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

Learn More →

Validation of an MRI-only planning workflow for definitive pelvic radiotherapy

Validation of an MRI-only planning workflow for definitive pelvic radiotherapy Purpose: Previous work on Magnetic Resonance Imaging (MRI) only planning has been applied to limited treat‑ ment regions with a focus on male anatomy. This research aimed to validate the use of a hybrid multi‑atlas synthetic computed tomography (sCT ) generation technique from a MRI, using a female and male atlas, for MRI only radiation therapy treatment planning of rectum, anal canal, cervix and endometrial malignancies. Patients and methods: Forty patients receiving radiation treatment for a range of pelvic malignancies, were sepa‑ rated into male (n = 20) and female (n = 20) cohorts for the creation of gender specific atlases. A multi‑atlas local weighted voting method was used to generate a sCT from a T1‑ weighted VIBE DIXON MRI sequence. The original treatment plans were copied from the CT scan to the corresponding sCT for dosimetric validation. Results: The median percentage dose difference between the treatment plan on the CT and sCT at the ICRU refer ‑ ence point for the male cohort was − 0.4% (IQR of 0 to − 0.6), and − 0.3% (IQR of 0 to − 0.6) for the female cohort. The mean gamma agreement for both cohorts was > 99% for criteria of 3%/2 mm and 2%/2 mm. With dose criteria of 1%/1 mm, the pass rate was higher for the male cohort at 96.3% than the female cohort at 93.4%. MRI to sCT anatomi‑ cal agreement for bone and body delineated contours was assessed, with a resulting Dice score of 0.91 ± 0.2 (mean ± 1 SD) and 0.97 ± 0.0 for the male cohort respectively; and 0.96 ± 0.0 and 0.98 ± 0.0 for the female cohort respec‑ tively. The mean absolute error in Hounsfield units (HUs) within the entire body for the male and female cohorts was 59.1 HU ± 7.2 HU and 53.3 HU ± 8.9 HU respectively. Conclusions: A multi‑atlas based method for sCT generation can be applied to a standard T1‑ weighted MRI sequence for male and female pelvic patients. The implications of this study support MRI only planning being applied more broadly for both male and female pelvic sites. Trial registration This trial was registered in the Australian New Zealand Clinical Trials Registry (ANZCTR) (www. anzctr. org. au) on 04/10/2017. Trial identifier ACTRN12617001406392. Keywords: MRI radiotherapy planning, Radiotherapy, Rectum neoplasms, Cervix neoplasms, Endometrium neoplasms, Anal canal neoplasms, Synthetic CT, Computer assisted radiotherapy planning, Image guided radiotherapy, Intensity modulated radiotherapy Background Computed tomography (CT) is the long established imaging modality used for radiation therapy treatment *Correspondence: Laura.OConnor@calvarymater.org.au Department of Radiation Oncology, Calvary Mater Hospital, Cnr Edith & planning. The inherent electron density [derived from Platt St, Waratah, , Newcastle, NSW 2298, Australia grey scale Hounsfield units (HU)] and anatomical data Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. O’Connor et al. Radiation Oncology (2022) 17:55 Page 2 of 11 from the CT scan, is used by commercial computer treatment volumes for colorectal and gynaecological planning systems to model and calculate radiation dose cancers are much more multifarious, traversing a more distribution within the patient’s body, using specific cal - variable body contour and bony anatomy than pros- culation algorithms [1]. The limited soft tissue contrast tate treatments. Rectum, anal canal and gynaecological and tumour delineation however, has given rise to the treatments routinely involve treating the gross tumour increasing use of Magnetic Resonance Imaging (MRI) in volume, surrounding tissue deemed to be at high risk of this context [2, 3]. As such, dedicated MRI scanners and tumour spread, the disease positive nodes and the sur- MRI-linear accelerator hybrid machines are increasingly rounding local nodal volumes to different radiation ther - being deployed in radiation oncology worldwide. apy prescriptions [16–19]. Given the treatment volumes Diagnostic MRI scans are used as a supplement to CT for these patients are comparatively larger than prostate datasets for radiation treatment planning, however this patients, there is a need to create a new atlas set for these introduces systematic inaccuracies in the planning pro- patients, as well as the requirement for gender specific cess due to positional differences between the scans [4, atlases. There has been limited work in the literature on 5]. The MRI simulator affords the ability to scan patients sCT creation for larger pelvic treatment sites, with small in the treatment position and potentially leads to a reduc- groups of patient numbers and no consideration of the tion in the registration errors. However due to unavoid- differences in male and female pelvic anatomy [9, 14, able differences in patient positioning between the two 20–22]. scans and inherent registration uncertainties, alignment This work investigated the application of a hybrid errors are still present. These errors are estimated to be multi-atlas approach for sCT creation for male and in the order of 2–4  mm for pelvic MRI to CT registra- female full pelvis treatments using forty patient datasets tions [6, 7]. This has resulted in increased research to acquired prospectively. Mean error and mean absolute incorporate MRI into radiotherapy planning, using MRI error in HU, volume comparisons and the Dice Similar- as primary imaging set rather than supplement. ity Co-efficient (DSC) were used to assess the anatomical The treatment planning system (TPS) uses electron accuracy while percentage dose difference at a reference density, resulting from the photoelectric effect and point, gamma dose comparison and dose volume histo- Compton scattering, to calculate the dosimetry. Unlike gram analysis for relevant structures was used to assess CT, the greyscale units of the MRI image do not cor- dosimetric accuracy of the sCT generation. relate with the electron density of tissues, therefore the treatment planning system (TPS) is unable to accurately model the dose deposition on a MRI scan. Researchers Materials and methods have developed various methods to create a synthetic CT Patient data collection scan (sCT) from the MRI scan in order to estimate the Ethics approval for the study was obtained through the electron densities of structures, allowing for dose calcula- local health district human research ethics committee tions [5]. There has been an increasing focus in machine (ref:17/06/21/3.02). Forty-one patients receiving radia- learning methods of sCT creation, however, current tion treatment for histologically confirmed malignancy applications in the clinic have relied on atlas based meth- of either the rectum, anal canal, cervix or endometrium, ods [8–12]; this approach has matured and has been gave informed consent to participate in the trial. One validated across multiple sites as compared to machine patient was excluded after insufficient coverage of the learning methods. MRI scan due to user error. The remaining forty par - The atlas based approach of generating sCT, which has ticipants were separated into male (n = 20) and female been successfully translated to the clinic, involves the (n = 20) cohorts for the creation of gender specific creation of an atlas of matching CT and MRI pairs. The atlases. MRI scans in the atlas are deformed to the new MRI and CT scans were acquired on a SOMATOM Confidence the deformation vectors are then applied to the corre- CT scanner (Siemens Healthineers; Erlangen, Germany) sponding CT pairs. This was initially performed using a at 120  kV with 2.0  mm slice thickness. Patients were single (average) scan pair atlas [13]. Later Dowling et  al. positioned supine, legs flat, using a CIVCO vac-lok bag [10] and Arabi et al. [14] further improved on atlas-based (CIVCO Medical Instruments; Iowa, USA) under their sCT generation by presenting a hybrid approach in which legs. All patients were scanned with a full bladder and a library of CT-MRI pairs is used, combined with local empty rectum, and oral or intravenous contrast was weighting of atlas patch values to create the sCT scan. administered at the radiation oncologist discretion. Scan Prostate cancer has been the focus of the MRI plan- range included the whole lumbar spine superiorly, to mid ning research due to its prevalence, small target, and femur inferiorly. Three positioning tattoos were used to the lack of complex anatomical interfaces [10, 15]. The aid in patient setup for treatment alignment. O ’Connor et al. Radiation Oncology (2022) 17:55 Page 3 of 11 MRI scans were performed following the planning Table 1 MRI acquisition parameters CT scan (mean 17.6 ± 13.0 (1 SD) minutes between Parameter T1 VIBE DIXON scans), on a MAGNETOM Skyra 3T MRI scanner (Sie- Scan type VIBE DIXON mens Healthineers; Erlangen, Germany), to ensure TE (ms) 1.23/2.46 similar bowel and bladder filling. The MRI scanner was TR (ms) 4.19 equipped with a Qfix flat couch (Qfix; Pennsylvania Flip angle 9° USA) and DORADOnova MR 3T external laser bridge FOV (mm) 256 * 499 (LAP; Luneburg, Germany). Patients were positioned by Slice thickness (mm) 1.6 a radiation therapist and a MRI radiographer, using their Base resolution 160 custom vac-lok bag. Patients were aligned using the posi- Acquisition plane Coronal tional tattoos and the external laser bridge. A 32 channel Phase direction R > L spine coil was utilised under the flat couch top and two Bandwidth (Hz/px) 1200 18 channel body coils were used over the pelvic region. Fat–water shift (px) 0.3 To avoid compression of the external body contour, one Distortion correction 3D body coil was positioned in a Qfix INSIGHT MR Body Acquisition stages 2 coil holder and placed over the superior portion of the Overlap (mm) 48 field, while the second coil was positioned with the supe - rior edge on the inferior edge of the coil bridge and the Composing Inline inferior edge on sponges (Fig. 1). For sCT generation, an additional T1 VIBE DIXON formed using Siemens adaptive algorithm, with sequence was added to the patient’s scanning proto- 48 mm of overlap, col of a small field of view T2 weighted sequence, and 3. A high receive bandwidth of 1200 Hz/pixel was used included the entire