Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Systematisation of spatial uncertainties for comparison between a MR and a CT-based radiotherapy workflow for prostate treatments

Systematisation of spatial uncertainties for comparison between a MR and a CT-based radiotherapy... Background: In the present work we compared the spatial uncertainties associated with a MR- based workflow for external radiotherapy of prostate cancer to a standard CT-based workflow. The MR-based workflow relies on target definition and patient positioning based on MR imaging. A solution for patient transport between the MR scanner and the treatment units has been developed. For the CT-based workflow, the target is defined on a MR series but then transferred to a CT study through image registration before treatment planning, and a patient positioning using portal imaging and fiducial markers. Methods: An "open bore" 1.5T MRI scanner, Siemens Espree, has been installed in the radiotherapy department in near proximity to a treatment unit to enable patient transport between the two installations, and hence use the MRI for patient positioning. The spatial uncertainty caused by the transport was added to the uncertainty originating from the target definition process, estimated through a review of the scientific literature. The uncertainty in the CT-based workflow was estimated through a literature review. Results: The systematic uncertainties, affecting all treatment fractions, are reduced from 3-4 mm (1Sd) with a CT based workflow to 2-3 mm with a MR based workflow. The main contributing factor to this improvement is the exclusion of registration between MR and CT in the planning phase of the treatment. Conclusion: Treatment planning directly on MR images reduce the spatial uncertainty for prostate treatments. apy [6-12]. An "open bore" 1.5T MRI, has been installed Background MR images are well suited for target delineation, not only in direct connection to a treatment unit at the radiother- for the prostate [1], but also for many other tumours, such apy department in Umeå [13]. This installation allows us as brain lesions [2,3] and head and neck tumours [4,5], to image most of our patients in treatment position with which explains the growing interest for MR in radiother- the MR for the target delineation, and open the door for Page 1 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 development of an online treatment setup workflow flows are based on both a literature review and the results designed for soft tissue tumours. Figure 1 illustrates a MR- of our own experiments. only workflow and a more conventional CT-based work- flow. In the MR-based workflow, the target definition, the Methods treatment planning, and patient positioning at treatment In order to assess the total spatial uncertainties in the two delivery, are performed with MR aid only. The patient workflows, shown in figure 1, the workflow processes positioning utilize a transport trolley to move the patient were broken down into independent sub processes. Both from the imaging in the MR to the treatment table. A very workflows contain two main steps where uncertainties robust fixation of the patient provides control over the can be introduced, target definition for treatment plan- relation among the coordinate systems in the patient, in ning and patient positioning at treatment delivery. Our the MR, and in the treatment room. However, the trans- tools in the uncertainty analysis have been literature port does introduce uncertainties, which must be reviews, and when necessary own experiments. The own accounted for in an evaluation of the workflow and the experiments concern positioning with MRI, and are resulting geometric uncertainties. described in the section about MR guided delivery. An alternative workflow could be to plan on MR material An open-bore MRI scanner (Siemens Espree, 1.5T) was followed by positioning based on fiducial markers. This used for the MR imaging of the patients in connection intermediate workflow requires that the internal markers radiotherapy. For prostate patients, a T2-weighted SPACE are visible on the MR images and that the apparent marker sequence (Siemens), which is a 3D turbo spin-echo positions are correct. Parker et al. [14] shows that internal sequence with varying flip angle on the refocusing pulses, markers appear clearly on gradient echo sequences, while was used. The slice thickness was 1.7 mm, typical pixel- more difficulty to identify on T2-weighted turbo spin echo size was 1.0 × 1.0 mm , and the bandwidth was 592 Hz sequences. The visibility of the markers was increased per pixel. Distortions caused by gradient non-linearity when the TE time was reduced, giving higher signal but were corrected with an algorithm based on spherical har- compromising the T2-weighted contrast. Verified robust monic expansion of the fields generated by the gradient imaging of fiducial markers in MR would enable also this coils [15]. The 3D correction algorithm including repre- workflow. In the present study, this intermediate work- sentation of the coils was delivered by Siemens as a stand- flow will not be explicitly handled. ard clinical tool integrated in the scanner software (VB15). The scanner was set in an isocentric mode, which moves The purpose of this study is to investigate if a MR-only the table prior to the acquisition of each sequence, to radiotherapy workflow, in accordance with figure 1b, has place the MR isocenter in the centre of the volume of the potential to improve the spatial accuracy compared to interest. the more conventional CT-based workflow (figure 1a). The estimations of the uncertainties in the different work- The total spatial uncertainty consists of both a random part, varying in direction and magnitude from fraction to fraction, and a systematic part, which is invariant over the treatment period. The systematic and random uncertainty should be given different weight in the formation of mar- gins between the CTV and the PTV. In the present work we used the weight factor 2.5 for the systematic errors and 0.7 for random errors as proposed by van Herk et. al. [16,17]. The PTV margin is hence expressed as (1) m =∑ 25..+07s PTV where Σ is the systematic and σ is the random spatial Overview of th study Figure 1 e two workflows analyzed in the present uncertainty. The presented uncertainties are throughout Overview of the two workflows analyzed in the this paper presented in units of one standard deviation present study. (a) A widely used workflow utilizing regis- (1SD), thus inherently assuming normal distributed data. tration between MR and CT images in order to transfer the delineated prostate volume (GTV/CTV) from the MR study Uncertainty in target definition to the CT study. The CT study is used for treatment planning The total uncertainty in the target definition can be bro- and to generate DRR's for patient position. Typically, fiducial ken down to three subparts: uncertainty in prostate delin- markers are used. (b) The workflow is entirely based on MR, eation (MR-based on both workflows), spatial distortion both for planning and positioning. in MR images that can be scanner related and patient Page 2 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 induced, and for the CT-based workflow: uncertainty in mm. This uncertainty is approximately equal in all direc- registration between CT and MR images. tions provided that a 3D correction algorithm is used. Uncertainty in prostate delineation Registration uncertainties - MR/CT Rasch et al [18] has from a study with 18 patient analysed The workflow in figure 1a involves a registration between by 3 physicians reported an uncertainty, in the prostate a CT and MR study. Errors in this registration directly delineation on axial MR study, of 2 mm at the base of affect the spatial accuracy of the target definition. Registra- seminal vesicles and up 2.8 mm in the prostate apex. The tions between MR and CT for prostate patients can be per- uncertainty in the head-feat (HF) direction was 2.5 mm formed based on fiducial markers [14]. The trend is, with a slice thickness of 5-6 mm for the axial MR images. however, to use mutual information (MI) registration In a later study involving 7 physicians analysing 10 based directly on the patient anatomy [27,28]. The pros- patients Smith et al. [19] reported a radial uncertainty of tate position relative other anatomical structures is not fix, 0.6 - 1.6 mm for the delineation of the prostate where the therefore the registration should ideally be based on the larger value is for the apex. The inter-observer uncertainty prostate with just a small margin. However, this has been in the length (HF direction) of the prostate was 3.4 mm, reported problematic because of too limited morphologi- and the intra-observer variation was 2.6 mm; the slice cal information content in the CT representation of the thickness was 2.5 mm. prostate [29,30]. A few studies have been performed eval- uating the accuracy and precision of MI registration for CT In summary, the literature review indicates a prostate and MR studies of the prostate; the registration uncer- delineation uncertainty of 1.8 mm in the right-left (RL) tainty has been reported to be around 2 mm [29,31]. Rob- and anterior-posterior (AP) directions and 2.8 mm in the erson et al. [31] reported that registration results depend HF direction. on the starting point for a specific MI optimization soft- ware. The mean difference between different stating Geometrical Distortions in MR points was up to 1 mm in the RL direction. The corre- Geometrical distortions in MR images are a well known sponding number for MR-MR registration was 0.4 mm in phenomena [20-22]. In modern MR scanners, gradient the HF direction which could indicate that the mutual non-linearity is the main cause of image distortions [20], information maximum is more distinct for MR-MR regis- dominating over the effect of static field inhomogenity. tration compared to CT-MR registration. The distortions introduced by the gradient non-linearities are increasing with the distance from the MR isocenter In summary, the registration uncertainty for a CT - MR reg- [20,23] Without correction, the geometrical distortions in istration for a