Online Adaptive MRI-Guided Stereotactic Body Radiotherapy for Pancreatic and Other Intra-Abdominal Cancers

Simple Summary MRI can provide better visualization of tumors and nearby organs at risk (OAR) than CT for fast and accurate contouring during online adaptive MRI-guided stereotactic body radiation treatment (MRI-guided SBRT) for pancreatic and other intra-abdominal cancers. Pre-set MRI sequences provided in a 1.5T MRI scanner hybrid with a linear accelerator can be used during MRI-guided SBRT, but they often limit tumor and OAR visualization and require a long image acquisition time. This study retrospectively analyzed 26 patients with pancreatic and intra-abdominal cancers that underwent CT and MR simulations and 3–5 fractionated MRI-guided SBRT. The visualization of tumors and OAR was improved with T1W imaging, which is essential for online adaptive planning and resulted in fast and accurate contouring in a shorter imaging time. Abstract A 1.5T MRI combined with a linear accelerator (Unity®, Elekta; Stockholm, Sweden) is a device that shows promise in MRI-guided stereotactic body radiation treatment (SBRT). Previous studies utilized the manufacturer’s pre-set MRI sequences (i.e., T2 Weighted (T2W)), which limited the visualization of pancreatic and intra-abdominal tumors and organs at risk (OAR). Here, a T1 Weighted (T1W) sequence was utilized to improve the visualization of tumors and OAR for online adapted-to-position (ATP) and adapted-to-shape (ATS) during MRI-guided SBRT. Twenty-six patients, 19 with pancreatic and 7 with intra-abdominal cancers, underwent CT and MRI simulations for SBRT planning before being treated with multi-fractionated MRI-guided SBRT. The boundary of tumors and OAR was more clearly seen on T1W image sets, resulting in fast and accurate contouring during online ATP/ATS planning. Plan quality in 26 patients was dependent on OAR proximity to the target tumor and achieved 96 ± 5% and 92 ± 9% in gross tumor volume D90% and planning target volume D90%. We utilized T1W imaging (about 120 s) to shorten imaging time by 67% compared to T2W imaging (about 360 s) and improve tumor visualization, minimizing target/OAR delineation uncertainty and the treatment margin for sparing OAR. The average time-consumption of MRI-guided SBRT for the first 21 patients was 55 ± 15 min for ATP and 79 ± 20 min for ATS.