lumbar spine to mid femur, similar to to reduce the fat water shift to a sub pixel level (0.3 the field of view of the CT. VIBE is a volumetric imag - pixel). ing technique, which is a fast 3D gradient-echo sequence, producing a T1-weighted image. The T1 VIBE DIXON The T1 VIBE DIXON scan was acquired in the coro - sequence parameters are outlined in Table 1. To minimise nal plane, with an isotropic voxel size of 1.6 mm, as inline magnetic field related distortion in the MRI sequence for composing of an axial acquisition resulted in uneven sCT creation: slice thickness and missing slices at the overlap junction. The Siemens adaptive algorithm uses elastic matching to 1. Vendor supplied 3D distortion correction software correct for distortion caused by magnetic field inhomo - was applied, geneity [23]. The phase encoding direction ran right to 2. The scan was acquired in two stages in the coronal left and was extended to 195% of the read field of view, plane. Inline composing was automatically per- to allow for patient’s hips to be included laterally. The composed scans were then reconstructed axially for sCT creation and imported into the TPS. Treatment planning was performed as per department protocol on the CT scan using the Eclipse TPS (version 15.6; Varian Medical Systems). Three patients in the male cohort were planned as 6-MV, 7–9 field sliding-window Intensity Modulated Radiation Therapy (IMRT) while all other patients were planned as 6-MV, 2–3 arc Volumetric Modulated Arc Therapy (VMAT). sCT creation A leave-one-out cross validation approach was used to generate sCTs for each of the male and female groups (i.e. 19 patients were used to generate a sCT for each target patient MRI). The sCT generation method was similar to that used in Fig. 1 Patient positioning for pelvic MRI using 2 × 18 channel body Dowling et  al. [10] with some modifications to account coils with MRI coil bridge and sponges for the larger field of view and female anatomy. All MRIs O’Connor et al. Radiation Oncology (2022) 17:55 Page 4 of 11 were pre-processed with N4 bias field correction [24] co-registered and propagated CT-MR scans in the atlas with background masked to 0. All CT scans had their dataset. background masked to − 1000. Bone and bladder struc- tures were contoured on the CT and MRI by a radiation sCT validation therapist. For atlas generation, the CT was registered to The T1 VIBE DIXON MRI and sCT were imported into the matching MRI using structure guided registration the TPS. For each subject, the sCT was co-registered (using binary labels based on the bone and bladder con- to the MRI and the CT using rigid registration, with tours) for both rigid and non-rigid registration using cus- a registration boundary of the top of L2 to the greater tom code written in simpleITK [25]. For converting each trochanter. Co-registration of the sCT and MRI was target MRI to sCT, the MRI body contour was required required, as the frame of reference information was to help guide an initial rigid registration from each of the stripped from the data set during sCT generation. Due to 19 atlas cases due to the comparatively larger superior- a disparity in location of bowel gas on the CT and MRI inferior coverage of the data sets. This was followed by scans, bowel gas in the proximity of the treatment region deformable registration. The custom initial registration was overridden to average surrounding tissue HU value was written using the SimpleITK (https:// simpl eitk. org/) on both the CT and sCT for ten patients. The body was library and registered distance maps (SimpleITK Signed- contoured using image thresholding on the CT and sCT MaurerDistanceMapImageFilter) from the combined and used as the calculation volume for the correspond- binary labels from the bladder and bones from each ing data set. Two patients had a large discrepancy in body modality. These distance maps were initially registered contour of > 4 cm in the lateral posterior region, between using a rigid registration (6 DoF, metric = MSE), fol- the CT and MRI, due to tensing of the gluteal muscles in lowed by a Fast Symmetric Forces Demons Registration CT. This region of discrepancy in patient positioning was (standard deviation = 1). The transform and deformation removed from the sCT calculation volume for dosimetric fields from these steps were then applied to initial mov - analysis alone, so as to not affect the results. An in-house ing CT image to initialize the multi-modal registration. HU to electron density curve (Siemens BR38 kernel) was Finally the initial CT-MR registration results was refined applied to the CT and sCT. The CT based treatment plan, with a final step of non-rigid registration using NiftyReg International Commission of Radiation Units and Meas- ( h t t p : / / c m i c t i g . c s . u c l . a c . u k / w i k i / i n d e x . p h p / N i f t y R e g) urements (ICRU) reference point and structure set were reg_f3d with default parameters (free form deformation, copied from the original CT to the sCT [26]. The struc - multiscale scale approach, metric = normalized mutual tures were copied using the rigid registration between information). the CT and sCT. The treatment plan was then re-calcu - Each MRI to MRI registration was performed ini- lated with identical monitor unit values. tially using the body masks only using reg-aladin from Dosimetric accuracy was assessed using the CT based NiftyReg with default parameters apart from -rigOnly plan as the gold-standard. The dose difference at ICRU (6 degrees of freedom) and -rmask and -fmask with the reference point and dose volume histogram (DVH) respective body masks. Following this step, the moving analysis for relevant planning target volume (PTV) and MR image was propagated using the same transform and organ at risk (OAR) structures were assessed. The rel - then deformably registered to the target MRI (using the evant DVH parameters used for these structures were ITK diffeomorphic demons registration implementation as per standard guidelines for each treatment site (see (3 standard deviations, 3 multi-resolution levels). Additional file  1 for greater detail on DVH param- The co-registered CT scans were propagated using the eters assessed) [16–19]. Several DVH parameters were same deformation fields to the target MRI and then local evaluated for each structure, the average dose differ - weighted voting was applied to generate the final sCT ence for each structure is a combined average of each volume (using a radius of 2 voxels and a gain of 1). The of these parameters per structure. The percentage dose radius parameter defined the size of the patches used in difference was calculated by the formula (D − D )/ sCT CT the atlas-based local weighted voting process: the radius D * 100%. Statistical significance of the dose differ - CT is the offset from the centre voxel. For example, in 3D, ence at ICRU reference point was determined using a a radius of 2 results in a 5 × 5 × 5 patch of voxels. The Wilcoxon Signed-Rank Test with a significance level gain parameter was used to increase sensitivity with the of 0.05. Three-dimensional gamma analysis was used similarity measure between patches (increasing the gain to evaluate the dose impact of the sCT on the treat- can help differentiate patches with very similar inten - ment plan across the entire treatment volume. 3D sity values). The computed weighted similarity between gamma analysis was performed using an in-house patches in the registered MRI to the target MRI were MATLAB code (MATLAB; MathWorks), using a dose- used to combine the patches in the same location, from difference (%) and distance to agreement (mm) criteria O ’Connor et al. Radiation Oncology (2022) 17:55 Page 5 of 11 Dose impact of 3%/2  mm, 2%/2  mm, and 1%/1  mm. An erosion of There were no statistically significant dose difference at 15  mm of the body perimeter was applied to exclude reference point between the treatment plan calculated failures which occurred at skin edge due to small una- on the CT and sCT for the female cohort, while there voidable differences in body contour between data sets, was a statistically significant difference for the male. The and a 10% low dose threshold was applied. median percentage dose difference at the ICRU refer - Hounsfield Unit accuracy was assessed using mean ence point was − 0.4% (interquartile range (IQR) of 0.0 error (ME) and mean absolute error (MAE) in the to − 0.6, p = 0.01) in the male cohort, and − 0.3% (IQR of entire body, bone regions and soft tissue regions for 0.0 to − 0.6, p = 0.10) in the female cohort. The median each cohort, to assess the accuracy of the atlas-based DVH percentage dose difference of all DVH parameters sCT model. For ME and MAE calculations, the superior combined was − 0.2% (IQR of 0.2 to − 0.7, p = < 0.05) and inferior 3 cm of the sCT data sets was excluded to for the male cohort and − 0.4% (IQR of − 0.1 to − 0.9, avoid regions of image degradation due to differences in p = < 0.05) for the female cohort (Fig.  2). See Additional scan coverage in the atlas sets, and the density override file 2 for detailed separated DVH dose difference results. of the bowel gas was not applied for these calculations. The 3D gamma results with criteria of 3%/2 mm for all Due to differences in the body outline between the patients were within the American Association of Phys- sCT and CT, the body MAE and ME calculations were ics in Medicine (AAPM) TG218 report guidelines of performed within the MRI body contour between the > 95% (Table  3) [27]. Figure  3 visually represents the 3D registered CT-MRI and the sCT. To assess anatomical gamma with criteria of 1%/1 mm for the worst perform- accuracy a Dice similarity coefficient (DSC = 2[A ∩ B]/ ing gamma (patient 41). The area of colour wash in row [A + B]) of body and bone regions between the MRI C represents the regions which do not meet the gamma and sCT was calculated, and volume comparisons of criteria, this is occurring in a high dose region close to bone and body structures were performed between the a steep dose drop off and penumbra region at the lower sCT, CT. aspect of the field, mostly in the inferior aspect of the PTV high region. The DVH in row D shows similarity in dose to structures between the CT plan and sCT plan Results overlayed. Detailed patient demographics are outlined in