prostate case was estimated to be 2 mm modern MR scanners can cause deviations between phys- based on current reports in the scientific literature. ical and imaged distances of up to 20% in extreme situa- tions. However, there are methods for distortion Uncertainty in patient positioning correction which reduces the errors significantly. It is pos- The patient positioning at treatment, with the develop- sible to use a specially designed geometry phantom to ment of image guided radiotherapy, been in focus the characterize and correct the distortions for a specific scan- recent years. For prostate cancer patients the improve- ner [22,24]. In the present work, a gradient coil specific ments in spatial treatment accuracy has been considera- distortion correction algorithm was applied. Even though ble. Both the CT and the MR-based workflows, shown in this device specific corrections only correct for intrinsic figure 1, rely on imaging before each fraction. Intra-frac- gradient non-linearity connected to a specific type of scan- tion motion of the prostate is therefore an issue for both ner/gradient coil, it has been shown that this kind of cor- workflows. rection yields a spatial accuracy better than 2% [23,25], which is sufficient when region of interest in the patient is Intra-fraction prostate motion close to the MR isocenter. Patient anatomy, e.g. air pockets In a large investigation by Kotte et al [32] intra fraction in the rectal cavity, can generate susceptibility-generated motion larger than 2 mm was observed during 66% of the field changes up to ± 10 ppm [26]. With a bandwidth of fractions, this number is roughly in agreement with the 592 Hz per pixel this corresponds to distortions up to results presented in other studies [33,34]. However, approximately 1 pixel for a 1.5T scanner. Thus, magnetic reduction of the rectal filling has been showed to be of susceptibility related distortions are a minor effect for the great importance to achieve a stable prostate position sequence used. [33,35]; an uncertainty of 2 mm is therefore realistic for a 5-7 min treatment when patients are instructed to empty In summary, for a prostate with radius of 2.5 cm the geo- rectum prior to treatment. The position uncertainty due to metrical distortions can cause errors of up to 0.5 mm, prostate motion is most pronounced in the AP and HF which corresponds to a standard deviation of around 0.2 directions [32,36]. Page 3 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 In summary, the overall uncertainty for the prostate posi- patient is fixated on a shell, with a double vacuum system tion was estimated to 2 mm, which broken down in the (BodyFIX, Medical Intelligence an Elekta company), orthogonal directions corresponds to: 1.4 mm in AP and which can be slid from the trolley to the treatment or MR HF, and 0.4 mm in RL. table after docking. The shell has fixed positions both at the MR and the treatment table, which enables absolute Uncertainty with fiducial markers coordinate transformation between MR coordinates and There are numerous studies on the accuracy of patient treatment coordinates. The treatment table is a Siemens positioning using fiducial markers and portal or flat 550 TxT equipped with a modified TT-D table-top com- screen kV images. Several different sources of uncertainty patible with the Miyabi transport solution. The daily treat- need to be considered in order to correctly estimate the ment table coordinates are calculated as the absolute table overall accuracy of the workflow. Random positioning coordinates from the treatment planning corrected for errors are partly due to uncertainty in the registration daily variations in patient and prostate position. The daily between the reference image and the portal/kV flat screen correction is calculated based on a sub-volume-based image. Literature indicates that a manual registration typ- rigid mutual information registration between the refer- ically results in uncertainty of around 0.7 mm in the HF ence MR images used at treatment planning and daily and RL direction, and 1.4 mm in the AP direction [37,38]. positioning MR images. The same SPACE sequence was used both for treatment planning and for daily position- An investigation by Nichol et al [39] indicates that a sys- ing. Calibration of the system, i.e. determination of the tematic deformation of the prostate during radiotherapy absolute coordinate transformation vector, is an obvious leads to drift in the relation between the centre of mass for source for systematic uncertainty, while mechanical insta- the markers and centre of mass for the contoured prostate. bilities in the mounting mechanism at the MR and treat- This uncertainty is in the order of 1 mm, which is roughly ment table together with image distortion, image in agreement with other reports [40,41]. It should be registration errors and patient movement during transport noted that deformation of the prostate is in many respects mainly result in uncertainty of random nature. equivalent to marker migration within the prostate. These two effects are therefore not separated in the present work. Uncertainty in calibration vector determination The calibration vector is the relation between the coordi- Prostate deformation and marker migration are resulting in a systematic uncertainty in the patient position. nate for a specific point, in the MR coordinate system and the treatment table coordinates that brings the same point The uncertainty of clinical imaging systems are in the to the treatment isocenter. The calibration vector was order of 1 mm, accounting for limitations in resolution, determined using a phantom which is sketched in figure isocenter position and mechanical instability. Paulsen et 3. The centre point of the phantom is clearly visible on al [34] observed a systematic discrepancy of almost 1 mm MR, CT, portal images and can also be positioned using when comparing 2 different imaging modalities at 2 dif- lasers. We placed the phantom at various positions on the ferent accelerators. Kotte et al. [32] detected that the sag of Miyabi shell and carefully determined the position of the the gantry caused a systematic imaging deviations of centre point in both the MR coordinate system using MR almost 1 mm in the HF direction when the gantry was in images, and the treatment coordinate system using cali- 0 degree position compared to 180 degree position. brated lasers. The calibration vector was calculated, for each phantom position on the Miyabi shell, as the differ- In summary, it is estimated that the uncertainty in the day ence between the MR coordinates and the treatment table to day registrations between reference image and the por- coordinates for the central point in the phantom. The idea tal image is 0.7 mm in RL and HF direction and 1.4 mm with repeated measurements was to assess the precision of in AP direction. The estimated uncertainty for the marker the vector determination taking intrinsic inhomogeneities position in the prostate is 1 mm in all directions, and the in the magnetic field and position dependent distortions estimated total uncertainty for the imaging systems is 1 into account. In total 16 independent determinations of mm in all directions. the calibration vector was performed, for different phan- tom positions on the Miyabi shell. The measurements MR guided treatment delivery were performed with the phantom centre positioned at ± The MR positioning approach is novel; we therefore 25 mm in the AP direction, and ± 60 mm in the RL direc- describe the principals in detail below, as well as the tion and at 4 different positions along the HF direction experiments performed to estimate the uncertainties con- with a total span of 450 mm. The scanning of the phan- nected to the method. tom was performed in isocentric mode. Figure 2 shows the hardware configuration. The patient is Weight correction transported between the MR scanner and the treatment The calibration vector needs to be corrected based on the unit on a MR compatible trolley (Miyabi, TRUMPF). The patient's weight to account for the treatment table sag- Page 4 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 Schematic overview of the hardware config Figure 2 uration for the MR positioning of patients Schematic overview of the hardware configuration for the MR positioning of patients. There is a direct connection between the MR room and treatment room, which makes patient transport quick and simple. In parallel with the patient trans- port the treatment couch coordinates are calculated using dedicated image registration software, the transport in it self does therefore not prolong the procedure. ging. The magnitude of the sagging was investigated using Patient movement a set of 15 kg bricks which were distributed to approxi- Significant patient movements during the time interval mate the weight distribution of a typical patient. We var- from the imaging to the treatment are deemed highly ied the total load and the weight distribution on the table unlikely when using the double vacuum immobilization top, to simulate patient weight from 0 to 105 kg, and device. There is however a risk for prostate movements patient height from approximately 150 cm to 190 cm. within the body during this time interval as discussed above (see section about intra-fraction prostate move- Geometrical distortions ment) The prostate is typically located on the patient's central line and with the Miyabi shell together with the BodyFIX Position reproducibility vacuum pillow the height of the prostate for the typical The reproducibility of the Miyabi shell position on the MR patient will be very close to the isocenter. The internal MR and treatment table were investigated through measure- laser is used to position the patient in the HF direction ment of the maximum shell displacement under direct before imaging, thus the prostate will be close to the iso- force in different directions center also in the HF direction. If the prostate centre is Registration uncertainties MR/MR within a sphere of 5 cm around the MR isocenter and the maximum spatial distortion is 2% then the maximum The registration accuracy with mutual information algo- error will be approximately 1 mm, i.e. a standard devia- rithms