Introduction
Magnetic resonance imaging (MRI) combined with a linear accelerator (MRI-Linac) [1][2][3][4] yields a technique that shows promise in MRI-guided stereotactic body radiotherapy (SBRT) [5].MRI-Linac [6] provides superior visualization of target tumors and surrounding organs at risk (OAR) to improve delineation accuracy; MRI-guidance accounts for position, size, and shape Cancers 2023, 15, 5272 3 of 14 reference plan for online adaptive MRI-guided SBRT.Four-dimensional CT (4D-CT) image sets with an abdominal compression belt (ZiFix TM , Qfix, PA, USA) were used to measure tumor motion range [22].If tumor motion was equal to or less than 0.5 cm, the patient was eligible for MRI-guided SBRT and was asked to complete the second MRI screening sheet for MRI safety before the MRI simulation [31,32].On the same day as the CT simulation, MR image sets of T2W + Nav, T1-Weighted (T1W), and T1W + Fat Saturated (FS) were acquired using a 1.5T Unity ® , and one of them was chosen (typically T1W) to contour the target tumor and use it as a sequence of daily-MRI.The third step was SBRT planning (Figure 1c), during which a FB-CT was rigidly registered to a chosen T1W, and target tumors and OAR were contoured on a chosen T1W and FB-CT image set, respectively.A CT reference plan ( CT Ref) was then developed on a FB-CT for all patients.Additionally, an MRI reference plan ( MR Ref) was developed on a T1W for the first few patients.
ure tumor motion range [22].If tumor motion was equal to or less than 0.5 cm, the patient was eligible for MRI-guided SBRT and was asked to complete the second MRI screening sheet for MRI safety before the MRI simulation [31,32].On the same day as the CT simulation, MR image sets of T2W + Nav, T1-Weighted (T1W), and T1W + Fat Saturated (FS) were acquired using a 1.5T Unity ® , and one of them was chosen (typically T1W) to contour the target tumor and use it as a sequence of daily-MRI.The third step was SBRT planning (Figure 1c), during which a FB-CT was rigidly registered to a chosen T1W, and target tumors and OAR were contoured on a chosen T1W and FB-CT image set, respectively.A CT reference plan (CTRef) was then developed on a FB-CT for all patients.Additionally, an MRI reference plan (MRRef) was developed on a T1W for the first few patients.
The fourth step was treating the patient with online adaptive MRI-guided SBRT.In each fraction of MRI-guided SBRT (Figure 1d), 3 image sets for daily-MRI were acquired: 1 set for online adaptive planning (plan-MRI) and 2 sets for patient setup verifications before (verification-MRI) and after beam delivery (post-MRI).Orthogonal, sagittal, and coronal cine images were acquired to verify internal target volume (ITV) by measuring target motion range on MiM (v7.0.6,MiM Software Inc, Cleveland, OH, USA).In this study, the patient setup was maintained by using the same Unity ® couch top and MRIsafe or conditional immobilization devices during each CT and MRI simulation and multifractionated MRI-guided SBRT.The fourth step was treating the patient with online adaptive MRI-guided SBRT.In each fraction of MRI-guided SBRT (Figure 1d), 3 image sets for daily-MRI were acquired: 1 set for online adaptive planning (plan-MRI) and 2 sets for patient setup verifications before (verification-MRI) and after beam delivery (post-MRI).Orthogonal, sagittal, and coronal cine images were acquired to verify internal target volume (ITV) by measuring target motion range on MiM (v7.0.6,MiM Software Inc, Cleveland, OH, USA).In this study, the patient setup was maintained by using the same Unity ® couch top and MRI-safe or conditional immobilization devices during each CT and MRI simulation and multifractionated MRI-guided SBRT.

Patients
Patients were usually treated every other day using a 1.5T Unity ® .The imaging, planning, and treatment data of each patient's CT and MRI simulation and consecutive 3-5 fractionated MRI-guided SBRT were retrospectively analyzed.