Table  2. Of the 40 patients recruited to the trial, two patients in the male cohort had previous rectal resections, and six patients in the female cohort had previous hyster- Anatomical accuracy ectomies. Four patients in the male cohort and eleven Agreement in the average Dice similarity co-efficient for patients in the female cohort received iodine based the bone and body regions, between the MRI and sCT is oral contrast, while one patient in the female cohort shown in Table  4. The volume comparisons between the received iodine based IV contrast. sCT and CT resulted in a − 2.1% and − 1.4% whole body Table 2 Patient demographics Cohort size Age range BMI range (kg/ Relevant surgical Primary Staging range Prescribed dose m ) history treatment site Male cohort 20 49–88 (mean = 65) 20.5–33.6 Hernia repairs Rectum (n = 20) T1N0–T4N1 60 Gy/30fx (n = 1) (mean = 25.5) (n = 3) 50.4 Gy/28fx (n = 1) Rectal resections 50 Gy/25fx (n = 18) (n = 2) Appendectomy (n = 1) Female cohort 20 41–85 (mean = 61) 18.0–36.9 Hysterectomy Rectum (n = 4) T3N0–T3N2 50 Gy/25fx (n = 4) (mean = 26.2) (n = 6) Anal Canal (n = 4) T1N0–T3N1 54 Gy/30fx (n = 2) Common iliac 50.4 Gy/28fx (n = 1) stent (n = 1) 50 Gy/25fx (n = 1) Caesarean (n = 1) Cervix (n = 8) IIA–IIB 55 Gy/25fx (n = 1) Hernia Repair 50 Gy/25fx (n = 4) (n = 2) 45 Gy/25fx (n = 3) Appendectomy (n = 3) Endometrium IIIA–IIIC 54 Gy/30fx (n = 1) (n = 4) 50 Gy/25fx (n = 1) 45 Gy/25fx (n = 2) O’Connor et al. Radiation Oncology (2022) 17:55 Page 6 of 11 Fig. 2 Percentage DVH dose difference by structure (each structure parameters combined) for male and female cohorts. PTV High = Planning target volume higher prescribed dose, PTV Low = Planning target volume lower prescribed dose, GTV = Gross tumour volume, CTV High = Clinical target volume higher prescribed dose, CTV Low = Clinical target volume lower prescribed dose, RT NOF = Right neck of femur, LT NOF = Left neck of femur Table 3 3D Gamma analysis results for male cohort (n = 20) and female cohort (n = 20) (mean ± 1 SD) 3%/2 mm 2%/2 mm 1%/1 mm Pass rate (%) Av Gamma Pass rate (%) Av Gamma Pass rate (%) Av Gamma Male cohort 99.8 ± 0.2 0.10 ± 0.03 99.7 ± 0.3 0.15 ± 0.04 96.3 ± 3.1 0.31 ± 0.09 Range 100.0–99.3 0.07–0.18 100.0–99.0 0.11–0.25 99.2–88.7 0.21–0.52 Female cohort 99.8 ± 0.3 0.13 ± 0.04 99.7 ± 0.4 0.19 ± 0.05 93.4 ± 5.2 0.38 ± 0.12 Range 100.0–99.1 0.08–0.19 100.0–98.8 0.11–0.28 99.1–81.0 0.23–0.57 point of 1.5%. The same group furthered this work by volume difference and a − 3.1% and − 4.1% bone volume using a multi-atlas hybrid approach for sCT genera- difference for the male and female cohort respectively. tion on a similar cohort to the original study, and the The mean absolute error in HU of the body between results greatly improved [10]. This study applies Dowl - the CT and sCT was 59.1 ± 7.2 for the male cohort and ing et  al.’s [10] multi-atlas hybrid approach to greater 53.3 ± 8.9 for the female cohort (Fig. 4). The mean abso - pelvic regions and male and female cohorts. The lute error in the bone regions for the male and female reported MAE and ME in HU of the body contours was cohort was 166.7 ± 19.8 and 171.2 ± 26.6 respectively. of greater accuracy in Dowling et  al.’s [10] study than both cohorts in this study, however the dose differ - Discussion ence at ICRU reference point of − 0.3% ± 0.8%, DVH The results presented in this article are comparable to dose difference < 0.5% and gamma results > 95.0% at previous studies on MRI only planning for pelvic treat- 1%/1  mm were very similar to this study. While Dowl- ments. Dowling et  al. [13] originally applied a single ing et  al. previously used a T2 weighted image, this atlas approach to sCT generation for prostate cancer study utilised a T1-weighted VIBE DIXON imaging treatments, with a dose difference at ICRU reference O ’Connor et al. Radiation Oncology (2022) 17:55 Page 7 of 11 Fig. 3 Results for patient 41 (worst performing gamma). Row A Original patient CT scan with dose overlayed. Row B sCT with dose overlayed. Row C CT scan and critical structure outlines with Gamma map overlayed (1%/1 mm) colour wash showing regions which do not meet the gamma pass rate (values between 1 and − 1 not displayed). Row D Dose volume histogram results technique, which is increasingly favoured in recent of having an in-phase, out-of- phase, fat-weighted and MRI planning studies due to; better anatomical defini - water-weighted image sets. tion of T1-weighting; typically shorter echo times (TE), Studies which have investigated atlas-based MRI only to include tissues with short TE properties; reduced planning for larger pelvic treatment sites have had rela- scan time for a larger field of view; as well as the benefit tively low patient numbers compared to this study, and O’Connor et al. Radiation Oncology (2022) 17:55 Page 8 of 11 Table 4 DSC and volume comparison for body and bone structures. Mean absolute error and mean error in HU (mean ± 1 SD) sCT versus MRI sCT versus CT DSC Vol. difference (%) MAE (HU) ME (HU) Male Female Male Female Male Female Male Female Body 0.97 ± 0.0 0.98 ± 0.0 − 2.1 ± 2.0 − 1.4 ± 1.8 59.1 ± 7.2 53.3 ± 8.9 − 18.8 ± 11.0 − 16.7 ± 14.3 Bone 0.91 ± 0.2 0.96 ± 0.0 − 3.1 ± 2.9 − 4.1 ± 2.1 166.7 ± 19.8 171.2 ± 26.6 − 118.5 ± 33.6 − 129.1 ± 34.1 CT computed tomography, sCT synthetic CT, MRI magnetic resonance imaging, DSC dice similarity co-efficient, MAE mean absolute error, ME mean error, HU hounsfield unit did not focus on optimising sCT methods for separate well. The anatomical accuracy measured with the DSC genders. Arabi et  al. [14] utilised an atlas based sCT between the MRI and sCT was high, with the bony anat- generation with local weighted voting using a T1 VIBE omy showing the greatest variation in scores. The cases DIXON MRI on 12 patients with rectal cancer (2 female, showing a lower DSC score for bone regions could be 10 male). Using a single atlas for both genders, Arabi due to the difficulty in identifying and contouring bone et al. [14] achieved a bone Dice of 0.89 and the OAR dose regions on MRI, introducing some inaccuracies, as well difference mean was less than 0.9%, with the gamma cri - as the ability of the atlas to account for greater varia- teria of 2%/2  mm at 99.86% ± 0.27%, and 1%/1  mm at tions in anatomy from the atlas sets. This factor could 97.67% ± 3.60%. be resolved with a greater number of data sets within Other studies investigating MRI planning for rectal and the atlas to represent a greater variety of anatomy differ - gynaecological treatments have focused on a tissue class ences. In this study it is difficult to isolate the bony disa - segmentation approach. Maspero et  al. [21], and Kemp- greement as the reason for lower dosimetry agreement painen et  al. [20] both utilised a commercially avail- between the CT and sCT due to several compounding able product by Philips healthcare, MRCAT. Both studies factors affecting dosimetry. Further work could be done found similar results when applying this method to male to isolate the bony anatomy to determine the effect it has and female cohorts. Maspero et  al. applied MRCAT to on dosimetry alone. fifteen male and five female patients with rectal malig - Although an attempt was made to account for una- nancies, the resulting gamma pass rate at 2%/2  mm was voidable differences between the CT and MRI, such as 94.7% ± 1.7%, and on average a mean increase of 0.3% adjusting the body contour for set up variations; other to the dose to target [21]. Kemppainen et  al. [20] found variations which may affect results, such as the pres - better agreement using MRCAT software for rectal and ence of oral contrast being greater in the female cohort gynaecological treatment, with median relative dose dif- (55%) than the male cohort (20%) and previous surgeries ference to PTV less than 0.8% and the median relative can affect results. We did not attempt to control these, dose difference to OARs was less than 1.2%. A higher so as to mirror routine clinical presentations. Although gamma pass rate was found with a criteria of 2%/2  mm, it is conceivable that this could lead to different pass with the rectum cohort and gynaecological cohort being rates, however, most differences were found at the field 99.3% and 99.2% respectively. edges bearing the penumbra regions, which is a known Liu et  al. [9] applied a shape model for bone with a area of failure with CT based planning. An attempt was modified probabilistic tissue classification using shape also made to control the difference in bowel gas place - classification to T1 VIBE DIXON images of 10 female ment between the CT and MRI by performing a den- patients with pelvic malignancies, resulting in a maximal sity override of bowel gas on both the CT and sCT for mean dose difference of 0.3  Gy (0.5%). Wang et  al. [22] dosimetric analysis alone. Importance was not placed on applied a bone mask and tissue class segmentation to the sCT ability to accurately recreate bowel gas from the VIBE DIXON images of 11 patients with rectal cancer (9 MRI scan in this study, due to the variability in bowel gas female, 2 male). The reported median dose difference in placement between simulation and treatment day to day. the target volume was 0.3% and the median gamma pass Due to this, it is practice in our department to override rate was higher than 99% for 2%/2 mm criteria. bowel gas on simulation scans for treatment planning to In this study we reported a high level of agreement in account for this day to day variability, and therefore this dosimetry and gamma pass rate for both male and female practice was mirrored in the analysis of the sCT. cohorts, which compares equivalently to the above men- In this study, a stitched T1 VIBE DIXON sequence tioned MRI planning studies for the same body region. was utilised for sCT generation. To minimise distortion The sCT created from both atlases performed equally effects associated with the magnetic field inhomogeneity O ’Connor et al. Radiation Oncology (2022) 17:55 Page 9 of 11 Fig. 4 HU difference results for the best and worst performing gamma. Column A Original T1 VIBE DIXON in‑phase MRI from patient 16 (best gamma) and patient 41 (worst gamma); B sCT generated from MRI; C planning CT scan; D HU difference between sCT and CT and patient susceptibility a read-out bandwidth of distortion effects. On phantom testing, geometric