx.doi.org/10.1700/950.10404
- M. Axente, A. Paidi, R. Von Eyben, C. Zeng, A. Bani-Hashemi, A. Krauss, "Clinical Evaluation of the Iterative Metal Artifact Reduction Algorithm for CT Simulation in Radiotherapy", Medical physics, Vol. 42, No. 3, pp. 1170-1183, 2015. http://dx.doi.org/10.1118/1.4906245
- M. Hagen, M. Kretschmer, F. Wurschmidt, T. Gauer, C. Giro, E. Karsten, "Clinical Relevance of Metal Artefact Reduction in Computed Tomography (iMAR) in the Pelvic and Head and Neck region: Multi institutional Contouring Study of Gross Tumor Volumes and Organs at Risk on Clinical Cases", Journal of Medical Imaging and Radiation Oncology, Vol. 63, No. 6, pp. 842-851, 2019. http://dx.doi.org/10.1111/1754-9485.12924
- E. Bar, A. Schwahofer, S. Kuchenbecker, P. Haring, "Improving Radiotherapy Planning in Patients with Metallic Implants using the Iterative Metal Artifact Reduction (iMAR) Algorithm", Biomedical Physics & Engineering Express, Vol. 1, No. 2, pp. 025206, 2015. http://dx.doi.org/10.1088/2057-1976/1/2/025206
- M. R. Paudel, M. Mackenzie, B. G. Fallone, S. Rathee, "Evaluation of Metal Artifacts in MVCT Systems using a Model Based Correction Method", Medical Physics, Vol. 39, No. 10, pp. 6297-6308, 2012. http://dx.doi.org/10.1118/1.4754647