CT and MRI Simulations with Immobilization Devices
On the same day, all patients underwent both a CT and an MRI simulation to acquire FB-CT images with and without a gadolinium-based contrast agent for OAR contouring, 4D-CT images for measuring target motion range, and multiple MR image sets of T2W + Nav, T1W, and T1W + FS for target tumor contouring.In addition, a 2D orthogonal coronal and sagittal image set (2D-Cine) was acquired to manually measure the range of target motion for individual patients and determine the margin of ITV.The target motion range was used to verify and adjust the initial target motion range measured in 4D-CT.An abdominal compression belt, which is safe for both CT and MRI scanning, was used for abdominal imaging to manage target and organ motion up to 0.5 cm in all directions induced by respiration [22].The pressure level of an abdominal compression belt for each patient, measured in a CT simulation, was used to set up that patient in an MRI simulation and throughout the entire MRI-guided SBRT.Two patients were scanned without the abdominal compression belt due to their discomfort.
Imaging parameters of the T1W and T1W + FS sequences were optimized during scans of volunteers and the first few patients in the study.The optimized MRI sequences were then used for all patients.For the MRI simulation, we used a T2W + Nav with a 3D turbo spin echo (TSE) MRI pulse sequence and a T1W and T1W + FS of a 3D turbo field echo (TFE) MRI pulse sequence on a 1.5T Unity ® , with 2 MRI receiver coils (a 4-channel anterior coil and a 4-channel posterior coil).Imaging parameters of T2W + Nav were repetition time (TR)/echo time (TE) = 1800/205 ms, field of view (FOV) = 400 × 400 mm 2 , pixel size = 1.56 × 1.56 mm 2 , image matrix = 480 × 480, thickness = 2.4 mm, flip angle = 90 • , bandwidth = 727 Hz, and number of signals (average = 5).Each MR image set took approximately 228 to 410 s; 233 images were acquired in total.A navigating window was set at the liver dome scout (1/3 on the lung side and 2/3 on the liver side).
Imaging parameters of T1W were TR/TE = 4.5/2.2ms, FOV = 400 × 400 mm 2 , pixel size = 1.1 × 1.1 mm 2 , image matrix = 280 × 280, thickness = 2 mm, flip angle = 10 On the MiM software, the OAR delineated on a FB-CT image set were heart, kidneys, liver, spinal cord, duodenum, small bowel, stomach, jejunum, colon, bone, body, spleen, and extras (i.e., air, contrast, vein, and celiac).Gross tumor volume (GTV), delineated on a T1W (or T1W + FS) MR image set, was determined for pancreatic tumors, left adrenal tumors, and pancreatic lymph nodes.The GTV contour was transferred from the T1W image set to the FB-CT image set for addition to the OAR contours.A FB-CT image set with all contours was exported to the treatment planning system (TPS, MR-Linac Monaco v5.51.11,CMS; St. Louis, MO, USA), and a patient-specific SBRT plan was developed.Each SBRT plan with 15-45 Gy in 3-5 fractions (Table 1) was calculated using Monte Carlo ® and used 7-14 beams delivered in a step-and-shoot intensity-modulated radiation therapy (IMRT) technique with 7FFF (flattening filter-free) photons.A reference plan developed on the FB-CT was exported to the oncology system (Mosaiq v2.8.3, CMS; St. Louis, MO, USA) for a treatment schedule every other day.
Table 1.The dose constraints of planning target volume (PTV) and organs at risk for stereotactic body radiotherapy for pancreatic and other intra-abdominal cancers.This is an example of 40 Gy in 5 fractions, so the dose constraints varied on the prescription dose (15 Gy to 45 Gy).

Organs
Dose Constraints (5 Fractions) of 40 Gy in 8 Gy × 5 Fractions A patient-specific quality assurance (QA) was performed for CT Ref and MR Ref plans and the first adaptive ATS plan ( MR ATS) to verify the applicability and deliverability of these QA plans on the Unity ® .This QA test was performed using an ion chamber (Exradin A1SLMR, Standard Imaging, Inc., Middleton, WI, USA) [33] and using an MRI-conditional cylindrical diode array dosimeter (ArcCHECK-MR, Sun Nuclear Corporation, Melbourne, FL, USA) [34].

Online Adaptive MRI-Guided SBRT
Patients were treated with 3-5 fractionated MRI-guided SBRT, using our Unity ® , between May 2020 and May 2023.For each fraction, 1 CT Ref or MR Ref plan was chosen for online adaptive ATP or ATS MRI-guided SBRT, and 1 or more MR ATS(s) was added to the list of reference plans for the next online adaptive MRI-guided SBRT.Three T1W (or T1W + FS) image sets were acquired during each fraction.The first T1W image set, the plan-MRI, was used to account for inter-fraction changes of targets and OAR shape, position, and size.After image fusion between the CT Ref (or MR Ref) and plan-MRI by a physicist (or a therapist), an attending physician determined ATP or ATS for further online adaptive planning.Then, GTV and OAR contours, rigidly or deformably transferred from the CT Ref (or MR Ref), were adjusted to match the plan-MRI.For ATS, GTV and the contours of the stomach, duodenum, small bowel, colon, and jejunum were usually adjusted or re-delineated by an attending physician, and the contours of the air and body were adjusted or removed by an attending physicist.
Next, 2 T1W image sets, the verification-MRI and post-MRI images, were acquired before and after the beam delivery to verify patient setup.Once the patient setup was verified, the target motion, moving within the planning target volume (PTV), was visually evaluated by the attending physician prior to the beam delivery.If there were patient setup differences between the plan-MRI and verification-MRI in GTV and OAR contours, the verification-MRI was used as the new plan-MRI to repeat online adaptive planning.A little over 10% of the fractions in all patients enrolled in this study experienced the changes in anatomical position and size found in the verification-MRI.