distor - 1200 Hz/Px was selected, with a fat–water shift of 0.3px; tion, non-linear gradient fields and distortion relating to below 1  mm. Furthermore a two stage acquisition with bandwidth was measured in the order of 1–2 mm for the overlapping stages to scan at isocentre and 3D distortion field of view used in this study. As the geometrical uncer - correction was applied to reduce gradient non-linearity tainties in the MRI translate to the sCT, this analysis of O’Connor et al. Radiation Oncology (2022) 17:55 Page 10 of 11 Availability of data and materials the sCT also addresses the effect this inherent distortion The datasets used during the study are available from the corresponding has when carried through to treatment planning. author on reasonable request. All of factors mentioned were found to minimally affect the DVH dose impact to the target volumes and OARs Declarations and be within acceptable guidelines. We found this Ethics approval and consent to participate method to be relatively simple to implement as part of an Ethics approval for this study was granted by the Hunter New England Human MRI planning workflow, using a standard MRI sequence Research Ethics Committee, study Reference Number: 17/06/21/3.02. All and with a high level of DSC agreement in the bony con- patients gave written and informed consent to participate in this study. tours which is also important for image guided radiation Consent for publication therapy treatment. Written, informed consent was obtained from all participants in this trial. Competing interests The authors declare that they have no competing interests. Conclusion This study has shown that a multi-atlas based method for Author details Department of Radiation Oncology, Calvary Mater Hospital, Cnr Edith & Platt sCT created from routinely employed MRI sequences can St, Waratah, , Newcastle, NSW 2298, Australia. School of Health Sciences, be used for definitive pelvic radiotherapy planning for University of Newcastle, University Drive, Newcastle, NSW 2308, Australia. male and female patients. The implications of this study Australian E‑Health Research Centre, Commonwealth Scientific and Industrial Research Organisation (CSIRO), Bowen Bridge Rd, Herston, QLD 4029, Australia. means that MRI planning can be applied more broadly to School of Mathematical and Physical Sciences, University of Newcastle, male and female cohorts and more treatment regions in University Drive, Newcastle, NSW 2308, Australia. Department of Radiology, the pelvis, therefore greatly expanding the scope of MRI Calvary Mater Hospital, Edith Street, Waratah, Newcastle, NSW 2298, Australia. only planning. Received: 2 June 2021 Accepted: 3 March 2022 Abbreviations CT: Computed tomography; DSC: Dice similarity coefficient; DVH: Dose volume histogram; HU: Hounsfield units; ICRU : International Commission of References Radiation Units and Measurements; IQR: Interquartile range; IMRT: Intensity 1. Khan FM. The physics of radiation therapy. 4th ed. Baltimore: Lippincott modulated radiation therapy; MRI: Magnetic resonance imaging; MAE: Mean Williams & Wilkins; 2009. absolute error; ME: Mean error; OAR: Organ at risk; PTV: Planning target vol‑ 2. Kessler ML, Pitluck S, Petti P, Castro JR. Integration of multimodality imag‑ ume; sCT: Synthetic CT; TPS: Treatment planning system; VIBE: Volumetric inter‑ ing data for radiotherapy treatment planning. Int J Radiat Oncol Biol Phys. polated breath‑hold examination; VMAT: Volumetric modulated arc therapy. 1991;21(6):1653–67. 3. Metcalfe P, Liney GP, Holloway L, Walker A, Barton M, Delaney GP, et al. The potential for an enhanced role for MRI in radiation‑therapy treatment Supplementary Information planning. Technol Cancer Res Treat. 2013;12(5):429–46. The online version contains supplementary material available at https:// doi. 4. Dirix P, Haustermans K, Vandecaveye V. The value of magnetic org/ 10. 1186/ s13014‑ 022‑ 02023‑4. resonance imaging for radiotherapy planning. Semin Radiat Oncol. 2014;24(3):151–9. 5. Greer PB, Dowling JA, Lambert JA, Fripp J, Parker J, Denham JW, et al. A Additional file 1. DVH data collection table. Organ at risk and target magnetic resonance imaging‑based workflow for planning radiation volume DVH parameters assessed by treatment site. therapy for prostate cancer. Med J Aust. 2011;194(4):S24–7. Additional file 2. Detailed DVH results table. Organ at risk and target 6. Dean CJ, Sykes JR, Cooper RA, Hatfield P, Carey B, Swift S, et al. An evalua‑ volume DVH dose difference separated by individual parameters for male tion of four CT‑MRI co ‑registration techniques for radiotherapy treatment and female cohorts. Note: Some structure parameters are not included as planning of prone rectal cancer patients. Br J Radiol. 2012;85(1009):61–8. sample size is too small for individual analysis. 7. Krempien RC, Daeuber S, Hensley FW, Wannenmacher M, Harms W. Image fusion of CT and MRI data enables improved target volume definition in 3D ‑brachytherapy treatment planning. Brachytherapy. Acknowledgements 2003;2(3):164–71. Not applicable. 8. Maspero M, Savenije MHF, Dinkla AM, Seevinck PR, Intven MPW, Jurgenliemk‑Schulz IM, et al. Dose evaluation of fast synthetic‑ CT genera‑ Authors’ contributions tion using a generative adversarial network for general pelvis MR‑ only LOC was involved in the study design, patient recruitment, data collection, radiotherapy. Phys Med Biol. 2018;63(18):185001. data analysis and wrote the manuscript. JD developed the atlas base synthetic 9. Liu L, Jolly S, Cao Y, Vineberg K, Fessler JA, Balter JM. Female pelvic CT creation method used in this study. JD, JM, PBG, HW‑F, HR contributed synthetic CT generation based on joint intensity and shape analysis. Phys to the study design data analysis and revised the manuscript critically for Med Biol. 2017;62(8):2935–49. important intellectual content. JHC contributed to the formal analysis. LB and 10. Dowling JA, Sun J, Pichler P, Rivest‑Henault D, Ghose S, Richardson H, KS assisted in the development and ongoing MRI imaging of participants. GG, et al. Automatic substitute computed tomography generation and SS, MK assisted in participant recruitment. All authors read and approved the contouring for magnetic resonance imaging (MRI)‑alone external beam final manuscript. radiation therapy from standard MRI sequences. Int J Radiat Oncol Biol Phys. 2015;93(5):1144–53. Funding 11. MR‑ only RT planning for brain and pelvis with Synthetic CT · White paper: This study was funded by the Calvary Mater, Margaret Mitchell research grant Siemens Healthcare GmbH; 2019. scheme. The funding body had no role in the design of the study, collection, 12. Köhler M, Vaara T, Grootel MV, Hoogeveen R, Kemppainen R, Renisch analysis, interpretation of data or writing of the, manuscript. S. MR‑ only simulation for radiotherapy planning. White paper: Philips O ’Connor et al. Radiation Oncology (2022) 17:55 Page 11 of 11 MRCAT for prostate dose calculations using only MRI data: Koninklijke Philips N.V.; 2015. 13. Dowling JA, Lambert J, Parker J, Salvado O, Fripp J, Capp A, et al. An atlas‑ based electron density mapping method for magnetic resonance imag‑ ing (MRI)‑alone treatment planning and adaptive MRI‑based prostate radiation therapy. Int J Radiat Oncol Biol Phys. 2012;83(1):e5‑11. 14. Arabi H, Koutsouvelis N, Rouzaud M, Miralbell R, Zaidi H. Atlas‑ guided generation of pseudo‑ CT images for MRI‑ only and hybrid PET‑MRI‑ guided radiotherapy treatment planning. Phys Med Biol. 2016;61(17):6531–52. 15. Johnstone E, Wyatt JJ, Henry AM, Short SC, Sebag‑Montefiore D, Murray L, et al. Systematic review of synthetic computed tomography generation methodologies for use in magnetic resonance imaging‑ only radiation therapy. Int J Radiat Oncol Biol Phys. 2018;100(1):199–217. 16. NSW Governement, Cancer Institute NSW. Clinical resource: colorectal rectum neoadjuvant EBRT chemoradiation pre‑ operative long‑ course V.5 2018. https:// www. eviq. org. au/ radia tion‑ oncol ogy/ color ectal/ 1863‑ rectal‑ neoad juvant‑ ebrt‑ chemo radia tion‑ pre‑ op . 17. NSW Governement, Cancer Institute NSW. Clinical resource: gynaeco‑ logical endometrium adjuvant EBRT V.6 2019. www. eviq. org. au/ radia tion‑ oncol ogy/ gynae colog ical/ 233‑ gynae colog ical‑ endom etrium‑ adjuv ant‑ ebrt. 18. NSW Governement, Cancer Institute NSW. Clinical resource: colorectal anal carcnoma definitive EBRT chemoradiation IMRT V.4 2018. www. eviq. org. au/ radia tion‑ oncol ogy/ color ectal/ 1860‑ color ectal‑ anal‑ carci noma‑ defin itive‑ ebrt‑ che . 19. NSW Governement, Cancer Institute NSW. Clinical resource: gynaecologi‑ cal cervix adjuvant EBRT 2019. www. eviq. org. au/ radia tion‑ oncol ogy/ gynae colog ical/ 1471‑ gynae colog ical‑ cervi‑ adjuv ant‑ ebrt. 20. Kemppainen R, Suilamo S, Tuokkola T, Lindholm P, Deppe MH, Keyrilainen J. Magnetic resonance‑ only simulation and dose calculation in external beam radiation therapy: a feasibility study for pelvic cancers. Acta Oncol. 2017;56(6):792–8. 21. Maspero M, Tyyger MD, Tijssen RHN, Seevinck PR, Intven MPW, van den Berg CAT. Feasibility of magentic resonance imaging‑ only rectum radio‑ therapy with a commerical synthetic computed tomography generation solution. Phys Imaging Radiat Oncol. 2018;7:58–64. 22. Wang H, Du K, Qu J, Chandarana H, Das IJ. Dosimetric evaluation of mag‑ netic resonance‑ generated synthetic CT for radiation treatment of rectal cancer. PLoS ONE. 2018;13(1):e0190883. 23. Aulesjord A, et al. syngo MR E11 Operator Manual ‑ Scanning and post ‑ processing. Erlangen: Siemens Healthcare; 2014. 24. Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, et al. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging. 2010;29(6):1310–20. 25. Lowekamp BC, Chen DT, Ibanez L, Blezek D. The design of SimpleITK. Front Neuroinform. 2013;7:45. 26. ICRU International Commission on Radiation Units and Measurements. Prescribing, recording, and reporting photon‑beam intensity‑modulated radiation therapy (IMRT ). ICRU report 83. J ICRU. 2010;10:1–106. 27. Miften M, Olch A, Mihailidis D, Moran J, Pawlicki T, Molineu A, et al. Toler‑ ance limits and methodologies for IMRT measurement‑based verifica‑ tion QA: recommendations of AAPM Task Group No. 218. Med Phys. 2018;45(4):e53–83. Publisher’s Note Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Radiation Oncology Springer Journals

Loading next page...