has been discussed above in the section about tion around 0.5 mm. The geometrical distortions system- uncertainty in target definition. Based on the high soft tis- atically affect the entire treatment through the reference sue contrast in the MR images and the similar information images, and do in addition contribute to random errors at content in the reference and positioning image it was each fraction. assumed that the accuracy is limited by the size of the vox- els. A voxel size of 1.0 × 1.0 × 2.5 mm gives a registration Page 5 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 tematic uncertainty was estimated to 0.6 mm in the AP direction, 0.2 mm in the HF direction, and neglectable in the RL direction. Position reproducibility Under direct force it was possible to displace the Miyabi shell slightly below 1 mm in the HF direction; this maxi- mum displacement corresponds to an uncertainty under normal distribution assumption of around 0.5 mm. It was not possible to measure any positioning inaccuracies in the RL and AP directions. The uncertainty in the HF direc- tion results in systematic uncertainties in the imaging for the treatment planning with a magnitude of 0.5 mm, and does in addition result in fraction to fraction positioning uncertainties of 0.7 mm (both MR and treatment table docking). Ca Figure 3 libration phantom Calibration phantom. The phantom which was used for coordinate calibration is 15 × 15 × 15 cm3 and filled with Comparison with established technique water. The central point is defined with lead bullet of 1 mm Table 1 summarizes results from the literature review in diameter which is fasten with 6 thin plexi rods creating a 3D section 3 and results presented in section 4. The total esti- hair cross. mated positioning uncertainty for a CT-based workflow, illustrated in figure 1a, is substantially larger than the esti- uncertainty of 0.5, 0.5, and 1.25 mm in the RL, AP and HF mated uncertainty using the MR-based workflow (figure directions respectively. 1b). The clinical implication of spatial uncertainties is the use of margins, dependent on both the random and sys- Results tematic part. In the present work we use the model Uncertainties associated with MR transport described through equation (1). The CT-based workflow Calibration vector should according to equation (1) be associated with the The calibration vector relates the coordinate system in the following margins: RL - 8.1 mm, AP - 8.7 mm, and HF - MR scanner with the treatment table coordinate system. 10.7 mm. The corresponding margin for the MR-based The estimated uncertainty for the calibration vector, based workflow should be: RL - 5.3 mm, AP - 6.1 mm, and HF - on the 16 independent measurements, was 0.5 mm, 0.4 8.7 mm. mm resp. 0.8 mm in the RL, HF and AP directions. The mean value of the 16 observations is connected to a sys- tematic uncertainty of 0.1 to 0.2 mm. Correction for weight The calibration vector was measured without load. There- fore there is a need to correct for the sagging of the treat- ment table under the weight of the patient. We found that the sagging of the treatment table could be modelled as a linear function of the patient weight (w) and the longitu- dinal coordinate for the prostate (l) in the MR coordinate system, according to: (2) d =−0.* 000178 wl*( − 1178.6) where the units are kg and mm respectively. Sagging of treatmen Figure 4 t table For simulated patients in the weight interval between 60 Sagging of treatment table. Modelled table sagging, the and 110 kg with their prostate located approximately 700- lines, is compared with observed sag, the points, for different 900 mm from the top of the skull, residual errors of max- simulated patient weights and prostate positions. The param- imum 1.2 mm was observed in the AP direction (figure 4), eter "Long" describes the distance from the head end of the Miyabi shell to the prostate. and 0.4 mm in the HF direction. In general the residual errors were small and the standard deviation of this sys- Page 6 of 9 (page number not for citation purposes) Table 1: Estimated positioning uncertainties CT resp. MR based treatment procedure CT based workflow MR based workflow CT/MR-systematic CT/MR-Random MR-systematic MR-random Contributing ΣRL mm ΣAP mm ΣHF mm σRL Mm σAP mm σgHF mm ΣRL mm ΣAP mm ΣHF mm σRL mm σAP mm σHF mm factor Prostate delineation 1.8 1.8 2.8 ----- ----- ----- 1.8 1.8 2.8 ----- ----- ----- Geometrical 0.2 0.2 0.2 ----- ----- ----- 0.2 0.2 0.2 ----- ----- ----- distortions MR to CT 2 2 2 ----- ----- ----- ----- ----- ----- ----- ----- ----- registration Total treatment 2.7 2.7 3.4 ----- ----- ----- 1.8 1.8 2.8 ----- ----- ----- planning uncertainty Intra-fraction motion ----- ----- ----- 0.4 1.4 1.4 ----- ----- ----- 0.4 1.4 1.4 CT to X-ray ----- ----- ----- 0.7 0.7 1.4 ----- ----- ----- ----- ----- ----- registration Fidutial marker 1.0 1.0 1.0 ----- ----- ----- ----- ----- ----- ----- ----- ----- uncertainty X-ray Imaging 1.0 1.0 1.0 ----- ----- ----- ----- ----- ----- ----- ----- ----- uncertainty MR Imaging ----- ----- ----- ----- ----- ----- 0.5 0.5 0.5 0.5 0.5 0.5 uncertainty/ distortion MR to MR ----- ----- ----- ----- ----- ----- ----- ----- ----- 0.5 0.5 1.25 registration Calibration vector ----- ----- ----- ----- ----- ----- 0.1 0.2 0.1 ----- ----- ----- determination Weight correction ----- ----- ----- ----- ----- ----- ----- 0.6 0.2 ----- ----- ----- Docking mechanism ----- ----- ----- ----- ----- ----- ----- ----- 0.5 ----- ----- 0.7 Total Set-up 1.4 1.4 1.4 0.8 1.6 2.0 0.5 0.8 0.7 0.8 1.6 2.1 uncertainty Total uncertainty 3.0 3.0 3.7 0.8 1.6 2.0 1.9 2.0 2.9 0.8 1.6 2.1 Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 Page 7 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 tration method was not included in the present study for Discussion Through this literature review together with our analysis several reasons. -The markers are not clearly visible with of the positioning procedure with MR, we claim that the the T2 weighted 3D sequence that is we use for target MR-only treatment workflow, shown in figure 1b, allows delineation. -Introduction of a dedicated sequence for vis- for significantly smaller PTV margins than the CT-based ualization of the markers gives a systematic spatial uncer- workflow (figure 1a). This conclusion has been reached tainty because of prostate movement between the through estimations of the uncertainty for each sub proc- sequences. -Use of a multi-echo sequence to acquire both ess in the treatment chain and sum-up's of the total spatial T2 weighted images for delineation and proton density uncertainty assuming that the errors from the sub proc- weighted images for visualization of the makers compro- esses are uncorrelated. This method yields results compa- mise the quality of the images used for delineation com- rable to other studies, for example, the resulting margins pared to present 3D sequence. -Finally, there is still a need for the positioning using CT-based workflow and gold for an in-depth investigation of the spatial uncertainties in markers are comparable with the results presented by Bel- the apparent marker position in the MR images, specifi- tran et al. [42]. Excluding the uncertainty in the delinea- cally, with respect to variations in frequency encoding tion of the prostate both Beltran et al. and the present direction, bandwidth, slice encoding method, and marker study estimate the proper margins to between 4 mm and shape and orientation relative the main magnetic field. 5 mm in all directions. The contributions from different Conclusion sources of uncertainty do however differ. It was shown that, from a spatial uncertainty point of The reduced uncertainty does not necessarily mean that view, the MR-only prostate treatment workflow is to be MR-only is the optimal workflow as other aspects also preferred in front of a MR/CT-based procedure. The sys- needs to be considered. It is not feasible to introduce a tematic uncertainties introduced by the MR/CT-registra- positioning method which requires considerably more tion are affecting the entire treatment but are avoided with patient time for all the 30-40 fractions than what are the MR-based workflow, while the random uncertainties standard at many departments. However, the importance from fraction to fraction are approximately the same as for of occupation time per treatment would be reduced if the the MR/CT workflow. hypo-fractionation of prostate treatments becomes clini- Competing interests cal standard. The authors declare that they have no competing interests. The delineation uncertainty is dominating the systematic overall uncertainty also for the MR only workflow. It is Authors' contributions clear that more effort needs to be spent on reducing uncer- TN Participated in the design of the study participated in tainty in the target delineation procedure. the literature review and drafted the manuscript. MN Par- ticipated in the design of the study and performed the In the present study we have used a generic algorithm for experimental work 3D distortions correction provided as a standard routine in the VB15 package delivered by Siemens. The accuracy of MGK Participated in the design of the study and in the lit- this correction was validated using a Philips PIQT phan- erature review. MK Participated in the design of the study tom, through comparison with CT and through direct dis- and in the literature review. All authors read and approved tance measurements in the images. The results were in the final manuscript agreement with the results reported by Krager et al. [23]. It Acknowledgements can be expected that the accuracy of generic distortion cor- We thank Cenneth Forsmark for the construction of the equipment, Mag- rection algorithms may vary between individual scanners, nus Karlsson (Siemens Healthcare, Sweden) for discussions and comments, it is thus important to validate the geometrical accuracy and the Swedish Cancer Society and the Cancer Research Foundation North for each MR-scanner before any clinical implementation. Sweden for financial support. Equally important is verification of the site specific regis- tration accuracy, which can differ depending of algorithm, References region of interest, and clinical