Statistical Analysis of Plan Quality
For each patient, we compared the image quality of T1W, FB-CT, and T2W + Nav image sets by inspecting tumor and OAR visualization on the plan-MRI acquired during 3-5 fractionated MRI-guided SBRT.The changes in tumor and OAR contours adjusted during ATS were quantified as a function of their variability in volume.Next, the quality of online adaptive SBRT plans was evaluated by quantifying the coverage of radiation dose to tumors and OAR.Both image quality and plan quality were evaluated for all patients using MiM.Lastly, the time-consumption of each step in our workflow was analyzed in individual, online adaptive ATP/ATS plans to determine our workflow's efficiency.

Results
All patients successfully completed CT and MRI simulations to acquire planning image sets and were treated with 3-5 fractionated online adaptive MRI-guided SBRT.

Patients
Twenty-six patients with pancreatic (n = 19), left adrenal (n = 3), and lymph node cancer (n = 4), who were treated using Unity ® between May 2021 and May 2023, were included in this study.Of the 26 patients, 16 were male and 10 were female, and the cohort had an average age of 71 years [range: 57-95] (Table 2).Average tumor volumes of reference plans ( CT Ref or MR Ref) and adaptive ATP/ATS plans were measured at 37.1 mL [range: 3.4 mL to 105.5 mL] and 35.4 mL [range: 3.4 mL to 106.4 mL], respectively.The same air-pressure of an abdominal compression belt recorded at the CT simulation was reproduced during the MRI simulation and consecutive 3-5 fractionated MRI-guided SBRT.

CT and MR Image Sets and Target Contouring
Figure 2 shows an example of pancreatic tumor and OAR visualization on FB-CT, T2W + Nav, and T1W image sets.Like FB-CT, the boundaries of a pancreatic tumor and OAR were well visualized on T1W, but not on T2W + Nav.Regarding tumor and OAR contouring, T1W images more clearly showed the boundaries of tumors in all 26 patients than the T2W + Nav images (Figure 3).The clear interface of organs required for higher accuracy and precision of SBRT planning was shown in T1W images, but it was unclear in T2W + Nav images.

CT and MR Image Sets and Target Contouring
Figure 2 shows an example of pancreatic tumor and OAR visualization on FB-CT, T2W + Nav, and T1W image sets.Like FB-CT, the boundaries of a pancreatic tumor and OAR were well visualized on T1W, but not on T2W + Nav.Regarding tumor and OAR contouring, T1W images more clearly showed the boundaries of tumors in all 26 patients than the T2W + Nav images (Figure 3).The clear interface of organs required for higher accuracy and precision of SBRT planning was shown in T1W images, but it was unclear in T2W + Nav images.The average GTV and PTV, measured in the CTRef or MRRef across all 26 patients, were 36.6 mL (range: 3.4 mL to 106.4 mL) and 74.9 mL (range: 7.1 mL to 181.4 mL), respectively.A target motion of 0.2-0.4cm was measured with 4D-CT or 2D-Cine and added to the GTV as a margin along all directions for ITV, accounting for respiratory-induced target motion when using the abdominal compression belt.

Plan Quality of Reference SBRT Plans
Figure 4 shows 2 examples of patient CTRef and MRRef plans.All QA testing CTRef, MRRef, and MRATS plans showed a >95% passing rate for the 3%-3mm gamma analysis and a <2% point dose difference between data measured by an ion chamber and data calculated by TPS.The average GTV and PTV, measured in the CT Ref or MR Ref across all 26 patients, were 36.6 mL (range: 3.4 mL to 106.4 mL) and 74.9 mL (range: 7.1 mL to 181.4 mL), respectively.A target motion of 0.2-0.4cm was measured with 4D-CT or 2D-Cine and added to the GTV as a margin along all directions for ITV, accounting for respiratory-induced target motion when using the abdominal compression belt.and MR ATS plans showed a >95% passing rate for the 3%-3mm gamma analysis and a <2% point dose difference between data measured by an ion chamber and data calculated by TPS.