 
/lp/springer-journals/validation-of-an-mri-only-planning-workflow-for-definitive-pelvic-VjkiEQmwcq
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
1748-717X
DOI
10.1186/s13014-022-02023-4
Publisher site
See Article on Publisher Site

Abstract

Purpose: Previous work on Magnetic Resonance Imaging (MRI) only planning has been applied to limited treat‑ ment regions with a focus on male anatomy. This research aimed to validate the use of a hybrid multi‑atlas synthetic computed tomography (sCT ) generation technique from a MRI, using a female and male atlas, for MRI only radiation therapy treatment planning of rectum, anal canal, cervix and endometrial malignancies. Patients and methods: Forty patients receiving radiation treatment for a range of pelvic malignancies, were sepa‑ rated into male (n = 20) and female (n = 20) cohorts for the creation of gender specific atlases. A multi‑atlas local weighted voting method was used to generate a sCT from a T1‑ weighted VIBE DIXON MRI sequence. The original treatment plans were copied from the CT scan to the corresponding sCT for dosimetric validation. Results: The median percentage dose difference between the treatment plan on the CT and sCT at the ICRU refer ‑ ence point for the male cohort was − 0.4% (IQR of 0 to − 0.6), and − 0.3% (IQR of 0 to − 0.6) for the female cohort. The mean gamma agreement for both cohorts was > 99% for criteria of 3%/2 mm and 2%/2 mm. With dose criteria of 1%/1 mm, the pass rate was higher for the male cohort at 96.3% than the female cohort at 93.4%. MRI to sCT anatomi‑ cal agreement for bone and body delineated contours was assessed, with a resulting Dice score of 0.91 ± 0.2 (mean ± 1 SD) and 0.97 ± 0.0 for the male cohort respectively; and 0.96 ± 0.0 and 0.98 ± 0.0 for the female cohort respec‑ tively. The mean absolute error in Hounsfield units (HUs) within the entire body for the male and female cohorts was 59.1 HU ± 7.2 HU and 53.3 HU ± 8.9 HU respectively. Conclusions: A multi‑atlas based method for sCT generation can be applied to a standard T1‑ weighted MRI sequence for male and female pelvic patients. The implications of this study support MRI only planning being applied more broadly for both male and female pelvic sites. Trial registration This trial was registered in the Australian New Zealand Clinical Trials Registry (ANZCTR) (www. anzctr. org. au) on 04/10/2017. Trial identifier ACTRN12617001406392. Keywords: MRI radiotherapy planning, Radiotherapy, Rectum neoplasms, Cervix neoplasms, Endometrium neoplasms, Anal canal neoplasms, Synthetic CT, Computer assisted radiotherapy planning, Image guided radiotherapy, Intensity modulated radiotherapy Background Computed tomography (CT) is the long established imaging modality used for radiation therapy treatment *Correspondence: Laura.OConnor@calvarymater.org.au Department of Radiation Oncology, Calvary Mater Hospital, Cnr Edith & planning. The inherent electron density [derived from Platt St, Waratah, , Newcastle, NSW 2298, Australia grey scale Hounsfield units (HU)] and anatomical data Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. O’Connor et al. Radiation Oncology (2022) 17:55 Page 2 of 11 from the CT scan, is used by commercial computer treatment volumes for colorectal and gynaecological planning systems to model and calculate radiation dose cancers are much more multifarious, traversing a more distribution within the patient’s body, using specific cal - variable body contour and bony anatomy than pros- culation algorithms [1]. The limited soft tissue contrast tate treatments. Rectum, anal canal and gynaecological and tumour delineation however, has given rise to the treatments routinely involve treating the gross tumour increasing use of Magnetic Resonance Imaging (MRI) in volume, surrounding tissue deemed to be at high risk of this context [2, 3]. As such, dedicated MRI scanners and tumour spread, the disease positive nodes and the sur- MRI-linear accelerator hybrid machines are increasingly rounding local nodal volumes to different radiation ther - being deployed in radiation oncology worldwide. apy prescriptions [16–19]. Given the treatment volumes Diagnostic MRI scans are used as a supplement to CT for these patients are comparatively larger than prostate datasets for radiation treatment planning, however this patients, there is a need to create a new atlas set for these introduces systematic inaccuracies in the planning pro- patients, as well as the requirement for gender specific cess due to positional differences between the scans [4, atlases. There has been limited work in the literature on 5]. The MRI simulator affords the ability to scan patients sCT creation for larger pelvic treatment sites, with small in the treatment position and potentially leads to a reduc- groups of patient numbers and no consideration of the tion in the registration errors. However due to unavoid- differences in male and female pelvic anatomy [9, 14, able differences in patient positioning between the two 20–22]. scans and inherent registration uncertainties, alignment This work investigated the application of a hybrid errors are still present. These errors are estimated to be multi-atlas approach for sCT creation for male and in the order of 2–4  mm for pelvic MRI to CT registra- female full pelvis treatments using forty patient datasets tions [6, 7]. This has resulted in increased research to acquired prospectively. Mean error and mean absolute incorporate MRI into radiotherapy planning, using MRI error in HU, volume comparisons and the Dice Similar- as primary imaging set rather than supplement. ity Co-efficient (DSC) were used to assess the anatomical The treatment planning system (TPS) uses electron accuracy while percentage dose difference at a reference density, resulting from the photoelectric effect and point, gamma dose comparison and dose volume histo- Compton scattering, to calculate the dosimetry. Unlike gram analysis for relevant structures was used to assess CT, the greyscale units of the MRI image do not cor- dosimetric accuracy of the sCT generation. relate with the electron density of tissues, therefore the treatment planning system (TPS) is unable to accurately model the dose deposition on a MRI scan. Researchers Materials and methods have developed various methods to create a synthetic CT Patient data collection scan (sCT) from the MRI scan in order to estimate the Ethics approval for the study was obtained through the electron densities of structures, allowing for dose calcula- local health district human research ethics committee tions [5]. There has been an increasing focus in machine (ref:17/06/21/3.02). Forty-one patients receiving radia- learning methods of sCT creation, however, current tion treatment for histologically confirmed malignancy applications in the clinic have relied on atlas based meth- of either the rectum, anal canal, cervix or endometrium, ods [8–12]; this approach has matured and has been gave informed consent to participate in the trial. One validated across multiple sites as compared to machine patient was excluded after insufficient coverage of the learning methods. MRI scan due to user error. The remaining forty par - The atlas based approach of generating sCT, which has ticipants were separated into male (n = 20) and female been successfully translated to the clinic, involves the (n = 20) cohorts for the creation of gender specific creation of an atlas of matching CT and MRI pairs. The atlases. MRI scans in the atlas are deformed to the new MRI and CT scans were acquired on a SOMATOM Confidence the deformation vectors are then applied to the corre- CT scanner (Siemens Healthineers; Erlangen, Germany) sponding CT pairs. This was initially performed using a at 120  kV with 2.0  mm slice thickness. Patients were single (average) scan pair atlas [13]. Later Dowling et  al. positioned supine, legs flat, using a CIVCO vac-lok bag [10] and Arabi et al. [14] further improved on atlas-based (CIVCO Medical Instruments; Iowa, USA) under their sCT generation by presenting a hybrid approach in which legs. All patients were scanned with a full bladder and a library of CT-MRI pairs is used, combined with local empty rectum, and oral or intravenous contrast was weighting of atlas patch values to create the sCT scan. administered at the radiation oncologist discretion. Scan Prostate cancer has been the focus of the MRI plan- range included the whole lumbar spine superiorly, to mid ning research due to its prevalence, small target, and femur inferiorly. Three positioning tattoos were used to the lack of complex anatomical interfaces [10, 15]. The aid in patient setup for treatment alignment. O ’Connor et al. Radiation Oncology (2022) 17:55 Page 3 of 11 MRI scans were performed following the planning Table 1 MRI acquisition parameters CT scan (mean 17.6 ± 13.0 (1 SD) minutes between Parameter T1 VIBE DIXON scans), on a MAGNETOM Skyra 3T MRI scanner (Sie- Scan type VIBE DIXON mens Healthineers; Erlangen, Germany), to ensure TE (ms) 1.23/2.46 similar bowel and bladder filling. The MRI scanner was TR (ms) 4.19 equipped with a Qfix flat couch (Qfix; Pennsylvania Flip angle 9° USA) and DORADOnova MR 3T external laser bridge FOV (mm) 256 * 499 (LAP; Luneburg, Germany). Patients were positioned by Slice thickness (mm) 1.6 a radiation therapist and a MRI radiographer, using their Base resolution 160 custom vac-lok bag. Patients were aligned using the posi- Acquisition plane Coronal tional tattoos and the external laser bridge. A 32 channel Phase direction R > L spine coil was utilised under the flat couch top and two Bandwidth (Hz/px) 1200 18 channel body coils were used over the pelvic region. Fat–water shift (px) 0.3 To avoid compression of the external body contour, one Distortion correction 3D body coil was positioned in a Qfix INSIGHT MR Body Acquisition stages 2 coil holder and placed over the superior portion of the Overlap (mm) 48 field, while the second coil was positioned with the supe - rior edge on the inferior edge of the coil bridge and the Composing Inline inferior edge on sponges (Fig. 1). For sCT generation, an additional T1 VIBE DIXON formed using Siemens adaptive algorithm, with sequence was added to the patient’s scanning proto- 48 mm of overlap, col of a small field of view T2 weighted sequence, and 3. A high receive bandwidth of 1200 Hz/pixel was used included the entire lumbar spine to mid femur, similar to to reduce the fat water shift to a sub pixel level (0.3 the field of view of the CT. VIBE