implementation. The 1. Hricak H: MR imaging and MR spectroscopic imaging in the pre-treatment evaluation of prostate cancer. Br J Radiol 2005, uncertainty quantification presented in Table 1 are repre- 78(Spec No 2):S103-11. sentative for the described methodology, but should be 2. Hawighorst H, Debus J, Schreiber W, Knopp MV, Engenhart-Cabillic verified locally. R, Essig M, Brix G, van Kaick G: Contrast-enhanced magnetiza- tion transfer imaging: improvement of brain tumor conspi- cuity and delineation for radiosurgical target volume Registrations between MR and CT, and MR to MR, were in definition. Radiother Oncol 1997, 43(3):261-7. the present study performed using a MI based method. An 3. Prabhakar R, Haresh KP, Ganesh T, Joshi RC, Julka PK, Rath GK: Comparison of computed tomography and magnetic reso- alternative workflow uses the internal gold markers as ref- nance based target volume in brain tumors. J Cancer Res Ther erence points in a landmark based registration. This regis- 2007, 3(2):121-3. Page 8 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 4. Rasch C, Steenbakkers R, van Herk M: Target definition in pros- 25. Jovicich J, Czanner S, Greve D, Haley E, Kouwe A van der, Gollub R, tate, head, and neck. Semin Radiat Oncol 2005, 15(3):136-45. Kennedy D, Schmitt F, Brown G, Macfall J, Fischl B, Dale A: Reliabil- 5. Krempien RC, Daeuber S, Hensley FW, Wannenmacher M, Harms ity in multi-site structural MRI studies: effects of gradient W: Image fusion of CT and MRI data enables improved tar- non-linearity correction on phantom and human data. Neu- get volume definition in 3D-brachytherapy treatment plan- roimage 2006, 30(2):436-43. ning. Brachytherapy 2003, 2(3):164-71. 26. Schenck JF: The role of magnetic susceptibility in magnetic 6. Pasquier D, Betrouni N, Vermandel M, Lacornerie T, Lartigau E, resonance imaging: MRI magnetic compatibility of the first Rousseau J: MRI alone simulation for conformal radiation and second kinds. Med Phys 1996, 23(6):815-50. therapy of prostate cancer: technical aspects. Conf Proc IEEE 27. Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P: Multi- Eng Med Biol Soc 2006, 1:160-3. modality image registration by maximization of mutual 7. Chen L, Nguyen TB, Jones E, Chen Z, Luo W, Wang L, Price RA Jr, information. IEEE Trans Med Imaging 1997, 16(2):187-98. Pollack A, Ma CM: Magnetic resonance-based treatment plan- 28. Wells WM, Viola P, Atsumi H, Nakajima S, Kikinis R: Multi-modal ning for prostate intensity-modulated radiotherapy: crea- volume registration by maximization of mutual information. tion of digitally reconstructed radiographs. Int J Radiat Oncol Med Image Anal 1996, 1(1):35-51. Biol Phys 2007, 68(3):903-11. 29. McLaughlin PW, Narayana V, Kessler M, McShan D, Troyer S, Marsh 8. Chen L, Price RA Jr, Nguyen TB, Wang L, Li JS, Qin L, Ding M, Palacio E, L, Hixson G, Roberson PL: The use of mutual information in Ma CM, Pollack A: Dosimetric evaluation of MRI-based treatment registration of CT and MRI datasets post permanent planning for prostate cancer. Phys Med Biol 2004, 49(22):5157-70. implant. Brachytherapy 2004, 3(2):61-70. 9. Chen L, Price RA Jr, Wang L, Li J, Qin L, McNeeley S, Ma CM, Freed- 30. Vidakovic S, Jans HS, Alexander A, Sloboda RS: Post-implant com- man GM, Pollack A: MRI-based treatment planning for radio- puted tomography-magnetic resonance prostate image reg- therapy: dosimetric verification for prostate IMRT. Int J Radiat istration using feature line parallelization and normalized Oncol Biol Phys 2004, 60(2):636-47. mutual information. J Appl Clin Med Phys 2007, 8(1):21-32. 10. Chen Z, Ma CM, Paskalev K, Li J, Yang J, Richardson T, Palacio L, Xu 31. Roberson PL, McLaughlin PW, Narayana V, Troyer S, Hixson GV, X, Chen L: Investigation of MR image distortion for radiother- Kessler ML: Use and uncertainties of mutual information for apy treatment planning of prostate cancer. Phys Med Biol 2006, computed tomography/magnetic resonance (CT/MR) regis- 51(6):1393-403. tration post permanent implant of the prostate. Med Phys 11. Raaijmakers AJ, Raaymakers BW, Meer S van der, Lagendijk JJ: Integrating 2005, 32(2):473-82. a MRI scanner with a 6 MV radiotherapy accelerator: impact of 32. Kotte AN, Hofman P, Lagendijk JJ, van Vulpen M, Heide UA van der: the surface orientation on the entrance and exit dose due to the Intrafraction motion of the prostate during external-beam transverse magnetic field. Phys Med Biol 2007, 52(4):929-39. radiation therapy: analysis of 427 patients with implanted 12. Lagendijk JJ, Raaymakers BW, Raaijmakers AJ, Overweg J, Brown KJ, fiducial markers. Int J Radiat Oncol Biol Phys 2007, 69(2):419-25. Kerkhof EM, Put RW van der, Hardemark B, van Vulpen M, Heide UA 33. Ghilezan MJ, Jaffray DA, Siewerdsen JH, Van Herk M, Shetty A, Sharpe van der: MRI/linac integration. Radiother Oncol 2008, 86(1):25-9. MB, Zafar Jafri S, Vicini FA, Matter RC, Brabbins DS, Martinez AA: 13. Karlsson M, Karlsson MG, Nyholm T, Amies C, Zackrisson B: Dedi- Prostate gland motion assessed with cine-magnetic reso- cated MR in the Radiotherapy clinic. Int J Radiat Oncol Biol Phys nance imaging (cine-MRI). Int J Radiat Oncol Biol Phys 2005, 2009, 74(2):644-51. 62(2):406-17. 14. Parker CC, Damyanovich A, Haycocks T, Haider M, Bayley A, Catton 34. Poulsen PR, Muren LP, Hoyer M: Residual set-up errors and mar- CN: Magnetic resonance imaging in the radiation treatment gins in on-line image-guided prostate localization in radio- planning of localized prostate cancer using intra-prostatic therapy. Radiother Oncol 2007, 85(2):201-6. fiducial markers for computed tomography co-registration. 35. Fiorino C, Di Muzio N, Broggi S, Cozzarini C, Maggiulli E, Alongi F, Radiother Oncol 2003, 66(2):217-24. Valdagni R, Fazio F, Calandrino R: Evidence of Limited Motion of 15. Janke A, Zhao H, Cowin GJ, Galloway GJ, Doddrell DM: Use of the Prostate by Carefully Emptying the Rectum as Assessed spherical harmonic deconvolution methods to compensate by Daily MVCT Image Guidance with Helical Tomotherapy. for nonlinear gradient effects on MRI images. Magn Reson Med Int J Radiat Oncol Biol Phys 2008, 1;71(2):611-7. 2004, 52(1):115-22. 36. Britton KR, Takai Y, Mitsuya M, Nemoto K, Ogawa Y, Yamada S: 16. van Herk M, Remeijer P, Rasch C, Lebesque JV: The probability of Evaluation of inter- and intrafraction organ motion during correct target dosage: dose-population histograms for deriv- intensity modulated radiation therapy (IMRT) for localized ing treatment margins in radiotherapy. Int J Radiat Oncol Biol prostate cancer measured by a newly developed on-board Phys 2000, 47(4):1121-35. image-guided system. Radiat Med 2005, 23(1):14-24. 17. van Herk M: Errors and margins in radiotherapy. Semin Radiat 37. Berthelet E, Truong PT, Zavgorodni S, Moravan V, Liu MC, Runkel J, Oncol 2004, 14(1):52-64. Bendorffe B, Sayers D: Consistency in electronic portal imaging 18. Rasch C, Barillot I, Remeijer P, Touw A, van Herk M, Lebesque JV: registration in prostate cancer radiation treatment verifica- Definition of the prostate in CT and MRI: a multi-observer tion. Radiat Oncol 2006, 1:37. study. Int J Radiat Oncol Biol Phys 1999, 43(1):57-66. 38. Moseley DJ, White EA, Wiltshire KL, Rosewall T, Sharpe MB, Siew- 19. Smith WL, Lewis C, Bauman G, Rodrigues G, D'Souza D, Ash R, Ho erdsen JH, Bissonnette JP, Gospodarowicz M, Warde P, Catton CN, D, Venkatesan V, Downey D, Fenster A: Prostate volume con- Jaffray DA: Comparison of localization performance with touring: a 3D analysis of segmentation using 3DTRUS, CT, implanted fiducial markers and cone-beam computed tom- and MR. Int J Radiat Oncol Biol Phys 2007, 67(4):1238-47. ography for on-line image-guided radiotherapy of the pros- 20. Wang D, Strugnell W, Cowin G, Doddrell DM, Slaughter R: Geo- tate. Int J Radiat Oncol Biol Phys 2007, 67(3):942-53. metric distortion in clinical MRI systems Part I: evaluation 39. Nichol AM, Brock KK, Lockwood GA, Moseley DJ, Rosewall T, using a 3D phantom. Magn Reson Imaging 2004, 22(9):1211-21. Warde PR, Catton CN, Jaffray DA: A magnetic resonance imag- 21. Wang D, Strugnell W, Cowin G, Doddrell DM, Slaughter R: Geo- ing study of prostate deformation relative to implanted gold metric distortion in clinical MRI systems Part II: correction fiducial markers. Int J Radiat Oncol Biol Phys 2007, 67(1):48-56. using a 3D phantom. Magn Reson Imaging 2004, 22(9):1223-32. 40. Poggi MM, Gant DA, Sewchand W, Warlick WB: Marker seed 22. Wang D, Doddrell DM, Cowin G: A novel phantom and method migration in prostate localization. Int J Radiat Oncol Biol Phys for comprehensive 3-dimensional measurement and correc- 2003, 56(5):1248-51. tion of geometric distortion in magnetic resonance imaging. 41. Kitamura K, Shirato H, Shimizu S, Shinohara N, Harabayashi T, Magn Reson Imaging 2004, 22(4):529-42. Shimizu T, Kodama Y, Endo H, Onimaru R, Nishioka S, Aoyama H, 23. Karger CP, Hoss A, Bendl R, Canda V, Schad L: Accuracy of device- Tsuchiya K, Miyasaka K: Registration accuracy and possible specific 2D and 3D image distortion correction algorithms migration of internal fiducial gold marker implanted in pros- for magnetic resonance imaging of the head provided by a tate and liver treated with real-time tumor-tracking radia- manufacturer. Phys Med Biol 2006, 51(12):N253-61. tion therapy (RTRT). Radiother Oncol 2002, 62(3):275-81. 24. Doran SJ, Charles-Edwards L, Reinsberg SA, Leach MO: A complete 42. Beltran C, Herman MG, Davis BJ: Planning target margin calcu- distortion correction for MR images: I. Gradient warp cor- lations for prostate radiotherapy based on intrafraction and rection. Phys Med Biol 2005, 50(7):1343-61. interfraction motion using four localization methods. Int J Radiat Oncol Biol Phys 2008, 70(1):289-95. Page 9 of 9 (page number not for citation purposes) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Radiation Oncology Springer Journals

Systematisation of spatial uncertainties for comparison between a MR and a CT-based radiotherapy workflow for prostate treatments

Download PDF
9 pages

Loading next page...
 
/lp/springer-journals/systematisation-of-spatial-uncertainties-for-comparison-between-a-mr-Ts0svxNIhV

References (85)

Publisher
Springer Journals
Copyright
Copyright © 2009 by Nyholm et al; licensee BioMed Central Ltd.