Online Adaptive MRI-Guided SBRT
Superior visualization of tumors and OAR on T1W image sets contributed to minimizing contouring uncertainty and improving the efficiency of image fusion between a chosen reference plan and plan-MRI.Furthermore, the total acquisition time of plan-MRI, verification-MRI, and post-MRI was reduced three times from 360 s for T2W + Nav to 120 s for T1W (a 67% reduction of the total acquisition time).Figure 5 shows the visualization of tumors and OAR on 5 MR ATS plans created with 5 fractioned online adaptive plans.
The to 100.0%], respectively.The dose coverages of GTV and PTV were mainly dependent on the locations of OAR, such as duodenum or jejunum (n = 10), stomach (n = 7), small bowel or colon (n = 9), and spleen (n = 11).Other OAR, such as the liver, kidneys, and spinal cords, had a negligible effect on GTV and PTV dose coverages since the beams' gantry angles avoided these OAR.

Online Adaptive MRI-Guided SBRT
Superior visualization of tumors and OAR on T1W image sets contributed to minimizing contouring uncertainty and improving the efficiency of image fusion between a  chosen reference plan and plan-MRI.Furthermore, the total acquisition time of plan-MRI, verification-MRI, and post-MRI was reduced three times from 360 s for T2W + Nav to 120 s for T1W (a 67% reduction of the total acquisition time).Figure 5 shows the visualization of tumors and OAR on 5 MRATS plans created with 5 fractioned online adaptive plans.The difference in GTV and PTV contours in all MRATS(s), compared to CTRef (or MRRef), was minimal.The average GTV was 37.1 mL in all CTRef(s) and MRRef and 35.4 mL in all MRATS(s).Similarly, the average PTV was 74.1 mL in all CTRef(s) and MRRef and 72.1 mL in all MRATS(s).In addition, the dose coverages of GTV D95% and D90% for all CTRef(s) were 94.2% [range: 78.8% to 100.0%] and 96.9% [range: 86.1% to 100.0%], respectively.The dose coverages of PTV D95% and D90% were slightly lower at 89.7% [range: 69.3% to 99.7%] and 94% [range: 78.8% to 100%], respectively.For all MRATS(s), GTV/PTV D95% and D90% were approximately at 83.7% [range: 16.8% to 100.0%]/91.6%[range: 52.8% to 100.0%], respectively.The dose coverages of GTV and PTV were mainly dependent on the locations of OAR, such as duodenum or jejunum (n = 10), stomach (n = 7), small bowel or colon (n = 9), and spleen (n = 11).Other OAR, such as the liver, kidneys, and spinal cords, had a negligible effect on GTV and PTV dose coverages since the beams' gantry angles avoided these OAR.

Overall Time-Consumption of Online MRI-Guided SBRT in 10 Steps
The average time-consumption of online MRI-guided SBRT for the first 21 patients was 55 min for ATP and 79 min for ATS (Table 3).The most time-consuming steps were fusion/contouring and beam delivery (5 Gy to 9 Gy in each fraction with 7-13 beams), followed by patient setup, plan optimization, and plan QA/approval.All OAR coverages were achieved in all MR ATS(s).

Overall Time-Consumption of Online MRI-Guided SBRT in 10 Steps
The average time-consumption of online MRI-guided SBRT for the first 21 patients was 55 min for ATP and 79 min for ATS (Table 3).The most time-consuming steps were fusion/contouring and beam delivery (5 Gy to 9 Gy in each fraction with 7-13 beams), followed by patient setup, plan optimization, and plan QA/approval.