is a volumetric imag - pixel). ing technique, which is a fast 3D gradient-echo sequence, producing a T1-weighted image. The T1 VIBE DIXON The T1 VIBE DIXON scan was acquired in the coro - sequence parameters are outlined in Table 1. To minimise nal plane, with an isotropic voxel size of 1.6 mm, as inline magnetic field related distortion in the MRI sequence for composing of an axial acquisition resulted in uneven sCT creation: slice thickness and missing slices at the overlap junction. The Siemens adaptive algorithm uses elastic matching to 1. Vendor supplied 3D distortion correction software correct for distortion caused by magnetic field inhomo - was applied, geneity [23]. The phase encoding direction ran right to 2. The scan was acquired in two stages in the coronal left and was extended to 195% of the read field of view, plane. Inline composing was automatically per- to allow for patient’s hips to be included laterally. The composed scans were then reconstructed axially for sCT creation and imported into the TPS. Treatment planning was performed as per department protocol on the CT scan using the Eclipse TPS (version 15.6; Varian Medical Systems). Three patients in the male cohort were planned as 6-MV, 7–9 field sliding-window Intensity Modulated Radiation Therapy (IMRT) while all other patients were planned as 6-MV, 2–3 arc Volumetric Modulated Arc Therapy (VMAT). sCT creation A leave-one-out cross validation approach was used to generate sCTs for each of the male and female groups (i.e. 19 patients were used to generate a sCT for each target patient MRI). The sCT generation method was similar to that used in Fig. 1 Patient positioning for pelvic MRI using 2 × 18 channel body Dowling et  al. [10] with some modifications to account coils with MRI coil bridge and sponges for the larger field of view and female anatomy. All MRIs O’Connor et al. Radiation Oncology (2022) 17:55 Page 4 of 11 were pre-processed with N4 bias field correction [24] co-registered and propagated CT-MR scans in the atlas with background masked to 0. All CT scans had their dataset. background masked to − 1000. Bone and bladder struc- tures were contoured on the CT and MRI by a radiation sCT validation therapist. For atlas generation, the CT was registered to The T1 VIBE DIXON MRI and sCT were imported into the matching MRI using structure guided registration the TPS. For each subject, the sCT was co-registered (using binary labels based on the bone and bladder con- to the MRI and the CT using rigid registration, with tours) for both rigid and non-rigid registration using cus- a registration boundary of the top of L2 to the greater tom code written in simpleITK [25]. For converting each trochanter. Co-registration of the sCT and MRI was target MRI to sCT, the MRI body contour was required required, as the frame of reference information was to help guide an initial rigid registration from each of the stripped from the data set during sCT generation. Due to 19 atlas cases due to the comparatively larger superior- a disparity in location of bowel gas on the CT and MRI inferior coverage of the data sets. This was followed by scans, bowel gas in the proximity of the treatment region deformable registration. The custom initial registration was overridden to average surrounding tissue HU value was written using the SimpleITK (https:// simpl eitk. org/) on both the CT and sCT for ten patients. The body was library and registered distance maps (SimpleITK Signed- contoured using image thresholding on the CT and sCT MaurerDistanceMapImageFilter) from the combined and used as the calculation volume for the correspond- binary labels from the bladder and bones from each ing data set. Two patients had a large discrepancy in body modality. These distance maps were initially registered contour of > 4 cm in the lateral posterior region, between using a rigid registration (6 DoF, metric = MSE), fol- the CT and MRI, due to tensing of the gluteal muscles in lowed by a Fast Symmetric Forces Demons Registration CT. This region of discrepancy in patient positioning was (standard deviation = 1). The transform and deformation removed from the sCT calculation volume for dosimetric fields from these steps were then applied to initial mov - analysis alone, so as to not affect the results. An in-house ing CT image to initialize the multi-modal registration. HU to electron density curve (Siemens BR38 kernel) was Finally the initial CT-MR registration results was refined applied to the CT and sCT. The CT based treatment plan, with a final step of non-rigid registration using NiftyReg International Commission of Radiation Units and Meas- ( h t t p : / / c m i c t i g . c s . u c l . a c . u k / w i k i / i n d e x . p h p / N i f t y R e g) urements (ICRU) reference point and structure set were reg_f3d with default parameters (free form deformation, copied from the original CT to the sCT [26]. The struc - multiscale scale approach, metric = normalized mutual tures were copied using the rigid registration between information). the CT and sCT. The treatment plan was then re-calcu - Each MRI to MRI registration was performed ini- lated with identical monitor unit values. tially using the body masks only using reg-aladin from Dosimetric accuracy was assessed using the CT based NiftyReg with default parameters apart from -rigOnly plan as the gold-standard. The dose difference at ICRU (6 degrees of freedom) and -rmask and -fmask with the reference point and dose volume histogram (DVH) respective body masks. Following this step, the moving analysis for relevant planning target volume (PTV) and MR image was propagated using the same transform and organ at risk (OAR) structures were assessed. The rel - then deformably registered to the target MRI (using the evant DVH parameters used for these structures were ITK diffeomorphic demons registration implementation as per standard guidelines for each treatment site (see (3 standard deviations, 3 multi-resolution levels). Additional file  1 for greater detail on DVH param- The co-registered CT scans were propagated using the eters assessed) [16–19]. Several DVH parameters were same deformation fields to the target MRI and then local evaluated for each structure, the average dose differ - weighted voting was applied to generate the final sCT ence for each structure is a combined average of each volume (using a radius of 2 voxels and a gain of 1). The of these parameters per structure. The percentage dose radius parameter defined the size of the patches used in difference was calculated by the formula (D − D )/ sCT CT the atlas-based local weighted voting process: the radius D * 100%. Statistical significance of the dose differ - CT is the offset from the centre voxel. For example, in 3D, ence at ICRU reference point was determined using a a radius of 2 results in a 5 × 5 × 5 patch of voxels. The Wilcoxon Signed-Rank Test with a significance level gain parameter was used to increase sensitivity with the of 0.05. Three-dimensional gamma analysis was used similarity measure between patches (increasing the gain to evaluate the dose impact of the sCT on the treat- can help differentiate patches with very similar inten - ment plan across the entire treatment volume. 3D sity values). The computed weighted similarity between gamma analysis was performed using an in-house patches in the registered MRI to the target MRI were MATLAB code (MATLAB; MathWorks), using a dose- used to combine the patches in the same location, from difference (%) and distance to agreement (mm) criteria O ’Connor et al. Radiation Oncology (2022) 17:55 Page 5 of 11 Dose impact of 3%/2  mm, 2%/2  mm, and 1%/1  mm. An erosion of There were no statistically significant dose difference at 15  mm of the body perimeter was applied to exclude reference point between the treatment plan calculated failures which occurred at skin edge due to small una- on the CT and sCT for the female cohort, while there voidable differences in body contour between data sets, was a statistically significant difference for the male. The and a 10% low dose threshold was applied. median percentage dose difference at the ICRU refer - Hounsfield Unit accuracy was assessed using mean ence point was − 0.4% (interquartile range (IQR) of 0.0 error (ME) and mean absolute error (MAE) in the to − 0.6, p = 0.01) in the male cohort, and − 0.3% (IQR of entire body, bone regions and soft tissue regions for 0.0 to − 0.6, p = 0.10) in the female cohort. The median each cohort, to assess the accuracy of the atlas-based DVH percentage dose difference of all DVH parameters sCT model. For ME and MAE calculations, the superior combined was − 0.2% (IQR of 0.2 to − 0.7, p = < 0.05) and inferior 3 cm of the sCT data sets was excluded to for the male cohort and − 0.4% (IQR of − 0.1 to − 0.9, avoid regions of image degradation due to differences in p = < 0.05) for the female cohort (Fig.  2). See Additional scan coverage in the atlas sets, and the density override file 2 for detailed separated DVH dose difference results. of the bowel gas was not applied for these calculations. The 3D gamma results with criteria of 3%/2 mm for all Due to differences in the body outline between the patients were within the American Association of Phys- sCT and CT, the body MAE and ME calculations were ics in Medicine (AAPM) TG218 report guidelines of performed within the MRI body contour between the > 95% (Table  3) [27]. Figure  3 visually represents the 3D registered CT-MRI and the sCT. To assess anatomical gamma with criteria of 1%/1 mm for the worst perform- accuracy a Dice similarity coefficient (DSC = 2[A ∩ B]/ ing gamma (patient 41). The area of colour wash in row [A + B]) of body and bone regions between the MRI C represents the regions which do not meet the gamma and sCT was calculated, and volume comparisons of criteria, this is occurring in a high dose region close to bone and body structures were performed between the a steep dose drop off and penumbra region at the lower sCT, CT. aspect of the field, mostly in the inferior aspect of the PTV high region. The DVH in row D shows similarity in dose to structures between the CT plan and sCT plan Results overlayed. Detailed patient demographics are outlined in Table  2. Of the 40 patients recruited to the trial, two patients in the male cohort had previous rectal resections, and six patients in the female cohort had previous hyster- Anatomical accuracy ectomies. Four patients in the male cohort and eleven Agreement in the average Dice