Subject
Medicine & Public Health; Oncology; Radiotherapy
eISSN
1748-717X
DOI
10.1186/1748-717X-4-54
pmid
19919713
Publisher site
See Article on Publisher Site

Abstract

Background: In the present work we compared the spatial uncertainties associated with a MR- based workflow for external radiotherapy of prostate cancer to a standard CT-based workflow. The MR-based workflow relies on target definition and patient positioning based on MR imaging. A solution for patient transport between the MR scanner and the treatment units has been developed. For the CT-based workflow, the target is defined on a MR series but then transferred to a CT study through image registration before treatment planning, and a patient positioning using portal imaging and fiducial markers. Methods: An "open bore" 1.5T MRI scanner, Siemens Espree, has been installed in the radiotherapy department in near proximity to a treatment unit to enable patient transport between the two installations, and hence use the MRI for patient positioning. The spatial uncertainty caused by the transport was added to the uncertainty originating from the target definition process, estimated through a review of the scientific literature. The uncertainty in the CT-based workflow was estimated through a literature review. Results: The systematic uncertainties, affecting all treatment fractions, are reduced from 3-4 mm (1Sd) with a CT based workflow to 2-3 mm with a MR based workflow. The main contributing factor to this improvement is the exclusion of registration between MR and CT in the planning phase of the treatment. Conclusion: Treatment planning directly on MR images reduce the spatial uncertainty for prostate treatments. apy [6-12]. An "open bore" 1.5T MRI, has been installed Background MR images are well suited for target delineation, not only in direct connection to a treatment unit at the radiother- for the prostate [1], but also for many other tumours, such apy department in Umeå [13]. This installation allows us as brain lesions [2,3] and head and neck tumours [4,5], to image most of our patients in treatment position with which explains the growing interest for MR in radiother- the MR for the target delineation, and open the door for Page 1 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 development of an online treatment setup workflow flows are based on both a literature review and the results designed for soft tissue tumours. Figure 1 illustrates a MR- of our own experiments. only workflow and a more conventional CT-based work- flow. In the MR-based workflow, the target definition, the Methods treatment planning, and patient positioning at treatment In order to assess the total spatial uncertainties in the two delivery, are performed with MR aid only. The patient workflows, shown in figure 1, the workflow processes positioning utilize a transport trolley to move the patient were broken down into independent sub processes. Both from the imaging in the MR to the treatment table. A very workflows contain two main steps where uncertainties robust fixation of the patient provides control over the can be introduced, target definition for treatment plan- relation among the coordinate systems in the patient, in ning and patient positioning at treatment delivery. Our the MR, and in the treatment room. However, the trans- tools in the uncertainty analysis have been literature port does introduce uncertainties, which must be reviews, and when necessary own experiments. The own accounted for in an evaluation of the workflow and the experiments concern positioning with MRI, and are resulting geometric uncertainties. described in the section about MR guided delivery. An alternative workflow could be to plan on MR material An open-bore MRI scanner (Siemens Espree, 1.5T) was followed by positioning based on fiducial markers. This used for the MR imaging of the patients in connection intermediate workflow requires that the internal markers radiotherapy. For prostate patients, a T2-weighted SPACE are visible on the MR images and that the apparent marker sequence (Siemens), which is a 3D turbo spin-echo positions are correct. Parker et al. [14] shows that internal sequence with varying flip angle on the refocusing pulses, markers appear clearly on gradient echo sequences, while was used. The slice thickness was 1.7 mm, typical pixel- more difficulty to identify on T2-weighted turbo spin echo size was 1.0 × 1.0 mm , and the bandwidth was 592 Hz sequences. The visibility of the markers was increased per pixel. Distortions caused by gradient non-linearity when the TE time was reduced, giving higher signal but were corrected with an algorithm based on spherical har- compromising the T2-weighted contrast. Verified robust monic expansion of the fields generated by the gradient imaging of fiducial markers in MR would enable also this coils [15]. The 3D correction algorithm including repre- workflow. In the present study, this intermediate work- sentation of the coils was delivered by Siemens as a stand- flow will not be explicitly handled. ard clinical tool integrated in the scanner software (VB15). The scanner was set in an isocentric mode, which moves The purpose of this study is to investigate if a MR-only the table prior to the acquisition of each sequence, to radiotherapy workflow, in accordance with figure 1b, has place the MR isocenter in the centre of the volume of the potential to improve the spatial accuracy compared to interest. the more conventional CT-based workflow (figure 1a). The estimations of the uncertainties in the different work- The total spatial uncertainty consists of both a random part, varying in direction and magnitude from fraction to fraction, and a systematic part, which is invariant over the treatment period. The systematic and random uncertainty should be given different weight in the formation of mar- gins between the CTV and the PTV. In the present work we used the weight factor 2.5 for the systematic errors and 0.7 for random errors as proposed by van Herk et. al. [16,17]. The PTV margin is hence expressed as (1) m =∑ 25..+07s PTV where Σ is the systematic and σ is the random spatial Overview of th study Figure 1 e two workflows analyzed in the present uncertainty. The presented uncertainties are throughout Overview of the two workflows analyzed in the this paper presented in units of one standard deviation present study. (a) A widely used workflow utilizing regis- (1SD), thus inherently assuming normal distributed data. tration between MR and CT images in order to transfer the delineated prostate volume (GTV/CTV) from the MR study Uncertainty in target definition to the CT study. The CT study is used for treatment planning The total uncertainty in the target definition can be bro- and to generate DRR's for patient position. Typically, fiducial ken down to three subparts: uncertainty in prostate delin- markers are used. (b) The workflow is entirely based on MR, eation (MR-based on both workflows), spatial distortion both for planning and positioning. in MR images that can be scanner related and patient Page 2 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 induced, and for the CT-based workflow: uncertainty in mm. This uncertainty is approximately equal in all direc- registration between CT and MR images. tions provided that a 3D correction algorithm is used. Uncertainty in prostate delineation Registration uncertainties - MR/CT Rasch et al [18] has from a study with 18 patient analysed The workflow in figure 1a involves a registration between by 3 physicians reported an uncertainty, in the prostate a CT and MR study. Errors in this registration directly delineation on axial MR study, of 2 mm at the base of affect the spatial accuracy of the target definition. Registra- seminal vesicles and up 2.8 mm in the prostate apex. The tions between MR and CT for prostate patients can be per- uncertainty in the head-feat (HF) direction was 2.5 mm formed based on fiducial markers [14]. The trend is, with a slice thickness of 5-6 mm for the axial MR images. however, to use mutual information (MI) registration In a later study involving 7 physicians analysing 10 based directly on the patient anatomy [27,28]. The pros- patients Smith et al. [19] reported a radial uncertainty of tate position relative other anatomical structures is not fix, 0.6 - 1.6 mm for the delineation of the prostate where the therefore the registration should ideally be based on the larger value is for the apex. The inter-observer uncertainty prostate with just a small margin. However, this has been in the length (HF direction) of the prostate was 3.4 mm, reported problematic because of too limited morphologi- and the intra-observer variation was 2.6 mm; the slice cal information content in the CT representation of the thickness was 2.5 mm. prostate [29,30]. A few studies have been performed eval- uating the accuracy and precision of MI registration for CT In summary, the literature review indicates a prostate and MR studies of the prostate; the registration uncer- delineation uncertainty of 1.8 mm in the right-left (RL) tainty has been reported to be around 2 mm [29,31]. Rob- and anterior-posterior (AP) directions and 2.8 mm in the erson et al. [31] reported that registration results depend HF direction. on the starting point for a specific MI optimization soft- ware. The mean difference between different stating Geometrical Distortions in MR points was up to 1 mm in the RL direction. The corre- Geometrical distortions in MR images are a well known sponding number for MR-MR registration was 0.4 mm in phenomena [20-22]. In modern MR scanners, gradient the HF direction which could indicate that the mutual non-linearity is the main cause of image distortions [20], information maximum is more distinct for MR-MR regis- dominating over the effect of static field inhomogenity. tration compared to CT-MR registration. The distortions introduced by the gradient non-linearities are increasing with the distance from the MR isocenter In summary, the registration uncertainty for a CT - MR reg- [20,23] Without correction, the geometrical distortions in istration for a prostate case