Discussion
SBRT planning requires that the tumor be superiorly visualized in CT or MR image sets to determine the radiation dose limit for OAR in patients with pancreatic and intraabdominal caners [5,9,[12][13][14].In this study, we utilized T1W imaging (about 120 s) to reduce the imaging time of plan-MRI, verification-MRI, and post-MRI by approximately 67%, compared to T2W + Nav imaging (about 360 s).T1W imaging also improved tumor visualization to (1) minimize delineation uncertainty, (2) reduce GTV, and (3) spare OAR during MRI-guided SBRT.We demonstrated the efficiency of T1W imaging by inspecting all images within the clear boundaries of contoured tumors and OAR.
Intra-and inter-fractional changes in a tumor's anatomical position and shape and a patient's setup are often found in image-guided radiotherapy, required re-planning, or online adaptive planning [35][36][37].To account for an inter-fractional change between a reference plan ( CT Ref, MR Ref, or MR ATS) and the plan-MRI, we performed online adaptive ATP or ATS planning, and we repeated adaptive planning within a fraction if we found an intra-fractional change between the plan-MRI and verification-MRI.To repeat adaptive planning, the latest verification-MRI was used as the plan-MRI and fused to a reference plan or the latest MR ATS, followed by contouring and/or plan optimization.More than one adaptive planning was required in 10% of all fractions, which could be improved by increasing patient comfort [38].For example, we minimized the time patients stayed on the treatment couch during online adaptive planning.T1W imaging can achieve fast imaging and efficient contouring of tumors and OAR, it still requires patient setup for those with pre-existing health conditions, causing discomfort and pain (i.e., surgery, injury, and claustrophobia), eliminating (or minimizing) the idle waiting time for the attending physician, and reducing the delivery time of uncomplicated SBRT plans [39][40][41].
The tumor contours delineated on T1W image sets matched the contours in FB-CT image sets because they account for the respiratory-induced motion during image acquisition.However, T2W + Nav image sets were acquired at exhalation in Unity ® , so the contour size of tumors was slightly smaller than it was in T1W image sets [42].Our ITV was determined at every fraction by encompassing GTV delineated on a T1W image set and motion range measured on orthogonal, sagittal, and coronal cine images in between inhalation and exhalation over the multiple breathing cycles (i.e., about 5-7 breathing cycles).This helps to clarify fractional breathing patterns for increasing the reliability of ITV margins in the presence of inter-fractional breathing variability [43].
Patient immobilization is critical for maintaining patient setup and preventing changes in target/OAR position and shape [32,44].A thin or medium-sized vac-bag was tested with the first 5 patients during patient setup, but this required a long setup time.The vac-bag also required the patient to rotate so that the ATS could be achieved during the long treatment time.Instead of a vac-bag, we used an abdominal compression belt to control the target motion moving within 5 mm [22].The setup consistency of an abdominal compression belt was dependent on the level of experience of individual therapists, and it was more consistent when the same therapist set it up every time across all simulations and multi-fractionated MRI-guided SBRT.
Our study has limitations.Limitations of this study include the basic analysis of image and plan quality using T1W imaging.However, our clinical protocol is continuously being improved to increase workflow efficiency using immobilization devices, MRI sequences, and auto-contouring and planning.Compared to our T1W sequence, the newly released MRI sequence (b3DVaneXD) and research sequences (compressed sensing and mDIXON) were not tested to compare tumor and OAR visualization.The present manual measurement of ITV margin may not be required when using a respiratory gating technique in the near future.This is an ongoing project at our institution and will include more quantitative analysis when compared with other MRI sequences, such as b3DVaneXD, compressed sensing, and mDIXON, to provide alternative imaging for less motion-dependent imaging, faster imaging, and fat/water suppression imaging, respectively.In addition, we will assess the complexity and dose coverage of our SBRT plans by comparing them with other plans using other imaging techniques for cross-validation.

Conclusions
This was the first study that utilized a customized and optimized T1W sequence to improve the visualization of tumors and OAR for reference planning and online adaptive planning on a 1.5T Unity ® .The tumor and OAR boundaries were clearly visible for delineation.Our results can facilitate consistent visualization of pancreatic and other intra-abdominal tumors to achieve fast and accurate MRI-guided SBRT.