similarity co-efficient for patients in the female cohort received iodine based the bone and body regions, between the MRI and sCT is oral contrast, while one patient in the female cohort shown in Table  4. The volume comparisons between the received iodine based IV contrast. sCT and CT resulted in a − 2.1% and − 1.4% whole body Table 2 Patient demographics Cohort size Age range BMI range (kg/ Relevant surgical Primary Staging range Prescribed dose m ) history treatment site Male cohort 20 49–88 (mean = 65) 20.5–33.6 Hernia repairs Rectum (n = 20) T1N0–T4N1 60 Gy/30fx (n = 1) (mean = 25.5) (n = 3) 50.4 Gy/28fx (n = 1) Rectal resections 50 Gy/25fx (n = 18) (n = 2) Appendectomy (n = 1) Female cohort 20 41–85 (mean = 61) 18.0–36.9 Hysterectomy Rectum (n = 4) T3N0–T3N2 50 Gy/25fx (n = 4) (mean = 26.2) (n = 6) Anal Canal (n = 4) T1N0–T3N1 54 Gy/30fx (n = 2) Common iliac 50.4 Gy/28fx (n = 1) stent (n = 1) 50 Gy/25fx (n = 1) Caesarean (n = 1) Cervix (n = 8) IIA–IIB 55 Gy/25fx (n = 1) Hernia Repair 50 Gy/25fx (n = 4) (n = 2) 45 Gy/25fx (n = 3) Appendectomy (n = 3) Endometrium IIIA–IIIC 54 Gy/30fx (n = 1) (n = 4) 50 Gy/25fx (n = 1) 45 Gy/25fx (n = 2) O’Connor et al. Radiation Oncology (2022) 17:55 Page 6 of 11 Fig. 2 Percentage DVH dose difference by structure (each structure parameters combined) for male and female cohorts. PTV High = Planning target volume higher prescribed dose, PTV Low = Planning target volume lower prescribed dose, GTV = Gross tumour volume, CTV High = Clinical target volume higher prescribed dose, CTV Low = Clinical target volume lower prescribed dose, RT NOF = Right neck of femur, LT NOF = Left neck of femur Table 3 3D Gamma analysis results for male cohort (n = 20) and female cohort (n = 20) (mean ± 1 SD) 3%/2 mm 2%/2 mm 1%/1 mm Pass rate (%) Av Gamma Pass rate (%) Av Gamma Pass rate (%) Av Gamma Male cohort 99.8 ± 0.2 0.10 ± 0.03 99.7 ± 0.3 0.15 ± 0.04 96.3 ± 3.1 0.31 ± 0.09 Range 100.0–99.3 0.07–0.18 100.0–99.0 0.11–0.25 99.2–88.7 0.21–0.52 Female cohort 99.8 ± 0.3 0.13 ± 0.04 99.7 ± 0.4 0.19 ± 0.05 93.4 ± 5.2 0.38 ± 0.12 Range 100.0–99.1 0.08–0.19 100.0–98.8 0.11–0.28 99.1–81.0 0.23–0.57 point of 1.5%. The same group furthered this work by volume difference and a − 3.1% and − 4.1% bone volume using a multi-atlas hybrid approach for sCT genera- difference for the male and female cohort respectively. tion on a similar cohort to the original study, and the The mean absolute error in HU of the body between results greatly improved [10]. This study applies Dowl - the CT and sCT was 59.1 ± 7.2 for the male cohort and ing et  al.’s [10] multi-atlas hybrid approach to greater 53.3 ± 8.9 for the female cohort (Fig. 4). The mean abso - pelvic regions and male and female cohorts. The lute error in the bone regions for the male and female reported MAE and ME in HU of the body contours was cohort was 166.7 ± 19.8 and 171.2 ± 26.6 respectively. of greater accuracy in Dowling et  al.’s [10] study than both cohorts in this study, however the dose differ - Discussion ence at ICRU reference point of − 0.3% ± 0.8%, DVH The results presented in this article are comparable to dose difference < 0.5% and gamma results > 95.0% at previous studies on MRI only planning for pelvic treat- 1%/1  mm were very similar to this study. While Dowl- ments. Dowling et  al. [13] originally applied a single ing et  al. previously used a T2 weighted image, this atlas approach to sCT generation for prostate cancer study utilised a T1-weighted VIBE DIXON imaging treatments, with a dose difference at ICRU reference O ’Connor et al. Radiation Oncology (2022) 17:55 Page 7 of 11 Fig. 3 Results for patient 41 (worst performing gamma). Row A Original patient CT scan with dose overlayed. Row B sCT with dose overlayed. Row C CT scan and critical structure outlines with Gamma map overlayed (1%/1 mm) colour wash showing regions which do not meet the gamma pass rate (values between 1 and − 1 not displayed). Row D Dose volume histogram results technique, which is increasingly favoured in recent of having an in-phase, out-of- phase, fat-weighted and MRI planning studies due to; better anatomical defini - water-weighted image sets. tion of T1-weighting; typically shorter echo times (TE), Studies which have investigated atlas-based MRI only to include tissues with short TE properties; reduced planning for larger pelvic treatment sites have had rela- scan time for a larger field of view; as well as the benefit tively low patient numbers compared to this study, and O’Connor et al. Radiation Oncology (2022) 17:55 Page 8 of 11 Table 4 DSC and volume comparison for body and bone structures. Mean absolute error and mean error in HU (mean ± 1 SD) sCT versus MRI sCT versus CT DSC Vol. difference (%) MAE (HU) ME (HU) Male Female Male Female Male Female Male Female Body 0.97 ± 0.0 0.98 ± 0.0 − 2.1 ± 2.0 − 1.4 ± 1.8 59.1 ± 7.2 53.3 ± 8.9 − 18.8 ± 11.0 − 16.7 ± 14.3 Bone 0.91 ± 0.2 0.96 ± 0.0 − 3.1 ± 2.9 − 4.1 ± 2.1 166.7 ± 19.8 171.2 ± 26.6 − 118.5 ± 33.6 − 129.1 ± 34.1 CT computed tomography, sCT synthetic CT, MRI magnetic resonance imaging, DSC dice similarity co-efficient, MAE mean absolute error, ME mean error, HU hounsfield unit did not focus on optimising sCT methods for separate well. The anatomical accuracy measured with the DSC genders. Arabi et  al. [14] utilised an atlas based sCT between the MRI and sCT was high, with the bony anat- generation with local weighted voting using a T1 VIBE omy showing the greatest variation in scores. The cases DIXON MRI on 12 patients with rectal cancer (2 female, showing a lower DSC score for bone regions could be 10 male). Using a single atlas for both genders, Arabi due to the difficulty in identifying and contouring bone et al. [14] achieved a bone Dice of 0.89 and the OAR dose regions on MRI, introducing some inaccuracies, as well difference mean was less than 0.9%, with the gamma cri - as the ability of the atlas to account for greater varia- teria of 2%/2  mm at 99.86% ± 0.27%, and 1%/1  mm at tions in anatomy from the atlas sets. This factor could 97.67% ± 3.60%. be resolved with a greater number of data sets within Other studies investigating MRI planning for rectal and the atlas to represent a greater variety of anatomy differ - gynaecological treatments have focused on a tissue class ences. In this study it is difficult to isolate the bony disa - segmentation approach. Maspero et  al. [21], and Kemp- greement as the reason for lower dosimetry agreement painen et  al. [20] both utilised a commercially avail- between the CT and sCT due to several compounding able product by Philips healthcare, MRCAT. Both studies factors affecting dosimetry. Further work could be done found similar results when applying this method to male to isolate the bony anatomy to determine the effect it has and female cohorts. Maspero et  al. applied MRCAT to on dosimetry alone. fifteen male and five female patients with rectal malig - Although an attempt was made to account for una- nancies, the resulting gamma pass rate at 2%/2  mm was voidable differences between the CT and MRI, such as 94.7% ± 1.7%, and on average a mean increase of 0.3% adjusting the body contour for set up variations; other to the dose to target [21]. Kemppainen et  al. [20] found variations which may affect results, such as the pres - better agreement using MRCAT software for rectal and ence of oral contrast being greater in the female cohort gynaecological treatment, with median relative dose dif- (55%) than the male cohort (20%) and previous surgeries ference to PTV less than 0.8% and the median relative can affect results. We did not attempt to control these, dose difference to OARs was less than 1.2%. A higher so as to mirror routine clinical presentations. Although gamma pass rate was found with a criteria of 2%/2  mm, it is conceivable that this could lead to different pass with the rectum cohort and gynaecological cohort being rates, however, most differences were found at the field 99.3% and 99.2% respectively. edges bearing the penumbra regions, which is a known Liu et  al. [9] applied a shape model for bone with a area of failure with CT based planning. An attempt was modified probabilistic tissue classification using shape also made to control the difference in bowel gas place - classification to T1 VIBE DIXON images of 10 female ment between the CT and MRI by performing a den- patients with pelvic malignancies, resulting in a maximal sity override of bowel gas on both the CT and sCT for mean dose difference of 0.3  Gy (0.5%). Wang et  al. [22] dosimetric analysis alone. Importance was not placed on applied a bone mask and tissue class segmentation to the sCT ability to accurately recreate bowel gas from the VIBE DIXON images of 11 patients with rectal cancer (9 MRI scan in this study, due to the variability in bowel gas female, 2 male). The reported median dose difference in placement between simulation and treatment day to day. the target volume was 0.3% and the median gamma pass Due to this, it is practice in our department to override rate was higher than 99% for 2%/2 mm criteria. bowel gas on simulation scans for treatment planning to In this study we reported a high level of agreement in account for this day to day variability, and therefore this dosimetry and gamma pass rate for both male and female practice was mirrored in the analysis of the sCT. cohorts, which compares equivalently to the above men- In this study, a stitched T1 VIBE DIXON sequence tioned MRI planning studies for the same body region. was utilised for sCT generation. To minimise distortion The sCT created from both atlases performed equally effects associated with the magnetic field inhomogeneity O ’Connor et al. Radiation Oncology (2022) 17:55 Page 9 of 11 Fig. 4 HU difference results for the best and worst performing gamma. Column A Original T1 VIBE DIXON in‑phase MRI from patient 16 (best gamma) and patient 41 (worst gamma); B sCT generated from MRI; C planning CT scan; D HU difference between sCT and CT and patient susceptibility a read-out bandwidth of distortion effects. On phantom testing, geometric distor - 1200 Hz/Px was selected, with a fat–water shift of 0.3px; tion, non-linear gradient fields and distortion relating to below 1  mm. Furthermore a two stage acquisition with bandwidth was measured in the order of 1–2 mm for the overlapping stages to scan at isocentre and 3D distortion field of view used in this study. As the geometrical uncer - correction was applied to reduce gradient non-linearity tainties in the MRI translate