was estimated to be 2 mm modern MR scanners can cause deviations between phys- based on current reports in the scientific literature. ical and imaged distances of up to 20% in extreme situa- tions. However, there are methods for distortion Uncertainty in patient positioning correction which reduces the errors significantly. It is pos- The patient positioning at treatment, with the develop- sible to use a specially designed geometry phantom to ment of image guided radiotherapy, been in focus the characterize and correct the distortions for a specific scan- recent years. For prostate cancer patients the improve- ner [22,24]. In the present work, a gradient coil specific ments in spatial treatment accuracy has been considera- distortion correction algorithm was applied. Even though ble. Both the CT and the MR-based workflows, shown in this device specific corrections only correct for intrinsic figure 1, rely on imaging before each fraction. Intra-frac- gradient non-linearity connected to a specific type of scan- tion motion of the prostate is therefore an issue for both ner/gradient coil, it has been shown that this kind of cor- workflows. rection yields a spatial accuracy better than 2% [23,25], which is sufficient when region of interest in the patient is Intra-fraction prostate motion close to the MR isocenter. Patient anatomy, e.g. air pockets In a large investigation by Kotte et al [32] intra fraction in the rectal cavity, can generate susceptibility-generated motion larger than 2 mm was observed during 66% of the field changes up to ± 10 ppm [26]. With a bandwidth of fractions, this number is roughly in agreement with the 592 Hz per pixel this corresponds to distortions up to results presented in other studies [33,34]. However, approximately 1 pixel for a 1.5T scanner. Thus, magnetic reduction of the rectal filling has been showed to be of susceptibility related distortions are a minor effect for the great importance to achieve a stable prostate position sequence used. [33,35]; an uncertainty of 2 mm is therefore realistic for a 5-7 min treatment when patients are instructed to empty In summary, for a prostate with radius of 2.5 cm the geo- rectum prior to treatment. The position uncertainty due to metrical distortions can cause errors of up to 0.5 mm, prostate motion is most pronounced in the AP and HF which corresponds to a standard deviation of around 0.2 directions [32,36]. Page 3 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 In summary, the overall uncertainty for the prostate posi- patient is fixated on a shell, with a double vacuum system tion was estimated to 2 mm, which broken down in the (BodyFIX, Medical Intelligence an Elekta company), orthogonal directions corresponds to: 1.4 mm in AP and which can be slid from the trolley to the treatment or MR HF, and 0.4 mm in RL. table after docking. The shell has fixed positions both at the MR and the treatment table, which enables absolute Uncertainty with fiducial markers coordinate transformation between MR coordinates and There are numerous studies on the accuracy of patient treatment coordinates. The treatment table is a Siemens positioning using fiducial markers and portal or flat 550 TxT equipped with a modified TT-D table-top com- screen kV images. Several different sources of uncertainty patible with the Miyabi transport solution. The daily treat- need to be considered in order to correctly estimate the ment table coordinates are calculated as the absolute table overall accuracy of the workflow. Random positioning coordinates from the treatment planning corrected for errors are partly due to uncertainty in the registration daily variations in patient and prostate position. The daily between the reference image and the portal/kV flat screen correction is calculated based on a sub-volume-based image. Literature indicates that a manual registration typ- rigid mutual information registration between the refer- ically results in uncertainty of around 0.7 mm in the HF ence MR images used at treatment planning and daily and RL direction, and 1.4 mm in the AP direction [37,38]. positioning MR images. The same SPACE sequence was used both for treatment planning and for daily position- An investigation by Nichol et al [39] indicates that a sys- ing. Calibration of the system, i.e. determination of the tematic deformation of the prostate during radiotherapy absolute coordinate transformation vector, is an obvious leads to drift in the relation between the centre of mass for source for systematic uncertainty, while mechanical insta- the markers and centre of mass for the contoured prostate. bilities in the mounting mechanism at the MR and treat- This uncertainty is in the order of 1 mm, which is roughly ment table together with image distortion, image in agreement with other reports [40,41]. It should be registration errors and patient movement during transport noted that deformation of the prostate is in many respects mainly result in uncertainty of random nature. equivalent to marker migration within the prostate. These two effects are therefore not separated in the present work. Uncertainty in calibration vector determination The calibration vector is the relation between the coordi- Prostate deformation and marker migration are resulting in a systematic uncertainty in the patient position. nate for a specific point, in the MR coordinate system and the treatment table coordinates that brings the same point The uncertainty of clinical imaging systems are in the to the treatment isocenter. The calibration vector was order of 1 mm, accounting for limitations in resolution, determined using a phantom which is sketched in figure isocenter position and mechanical instability. Paulsen et 3. The centre point of the phantom is clearly visible on al [34] observed a systematic discrepancy of almost 1 mm MR, CT, portal images and can also be positioned using when comparing 2 different imaging modalities at 2 dif- lasers. We placed the phantom at various positions on the ferent accelerators. Kotte et al. [32] detected that the sag of Miyabi shell and carefully determined the position of the the gantry caused a systematic imaging deviations of centre point in both the MR coordinate system using MR almost 1 mm in the HF direction when the gantry was in images, and the treatment coordinate system using cali- 0 degree position compared to 180 degree position. brated lasers. The calibration vector was calculated, for each phantom position on the Miyabi shell, as the differ- In summary, it is estimated that the uncertainty in the day ence between the MR coordinates and the treatment table to day registrations between reference image and the por- coordinates for the central point in the phantom. The idea tal image is 0.7 mm in RL and HF direction and 1.4 mm with repeated measurements was to assess the precision of in AP direction. The estimated uncertainty for the marker the vector determination taking intrinsic inhomogeneities position in the prostate is 1 mm in all directions, and the in the magnetic field and position dependent distortions estimated total uncertainty for the imaging systems is 1 into account. In total 16 independent determinations of mm in all directions. the calibration vector was performed, for different phan- tom positions on the Miyabi shell. The measurements MR guided treatment delivery were performed with the phantom centre positioned at ± The MR positioning approach is novel; we therefore 25 mm in the AP direction, and ± 60 mm in the RL direc- describe the principals in detail below, as well as the tion and at 4 different positions along the HF direction experiments performed to estimate the uncertainties con- with a total span of 450 mm. The scanning of the phan- nected to the method. tom was performed in isocentric mode. Figure 2 shows the hardware configuration. The patient is Weight correction transported between the MR scanner and the treatment The calibration vector needs to be corrected based on the unit on a MR compatible trolley (Miyabi, TRUMPF). The patient's weight to account for the treatment table sag- Page 4 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 Schematic overview of the hardware config Figure 2 uration for the MR positioning of patients Schematic overview of the hardware configuration for the MR positioning of patients. There is a direct connection between the MR room and treatment room, which makes patient transport quick and simple. In parallel with the patient trans- port the treatment couch coordinates are calculated using dedicated image registration software, the transport in it self does therefore not prolong the procedure. ging. The magnitude of the sagging was investigated using Patient movement a set of 15 kg bricks which were distributed to approxi- Significant patient movements during the time interval mate the weight distribution of a typical patient. We var- from the imaging to the treatment are deemed highly ied the total load and the weight distribution on the table unlikely when using the double vacuum immobilization top, to simulate patient weight from 0 to 105 kg, and device. There is however a risk for prostate movements patient height from approximately 150 cm to 190 cm. within the body during this time interval as discussed above (see section about intra-fraction prostate move- Geometrical distortions ment) The prostate is typically located on the patient's central line and with the Miyabi shell together with the BodyFIX Position reproducibility vacuum pillow the height of the prostate for the typical The reproducibility of the Miyabi shell position on the MR patient will be very close to the isocenter. The internal MR and treatment table were investigated through measure- laser is used to position the patient in the HF direction ment of the maximum shell displacement under direct before imaging, thus the prostate will be close to the iso- force in different directions center also in the HF direction. If the prostate centre is Registration uncertainties MR/MR within a sphere of 5 cm around the MR isocenter and the maximum spatial distortion is 2% then the maximum The registration