Figure 1 .
Figure 1.The workflow of simulations and multi-fractionated stereotactic body radiotherapy for pancreatic and intra-abdominal cancers (MRI-guided SBRT) on a 1.5T Unity ® .(a) Patient selection and 1st MRI screening to check the eligibility of each patient; (b) CT and MRI simulations with an abdominal compression belt to acquire CT and MR image sets; (c) contouring organs at risk (OAR) on FB-CT images and target tumors on T1W images to develop a CT reference plan (CTRef) and an MRI reference plan (MRRef); and (d) MRI-guided SBRT in 3-5 fractions.An identical Unity ® couch top and the same abdominal compression belt were used in CT and MRI simulations and across all SBRT fractions.Orthogonal sagittal and coronal 2D cine images were acquired to measure the range of target motion induced by respiration for determining an internal target volume (ITV) margin.

Figure 1 .
Figure 1.The workflow of simulations and multi-fractionated stereotactic body radiotherapy for pancreatic and intra-abdominal cancers (MRI-guided SBRT) on a 1.5T Unity ® .(a) Patient selection and 1st MRI screening to check the eligibility of each patient; (b) CT and MRI simulations with an abdominal compression belt to acquire CT and MR image sets; (c) contouring organs at risk (OAR) on FB-CT images and target tumors on T1W images to develop a CT reference plan ( CT Ref) and an MRI reference plan ( MR Ref); and (d) MRI-guided SBRT in 3-5 fractions.An identical Unity ® couch top and the same abdominal compression belt were used in CT and MRI simulations and across all SBRT fractions.Orthogonal sagittal and coronal 2D cine images were acquired to measure the range of target motion induced by respiration for determining an internal target volume (ITV) margin.
tumor volume; PRV: Planning of organ at-risk volume.

Figure 2 .
Figure 2. The pancreatic tumor and organs at risk (OAR) visualized on T1W for contouring and further SBRT planning.A CT image set (a) acquired in free-breathing (FB) with Gadolinium was compared to a T2W with exhalation-navigating (+Nav) image set (b) and a T1W image set (c).The target tumor areas were magnified, and the target tumors were contoured on a CT image set with the yellow color and copied to MR image sets.The boundary of the target tumor was clearly visible on the T1W image set acquired in a short period of time, but it was unclear on the T2W + Nav image set.The body shape due to the use of an abdominal compression belt was identical across all FB-CT, T2W + Nav, and T1W image sets.

Figure 2 .
Figure 2. The pancreatic tumor and organs at risk (OAR) visualized on T1W for contouring and further SBRT planning.A CT image set (a) acquired in free-breathing (FB) with Gadolinium was compared to a T2W with exhalation-navigating (+Nav) image set (b) and a T1W image set (c).The target tumor areas were magnified, and the target tumors were contoured on a CT image set with the yellow color and copied to MR image sets.The boundary of the target tumor was clearly visible on the T1W image set acquired in a short period of time, but it was unclear on the T2W + Nav image set.The body shape due to the use of an abdominal compression belt was identical across all FB-CT, T2W + Nav, and T1W image sets.

Figure 3 .
Figure 3.Comparison of tumor visualization between CT and 4 T1W image sets.The same target tumors of 4 patients, colored in red, are shown in free-breathing (FB)-CT and T1W: (a) left adrenal, (b) pancreatic node, (c) large pancreatic mass, and (d) pancreatic body.The target tumors are clearly seen in both the FB-CT and T1W image sets.

Figure 3 .
Figure 3.Comparison of tumor visualization between CT and 4 T1W image sets.The same target tumors of 4 patients, colored in red, are shown in free-breathing (FB)-CT and T1W: (a) left adrenal, (b) pancreatic node, (c) large pancreatic mass, and (d) pancreatic body.The target tumors are clearly seen in both the FB-CT and T1W image sets.