to the sCT, this analysis of O’Connor et al. Radiation Oncology (2022) 17:55 Page 10 of 11 Availability of data and materials the sCT also addresses the effect this inherent distortion The datasets used during the study are available from the corresponding has when carried through to treatment planning. author on reasonable request. All of factors mentioned were found to minimally affect the DVH dose impact to the target volumes and OARs Declarations and be within acceptable guidelines. We found this Ethics approval and consent to participate method to be relatively simple to implement as part of an Ethics approval for this study was granted by the Hunter New England Human MRI planning workflow, using a standard MRI sequence Research Ethics Committee, study Reference Number: 17/06/21/3.02. All and with a high level of DSC agreement in the bony con- patients gave written and informed consent to participate in this study. tours which is also important for image guided radiation Consent for publication therapy treatment. Written, informed consent was obtained from all participants in this trial. Competing interests The authors declare that they have no competing interests. Conclusion This study has shown that a multi-atlas based method for Author details Department of Radiation Oncology, Calvary Mater Hospital, Cnr Edith & Platt sCT created from routinely employed MRI sequences can St, Waratah, , Newcastle, NSW 2298, Australia. School of Health Sciences, be used for definitive pelvic radiotherapy planning for University of Newcastle, University Drive, Newcastle, NSW 2308, Australia. male and female patients. The implications of this study Australian E‑Health Research Centre, Commonwealth Scientific and Industrial Research Organisation (CSIRO), Bowen Bridge Rd, Herston, QLD 4029, Australia. means that MRI planning can be applied more broadly to School of Mathematical and Physical Sciences, University of Newcastle, male and female cohorts and more treatment regions in University Drive, Newcastle, NSW 2308, Australia. Department of Radiology, the pelvis, therefore greatly expanding the scope of MRI Calvary Mater Hospital, Edith Street, Waratah, Newcastle, NSW 2298, Australia. only planning. Received: 2 June 2021 Accepted: 3 March 2022 Abbreviations CT: Computed tomography; DSC: Dice similarity coefficient; DVH: Dose volume histogram; HU: Hounsfield units; ICRU : International Commission of References Radiation Units and Measurements; IQR: Interquartile range; IMRT: Intensity 1. Khan FM. The physics of radiation therapy. 4th ed. Baltimore: Lippincott modulated radiation therapy; MRI: Magnetic resonance imaging; MAE: Mean Williams & Wilkins; 2009. absolute error; ME: Mean error; OAR: Organ at risk; PTV: Planning target vol‑ 2. Kessler ML, Pitluck S, Petti P, Castro JR. Integration of multimodality imag‑ ume; sCT: Synthetic CT; TPS: Treatment planning system; VIBE: Volumetric inter‑ ing data for radiotherapy treatment planning. Int J Radiat Oncol Biol Phys. polated breath‑hold examination; VMAT: Volumetric modulated arc therapy. 1991;21(6):1653–67. 3. Metcalfe P, Liney GP, Holloway L, Walker A, Barton M, Delaney GP, et al. The potential for an enhanced role for MRI in radiation‑therapy treatment Supplementary Information planning. Technol Cancer Res Treat. 2013;12(5):429–46. The online version contains supplementary material available at https:// doi. 4. Dirix P, Haustermans K, Vandecaveye V. The value of magnetic org/ 10. 1186/ s13014‑ 022‑ 02023‑4. resonance imaging for radiotherapy planning. Semin Radiat Oncol. 2014;24(3):151–9. 5. Greer PB, Dowling JA, Lambert JA, Fripp J, Parker J, Denham JW, et al. A Additional file 1. DVH data collection table. Organ at risk and target magnetic resonance imaging‑based workflow for planning radiation volume DVH parameters assessed by treatment site. therapy for prostate cancer. Med J Aust. 2011;194(4):S24–7. Additional file 2. Detailed DVH results table. Organ at risk and target 6. Dean CJ, Sykes JR, Cooper RA, Hatfield P, Carey B, Swift S, et al. An evalua‑ volume DVH dose difference separated by individual parameters for male tion of four CT‑MRI co ‑registration techniques for radiotherapy treatment and female cohorts. Note: Some structure parameters are not included as planning of prone rectal cancer patients. Br J Radiol. 2012;85(1009):61–8. sample size is too small for individual analysis. 7. Krempien RC, Daeuber S, Hensley FW, Wannenmacher M, Harms W. Image fusion of CT and MRI data enables improved target volume definition in 3D ‑brachytherapy treatment planning. Brachytherapy. Acknowledgements 2003;2(3):164–71. Not applicable. 8. Maspero M, Savenije MHF, Dinkla AM, Seevinck PR, Intven MPW, Jurgenliemk‑Schulz IM, et al. Dose evaluation of fast synthetic‑ CT genera‑ Authors’ contributions tion using a generative adversarial network for general pelvis MR‑ only LOC was involved in the study design, patient recruitment, data collection, radiotherapy. Phys Med Biol. 2018;63(18):185001. data analysis and wrote the manuscript. JD developed the atlas base synthetic 9. Liu L, Jolly S, Cao Y, Vineberg K, Fessler JA, Balter JM. Female pelvic CT creation method used in this study. JD, JM, PBG, HW‑F, HR contributed synthetic CT generation based on joint intensity and shape analysis. Phys to the study design data analysis and revised the manuscript critically for Med Biol. 2017;62(8):2935–49. important intellectual content. JHC contributed to the formal analysis. LB and 10. Dowling JA, Sun J, Pichler P, Rivest‑Henault D, Ghose S, Richardson H, KS assisted in the development and ongoing MRI imaging of participants. GG, et al. Automatic substitute computed tomography generation and SS, MK assisted in participant recruitment. All authors read and approved the contouring for magnetic resonance imaging (MRI)‑alone external beam final manuscript. radiation therapy from standard MRI sequences. Int J Radiat Oncol Biol Phys. 2015;93(5):1144–53. Funding 11. MR‑ only RT planning for brain and pelvis with Synthetic CT · White paper: This study was funded by the Calvary Mater, Margaret Mitchell research grant Siemens Healthcare GmbH; 2019. scheme. The funding body had no role in the design of the study, collection, 12. Köhler M, Vaara T, Grootel MV, Hoogeveen R, Kemppainen R, Renisch analysis, interpretation of data or writing of the, manuscript. S. MR‑ only simulation for radiotherapy planning. White paper: Philips O ’Connor et al. Radiation Oncology (2022) 17:55 Page 11 of 11 MRCAT for prostate dose calculations using only MRI data: Koninklijke Philips N.V.; 2015. 13. Dowling JA, Lambert J, Parker J, Salvado O, Fripp J, Capp A, et al. An atlas‑ based electron density mapping method for magnetic resonance imag‑ ing (MRI)‑alone treatment planning and adaptive MRI‑based prostate radiation therapy. Int J Radiat Oncol Biol Phys. 2012;83(1):e5‑11. 14. Arabi H, Koutsouvelis N, Rouzaud M, Miralbell R, Zaidi H. Atlas‑ guided generation of pseudo‑ CT images for MRI‑ only and hybrid PET‑MRI‑ guided radiotherapy treatment planning. Phys Med Biol. 2016;61(17):6531–52. 15. Johnstone E, Wyatt JJ, Henry AM, Short SC, Sebag‑Montefiore D, Murray L, et al. Systematic review of synthetic computed tomography generation methodologies for use in magnetic resonance imaging‑ only radiation therapy. Int J Radiat Oncol Biol Phys. 2018;100(1):199–217. 16. NSW Governement, Cancer Institute NSW. Clinical resource: colorectal rectum neoadjuvant EBRT chemoradiation pre‑ operative long‑ course V.5 2018. https:// www. eviq. org. au/ radia tion‑ oncol ogy/ color ectal/ 1863‑ rectal‑ neoad juvant‑ ebrt‑ chemo radia tion‑ pre‑ op . 17. NSW Governement, Cancer Institute NSW. Clinical resource: gynaeco‑ logical endometrium adjuvant EBRT V.6 2019. www. eviq. org. au/ radia tion‑ oncol ogy/ gynae colog ical/ 233‑ gynae colog ical‑ endom etrium‑ adjuv ant‑ ebrt. 18. NSW Governement, Cancer Institute NSW. Clinical resource: colorectal anal carcnoma definitive EBRT chemoradiation IMRT V.4 2018. www. eviq. org. au/ radia tion‑ oncol ogy/ color ectal/ 1860‑ color ectal‑ anal‑ carci noma‑ defin itive‑ ebrt‑ che . 19. NSW Governement, Cancer Institute NSW. Clinical resource: gynaecologi‑ cal cervix adjuvant EBRT 2019. www. eviq. org. au/ radia tion‑ oncol ogy/ gynae colog ical/ 1471‑ gynae colog ical‑ cervi‑ adjuv ant‑ ebrt. 20. Kemppainen R, Suilamo S, Tuokkola T, Lindholm P, Deppe MH, Keyrilainen J. Magnetic resonance‑ only simulation and dose calculation in external beam radiation therapy: a feasibility study for pelvic cancers. Acta Oncol. 2017;56(6):792–8. 21. Maspero M, Tyyger MD, Tijssen RHN, Seevinck PR, Intven MPW, van den Berg CAT. Feasibility of magentic resonance imaging‑ only rectum radio‑ therapy with a commerical synthetic computed tomography generation solution. Phys Imaging Radiat Oncol. 2018;7:58–64. 22. Wang H, Du K, Qu J, Chandarana H, Das IJ. Dosimetric evaluation of mag‑ netic resonance‑ generated synthetic CT for radiation treatment of rectal cancer. PLoS ONE. 2018;13(1):e0190883. 23. Aulesjord A, et al. syngo MR E11 Operator Manual ‑ Scanning and post ‑ processing. Erlangen: Siemens Healthcare; 2014. 24. Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, et al. N4ITK: improved N3 bias correction. IEEE Trans Med Imaging. 2010;29(6):1310–20. 25. Lowekamp BC, Chen DT, Ibanez L, Blezek D. The design of SimpleITK. Front Neuroinform. 2013;7:45. 26. ICRU International Commission on Radiation Units and Measurements. Prescribing, recording, and reporting photon‑beam intensity‑modulated radiation therapy (IMRT ). ICRU report 83. J ICRU. 2010;10:1–106. 27. Miften M, Olch A, Mihailidis D, Moran J, Pawlicki T, Molineu A, et al. Toler‑ ance limits and methodologies for IMRT measurement‑based verifica‑ tion QA: recommendations of AAPM Task Group No. 218. Med Phys. 2018;45(4):e53–83. Publisher’s Note Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : Springer Nature remains neutral with regard to jurisdictional claims in pub‑ lished maps and institutional affiliations. fast, convenient online submission thorough peer review by experienced researchers in your field rapid publication on acceptance support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions

Journal

Radiation OncologySpringer Journals

Published: Mar 18, 2022

Keywords: MRI radiotherapy planning; Radiotherapy; Rectum neoplasms; Cervix neoplasms; Endometrium neoplasms; Anal canal neoplasms; Synthetic CT; Computer assisted radiotherapy planning; Image guided radiotherapy; Intensity modulated radiotherapy

References