accuracy with mutual information algo- error will be approximately 1 mm, i.e. a standard devia- rithms has been discussed above in the section about tion around 0.5 mm. The geometrical distortions system- uncertainty in target definition. Based on the high soft tis- atically affect the entire treatment through the reference sue contrast in the MR images and the similar information images, and do in addition contribute to random errors at content in the reference and positioning image it was each fraction. assumed that the accuracy is limited by the size of the vox- els. A voxel size of 1.0 × 1.0 × 2.5 mm gives a registration Page 5 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 tematic uncertainty was estimated to 0.6 mm in the AP direction, 0.2 mm in the HF direction, and neglectable in the RL direction. Position reproducibility Under direct force it was possible to displace the Miyabi shell slightly below 1 mm in the HF direction; this maxi- mum displacement corresponds to an uncertainty under normal distribution assumption of around 0.5 mm. It was not possible to measure any positioning inaccuracies in the RL and AP directions. The uncertainty in the HF direc- tion results in systematic uncertainties in the imaging for the treatment planning with a magnitude of 0.5 mm, and does in addition result in fraction to fraction positioning uncertainties of 0.7 mm (both MR and treatment table docking). Ca Figure 3 libration phantom Calibration phantom. The phantom which was used for coordinate calibration is 15 × 15 × 15 cm3 and filled with Comparison with established technique water. The central point is defined with lead bullet of 1 mm Table 1 summarizes results from the literature review in diameter which is fasten with 6 thin plexi rods creating a 3D section 3 and results presented in section 4. The total esti- hair cross. mated positioning uncertainty for a CT-based workflow, illustrated in figure 1a, is substantially larger than the esti- uncertainty of 0.5, 0.5, and 1.25 mm in the RL, AP and HF mated uncertainty using the MR-based workflow (figure directions respectively. 1b). The clinical implication of spatial uncertainties is the use of margins, dependent on both the random and sys- Results tematic part. In the present work we use the model Uncertainties associated with MR transport described through equation (1). The CT-based workflow Calibration vector should according to equation (1) be associated with the The calibration vector relates the coordinate system in the following margins: RL - 8.1 mm, AP - 8.7 mm, and HF - MR scanner with the treatment table coordinate system. 10.7 mm. The corresponding margin for the MR-based The estimated uncertainty for the calibration vector, based workflow should be: RL - 5.3 mm, AP - 6.1 mm, and HF - on the 16 independent measurements, was 0.5 mm, 0.4 8.7 mm. mm resp. 0.8 mm in the RL, HF and AP directions. The mean value of the 16 observations is connected to a sys- tematic uncertainty of 0.1 to 0.2 mm. Correction for weight The calibration vector was measured without load. There- fore there is a need to correct for the sagging of the treat- ment table under the weight of the patient. We found that the sagging of the treatment table could be modelled as a linear function of the patient weight (w) and the longitu- dinal coordinate for the prostate (l) in the MR coordinate system, according to: (2) d =−0.* 000178 wl*( − 1178.6) where the units are kg and mm respectively. Sagging of treatmen Figure 4 t table For simulated patients in the weight interval between 60 Sagging of treatment table. Modelled table sagging, the and 110 kg with their prostate located approximately 700- lines, is compared with observed sag, the points, for different 900 mm from the top of the skull, residual errors of max- simulated patient weights and prostate positions. The param- imum 1.2 mm was observed in the AP direction (figure 4), eter "Long" describes the distance from the head end of the Miyabi shell to the prostate. and 0.4 mm in the HF direction. In general the residual errors were small and the standard deviation of this sys- Page 6 of 9 (page number not for citation purposes) Table 1: Estimated positioning uncertainties CT resp. MR based treatment procedure CT based workflow MR based workflow CT/MR-systematic CT/MR-Random MR-systematic MR-random Contributing ΣRL mm ΣAP mm ΣHF mm σRL Mm σAP mm σgHF mm ΣRL mm ΣAP mm ΣHF mm σRL mm σAP mm σHF mm factor Prostate delineation 1.8 1.8 2.8 ----- ----- ----- 1.8 1.8 2.8 ----- ----- ----- Geometrical 0.2 0.2 0.2 ----- ----- ----- 0.2 0.2 0.2 ----- ----- ----- distortions MR to CT 2 2 2 ----- ----- ----- ----- ----- ----- ----- ----- ----- registration Total treatment 2.7 2.7 3.4 ----- ----- ----- 1.8 1.8 2.8 ----- ----- ----- planning uncertainty Intra-fraction motion ----- ----- ----- 0.4 1.4 1.4 ----- ----- ----- 0.4 1.4 1.4 CT to X-ray ----- ----- ----- 0.7 0.7 1.4 ----- ----- ----- ----- ----- ----- registration Fidutial marker 1.0 1.0 1.0 ----- ----- ----- ----- ----- ----- ----- ----- ----- uncertainty X-ray Imaging 1.0 1.0 1.0 ----- ----- ----- ----- ----- ----- ----- ----- ----- uncertainty MR Imaging ----- ----- ----- ----- ----- ----- 0.5 0.5 0.5 0.5 0.5 0.5 uncertainty/ distortion MR to MR ----- ----- ----- ----- ----- ----- ----- ----- ----- 0.5 0.5 1.25 registration Calibration vector ----- ----- ----- ----- ----- ----- 0.1 0.2 0.1 ----- ----- ----- determination Weight correction ----- ----- ----- ----- ----- ----- ----- 0.6 0.2 ----- ----- ----- Docking mechanism ----- ----- ----- ----- ----- ----- ----- ----- 0.5 ----- ----- 0.7 Total Set-up 1.4 1.4 1.4 0.8 1.6 2.0 0.5 0.8 0.7 0.8 1.6 2.1 uncertainty Total uncertainty 3.0 3.0 3.7 0.8 1.6 2.0 1.9 2.0 2.9 0.8 1.6 2.1 Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 Page 7 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 tration method was not included in the present study for Discussion Through this literature review together with our analysis several reasons. -The markers are not clearly visible with of the positioning procedure with MR, we claim that the the T2 weighted 3D sequence that is we use for target MR-only treatment workflow, shown in figure 1b, allows delineation. -Introduction of a dedicated sequence for vis- for significantly smaller PTV margins than the CT-based ualization of the markers gives a systematic spatial uncer- workflow (figure 1a). This conclusion has been reached tainty because of prostate movement between the through estimations of the uncertainty for each sub proc- sequences. -Use of a multi-echo sequence to acquire both ess in the treatment chain and sum-up's of the total spatial T2 weighted images for delineation and proton density uncertainty assuming that the errors from the sub proc- weighted images for visualization of the makers compro- esses are uncorrelated. This method yields results compa- mise the quality of the images used for delineation com- rable to other studies, for example, the resulting margins pared to present 3D sequence. -Finally, there is still a need for the positioning using CT-based workflow and gold for an in-depth investigation of the spatial uncertainties in markers are comparable with the results presented by Bel- the apparent marker position in the MR images, specifi- tran et al. [42]. Excluding the uncertainty in the delinea- cally, with respect to variations in frequency encoding tion of the prostate both Beltran et al. and the present direction, bandwidth, slice encoding method, and marker study estimate the proper margins to between 4 mm and shape and orientation relative the main magnetic field. 5 mm in all directions. The contributions from different Conclusion sources of uncertainty do however differ. It was shown that, from a spatial uncertainty point of The reduced uncertainty does not necessarily mean that view, the MR-only prostate treatment workflow is to be MR-only is the optimal workflow as other aspects also preferred in front of a MR/CT-based procedure. The sys- needs to be considered. It is not feasible to introduce a tematic uncertainties introduced by the MR/CT-registra- positioning method which requires considerably more tion are affecting the entire treatment but are avoided with patient time for all the 30-40 fractions than what are the MR-based workflow, while the random uncertainties standard at many departments. However, the importance from fraction to fraction are approximately the same as for of occupation time per treatment would be reduced if the the MR/CT workflow. hypo-fractionation of prostate treatments becomes clini- Competing interests cal standard. The authors declare that they have no competing interests. The delineation uncertainty is dominating the systematic overall uncertainty also for the MR only workflow. It is Authors' contributions clear that more effort needs to be spent on reducing uncer- TN Participated in the design of the study participated in tainty in the target delineation procedure. the literature review and drafted the manuscript. MN Par- ticipated in the design of the study and performed the In the present study we have used a generic algorithm for experimental work 3D distortions correction provided as a standard routine in the VB15 package delivered by Siemens. The accuracy of MGK Participated in the design of the study and in the lit- this correction was validated using a Philips PIQT phan- erature review. MK Participated in the design of the study tom, through comparison with CT and through direct dis- and in the literature review. All authors read and approved tance measurements in the images. The results were in the final manuscript agreement with the results reported by Krager et al. [23]. It Acknowledgements can be expected that the accuracy of generic distortion cor- We thank Cenneth Forsmark for the construction of the equipment, Mag- rection algorithms may vary between individual scanners, nus Karlsson (Siemens Healthcare, Sweden) for discussions and comments, it is thus important to validate the geometrical accuracy and the Swedish Cancer Society and the Cancer Research Foundation North for each MR-scanner before any clinical implementation. Sweden