Figure 4 shows 2
Figure 4 shows 2 examples of patient CT Ref and MR Ref plans.All QA testing CT Ref, MR Ref,and MR ATS plans showed a >95% passing rate for the 3%-3mm gamma analysis and a <2% point dose difference between data measured by an ion chamber and data calculated by TPS.

Figure 4 .
Figure 4.A CT reference plan ( CT Ref) of P12 and an MRI reference plan ( MR Ref) of P02 developed in a free-breathing (FB) CT image set (45 Gy = 9 Gy × 5 fractions with 12 beams) and a T1W image set (40 Gy = 8 Gy × 5 fractions with 13 beams), respectively.PTV and OAR (kidneys, liver, spleen, spinal cord, duodenum, bone, body (external), and small bowel) were contoured and shown in the FB-CT and T1W image sets.Both were successfully used for pancreatic and intra-abdominal cancers during 5 fractionated MRI-guided stereotactic body radiotherapy (SBRT).(a) The dose coverage of gross tumor volume (GTV) and planning target volume (PTV) in CT Ref achieved V 45Gy > 95% and V 45Gy > 93%, respectively.(b) The dose coverage of PTV High and PTV Low in MR Ref achieved V 40Gy > 92% and V 35Gy > 97%, respectively.

Figure 5 .
Figure5.Online adaptive MRI-guided stereotactic body radiotherapy for pancreatic and intra-abdominal cancers of (MRI-guided SBRT) for P07 using 35Gy in 5 fractions with 12 beams (95 years old, male, pancreas head).All MRI adapt-to-shape (MRATS) plans were created on T1W image sets.The target coverages of gross tumor volume (GTV) and planning target volume (PTV) were greater than 90% in 4 fractions (Fx#1, Fx#2, Fx#3, and Fx#5), except for Fx#4.Organs at risk (OAR), the duodenum and jejunum, were very close to the target tumor and resulted in less than 90% in Fx#4.All OAR coverages were achieved in all MRATS(s).

Figure 5 .
Figure5.Online adaptive MRI-guided stereotactic body radiotherapy for pancreatic and intraabdominal cancers of (MRI-guided SBRT) for P07 using 35Gy in 5 fractions with 12 beams (95 years old, male, pancreas head).All MRI adapt-to-shape ( MR ATS) plans were created on T1W image sets.The target coverages of gross tumor volume (GTV) and planning target volume (PTV) were greater than 90% in 4 fractions (Fx#1, Fx#2, Fx#3, and Fx#5), except for Fx#4.Organs at risk (OAR), the duodenum and jejunum, were very close to the target tumor and resulted in less than 90% in Fx#4.All OAR coverages were achieved in all MR ATS(s).

simulation image sets were used to develop 1 or more CT Ref or MR Ref plans for each patient. For the first 5 patients in the study, we developed both a CT Ref and a MR Ref, but we developed only a CT Ref for the rest of the patients. The T1W image sets were used to develop MR Ref plans through off-line adaptive ATS planning from the CT Ref
plans.Average electron densities of the tumor and OAR from the CT Ref were assigned to corresponding contours in the MR Ref.Both CT Ref and MR Ref plans were developed to meet the dose constraints listed in Table 1.One of the CT Ref and MR Ref plans was peer-reviewed in the department chart round at Radiation Oncology, Allegheny Health Network Cancer Center.

Table 3 .
Time-consumption of online adaptive MRI-guided stereotactic body radiotherapy for pancreatic and intra-abdominal cancers (SBRT) in 10 steps.The step of fusion and contouring included some waiting time of attending physicians.The step of plan QA and approval included an independent dose verification and a visual inspection of the target tumor moving within the PTV contour.

Table 3 .
Time-consumption of online adaptive MRI-guided stereotactic body radiotherapy for pancreatic and intra-abdominal cancers (SBRT) in 10 steps.The step of fusion and contouring included some waiting time of attending physicians.The step of plan QA and approval included an independent dose verification and a visual inspection of the target tumor moving within the PTV contour.