for financial support. Equally important is verification of the site specific regis- tration accuracy, which can differ depending of algorithm, References region of interest, and clinical implementation. The 1. Hricak H: MR imaging and MR spectroscopic imaging in the pre-treatment evaluation of prostate cancer. Br J Radiol 2005, uncertainty quantification presented in Table 1 are repre- 78(Spec No 2):S103-11. sentative for the described methodology, but should be 2. Hawighorst H, Debus J, Schreiber W, Knopp MV, Engenhart-Cabillic verified locally. R, Essig M, Brix G, van Kaick G: Contrast-enhanced magnetiza- tion transfer imaging: improvement of brain tumor conspi- cuity and delineation for radiosurgical target volume Registrations between MR and CT, and MR to MR, were in definition. Radiother Oncol 1997, 43(3):261-7. the present study performed using a MI based method. An 3. Prabhakar R, Haresh KP, Ganesh T, Joshi RC, Julka PK, Rath GK: Comparison of computed tomography and magnetic reso- alternative workflow uses the internal gold markers as ref- nance based target volume in brain tumors. J Cancer Res Ther erence points in a landmark based registration. This regis- 2007, 3(2):121-3. Page 8 of 9 (page number not for citation purposes) Radiation Oncology 2009, 4:54 http://www.ro-journal.com/content/4/1/54 4. Rasch C, Steenbakkers R, van Herk M: Target definition in pros- 25. Jovicich J, Czanner S, Greve D, Haley E, Kouwe A van der, Gollub R, tate, head, and neck. Semin Radiat Oncol 2005, 15(3):136-45. Kennedy D, Schmitt F, Brown G, Macfall J, Fischl B, Dale A: Reliabil- 5. Krempien RC, Daeuber S, Hensley FW, Wannenmacher M, Harms ity in multi-site structural MRI studies: effects of gradient W: Image fusion of CT and MRI data enables improved tar- non-linearity correction on phantom and human data. Neu- get volume definition in 3D-brachytherapy treatment plan- roimage 2006, 30(2):436-43. ning. Brachytherapy 2003, 2(3):164-71. 26. Schenck JF: The role of magnetic susceptibility in magnetic 6. Pasquier D, Betrouni N, Vermandel M, Lacornerie T, Lartigau E, resonance imaging: MRI magnetic compatibility of the first Rousseau J: MRI alone simulation for conformal radiation and second kinds. Med Phys 1996, 23(6):815-50. therapy of prostate cancer: technical aspects. Conf Proc IEEE 27. Maes F, Collignon A, Vandermeulen D, Marchal G, Suetens P: Multi- Eng Med Biol Soc 2006, 1:160-3. modality image registration by maximization of mutual 7. Chen L, Nguyen TB, Jones E, Chen Z, Luo W, Wang L, Price RA Jr, information. IEEE Trans Med Imaging 1997, 16(2):187-98. Pollack A, Ma CM: Magnetic resonance-based treatment plan- 28. Wells WM, Viola P, Atsumi H, Nakajima S, Kikinis R: Multi-modal ning for prostate intensity-modulated radiotherapy: crea- volume registration by maximization of mutual information. tion of digitally reconstructed radiographs. Int J Radiat Oncol Med Image Anal 1996, 1(1):35-51. Biol Phys 2007, 68(3):903-11. 29. McLaughlin PW, Narayana V, Kessler M, McShan D, Troyer S, Marsh 8. Chen L, Price RA Jr, Nguyen TB, Wang L, Li JS, Qin L, Ding M, Palacio E, L, Hixson G, Roberson PL: The use of mutual information in Ma CM, Pollack A: Dosimetric evaluation of MRI-based treatment registration of CT and MRI datasets post permanent planning for prostate cancer. Phys Med Biol 2004, 49(22):5157-70. implant. Brachytherapy 2004, 3(2):61-70. 9. Chen L, Price RA Jr, Wang L, Li J, Qin L, McNeeley S, Ma CM, Freed- 30. Vidakovic S, Jans HS, Alexander A, Sloboda RS: Post-implant com- man GM, Pollack A: MRI-based treatment planning for radio- puted tomography-magnetic resonance prostate image reg- therapy: dosimetric verification for prostate IMRT. Int J Radiat istration using feature line parallelization and normalized Oncol Biol Phys 2004, 60(2):636-47. mutual information. J Appl Clin Med Phys 2007, 8(1):21-32. 10. Chen Z, Ma CM, Paskalev K, Li J, Yang J, Richardson T, Palacio L, Xu 31. Roberson PL, McLaughlin PW, Narayana V, Troyer S, Hixson GV, X, Chen L: Investigation of MR image distortion for radiother- Kessler ML: Use and uncertainties of mutual information for apy treatment planning of prostate cancer. Phys Med Biol 2006, computed tomography/magnetic resonance (CT/MR) regis- 51(6):1393-403. tration post permanent implant of the prostate. Med Phys 11. Raaijmakers AJ, Raaymakers BW, Meer S van der, Lagendijk JJ: Integrating 2005, 32(2):473-82. a MRI scanner with a 6 MV radiotherapy accelerator: impact of 32. Kotte AN, Hofman P, Lagendijk JJ, van Vulpen M, Heide UA van der: the surface orientation on the entrance and exit dose due to the Intrafraction motion of the prostate during external-beam transverse magnetic field. Phys Med Biol 2007, 52(4):929-39. radiation therapy: analysis of 427 patients with implanted 12. Lagendijk JJ, Raaymakers BW, Raaijmakers AJ, Overweg J, Brown KJ, fiducial markers. Int J Radiat Oncol Biol Phys 2007, 69(2):419-25. Kerkhof EM, Put RW van der, Hardemark B, van Vulpen M, Heide UA 33. Ghilezan MJ, Jaffray DA, Siewerdsen JH, Van Herk M, Shetty A, Sharpe van der: MRI/linac integration. Radiother Oncol 2008, 86(1):25-9. MB, Zafar Jafri S, Vicini FA, Matter RC, Brabbins DS, Martinez AA: 13. Karlsson M, Karlsson MG, Nyholm T, Amies C, Zackrisson B: Dedi- Prostate gland motion assessed with cine-magnetic reso- cated MR in the Radiotherapy clinic. Int J Radiat Oncol Biol Phys nance imaging (cine-MRI). Int J Radiat Oncol Biol Phys 2005, 2009, 74(2):644-51. 62(2):406-17. 14. Parker CC, Damyanovich A, Haycocks T, Haider M, Bayley A, Catton 34. Poulsen PR, Muren LP, Hoyer M: Residual set-up errors and mar- CN: Magnetic resonance imaging in the radiation treatment gins in on-line image-guided prostate localization in radio- planning of localized prostate cancer using intra-prostatic therapy. Radiother Oncol 2007, 85(2):201-6. fiducial markers for computed tomography co-registration. 35. Fiorino C, Di Muzio N, Broggi S, Cozzarini C, Maggiulli E, Alongi F, Radiother Oncol 2003, 66(2):217-24. Valdagni R, Fazio F, Calandrino R: Evidence of Limited Motion of 15. Janke A, Zhao H, Cowin GJ, Galloway GJ, Doddrell DM: Use of the Prostate by Carefully Emptying the Rectum as Assessed spherical harmonic deconvolution methods to compensate by Daily MVCT Image Guidance with Helical Tomotherapy. for nonlinear gradient effects on MRI images. Magn Reson Med Int J Radiat Oncol Biol Phys 2008, 1;71(2):611-7. 2004, 52(1):115-22. 36. Britton KR, Takai Y, Mitsuya M, Nemoto K, Ogawa Y, Yamada S: 16. van Herk M, Remeijer P, Rasch C, Lebesque JV: The probability of Evaluation of inter- and intrafraction organ motion during correct target dosage: dose-population histograms for deriv- intensity modulated radiation therapy (IMRT) for localized ing treatment margins in radiotherapy. Int J Radiat Oncol Biol prostate cancer measured by a newly developed on-board Phys 2000, 47(4):1121-35. image-guided system. Radiat Med 2005, 23(1):14-24. 17. van Herk M: Errors and margins in radiotherapy. Semin Radiat 37. Berthelet E, Truong PT, Zavgorodni S, Moravan V, Liu MC, Runkel J, Oncol 2004, 14(1):52-64. Bendorffe B, Sayers D: Consistency in electronic portal imaging 18. Rasch C, Barillot I, Remeijer P, Touw A, van Herk M, Lebesque JV: registration in prostate cancer radiation treatment verifica- Definition of the prostate in CT and MRI: a multi-observer tion. Radiat Oncol 2006, 1:37. study. Int J Radiat Oncol Biol Phys 1999, 43(1):57-66. 38. Moseley DJ, White EA, Wiltshire KL, Rosewall T, Sharpe MB, Siew- 19. Smith WL, Lewis C, Bauman G, Rodrigues G, D'Souza D, Ash R, Ho erdsen JH, Bissonnette JP, Gospodarowicz M, Warde P, Catton CN, D, Venkatesan V, Downey D, Fenster A: Prostate volume con- Jaffray DA: Comparison of localization performance with touring: a 3D analysis of segmentation using 3DTRUS, CT, implanted fiducial markers and cone-beam computed tom- and MR. Int J Radiat Oncol Biol Phys 2007, 67(4):1238-47. ography for on-line image-guided radiotherapy of the pros- 20. Wang D, Strugnell W, Cowin G, Doddrell DM, Slaughter R: Geo- tate. Int J Radiat Oncol Biol Phys 2007, 67(3):942-53. metric distortion in clinical MRI systems Part I: evaluation 39. Nichol AM, Brock KK, Lockwood GA, Moseley DJ, Rosewall T, using a 3D phantom. Magn Reson Imaging 2004, 22(9):1211-21. Warde PR, Catton CN, Jaffray DA: A magnetic resonance imag- 21. Wang D, Strugnell W, Cowin G, Doddrell DM, Slaughter R: Geo- ing study of prostate deformation relative to implanted gold metric distortion in clinical MRI systems Part II: correction fiducial markers. Int J Radiat Oncol Biol Phys 2007, 67(1):48-56. using a 3D phantom. Magn Reson Imaging 2004, 22(9):1223-32. 40. Poggi MM, Gant DA, Sewchand W, Warlick WB: Marker seed 22. Wang D, Doddrell DM, Cowin G: A novel phantom and method migration in prostate localization. Int J Radiat Oncol Biol Phys for comprehensive 3-dimensional measurement and correc- 2003, 56(5):1248-51. tion of geometric distortion in magnetic resonance imaging. 41. Kitamura K, Shirato H, Shimizu S, Shinohara N, Harabayashi T, Magn Reson Imaging 2004, 22(4):529-42. Shimizu T, Kodama Y, Endo H, Onimaru R, Nishioka S, Aoyama H, 23. Karger CP, Hoss A, Bendl R, Canda V, Schad L: Accuracy of device- Tsuchiya K, Miyasaka K: Registration accuracy and possible specific 2D and 3D image distortion correction algorithms migration of internal fiducial gold marker implanted in pros- for magnetic resonance imaging of the head provided by a tate and liver treated with real-time tumor-tracking radia- manufacturer. Phys Med Biol 2006, 51(12):N253-61. tion therapy (RTRT). Radiother Oncol 2002, 62(3):275-81. 24. Doran SJ, Charles-Edwards L, Reinsberg SA, Leach MO: A complete 42. Beltran C, Herman MG, Davis BJ: Planning target margin calcu- distortion correction for MR images: I. Gradient warp cor- lations for prostate radiotherapy based on intrafraction and rection. Phys Med Biol 2005, 50(7):1343-61. interfraction motion using four localization methods. Int J Radiat Oncol Biol Phys 2008, 70(1):289-95. Page 9 of 9 (page number not for citation purposes)

Journal

Radiation Oncology – Springer Journals

Published: Nov 17, 2009

There are no references for this article.