Geometric and dosimetric impact of anatomical changes for MR‐only radiation therapy for the prostate

Abstract Purpose With the move towards magnetic resonance imaging (MRI) as a primary treatment planning modality option for men with prostate cancer, it becomes critical to quantify the potential uncertainties introduced for MR‐only planning. This work characterized geometric and dosimetric intra‐fractional changes between the prostate, seminal vesicles (SVs), and organs at risk (OARs) in response to bladder filling conditions. Materials and methods T2‐weighted and mDixon sequences (3–4 time points/subject, at 1, 1.5 and 3.0 T with totally 34 evaluable time points) were acquired in nine subjects using a fixed bladder filling protocol (bladder void, 20 oz water consumed pre‐imaging, 10 oz mid‐session). Using mDixon images, Magnetic Resonance for Calculating Attenuation (MR‐CAT) synthetic computed tomography (CT) images were generated by classifying voxels as muscle, adipose, spongy, and compact bone and by assignment of bulk Hounsfield Unit values. Organs including the prostate, SVs, bladder, and rectum were delineated on the T2 images at each time point by one physician. The displacement of the prostate and SVs was assessed based on the shift of the center of mass of the delineated organs from the reference state (fullest bladder). Changes in dose plans at different bladder states were assessed based on volumetric modulated arc radiotherapy (VMAT) plans generated for the reference state. Results Bladder volume reduction of 70 ± 14% from the final to initial time point (relative to the final volume) was observed in the subject population. In the empty bladder condition, the dose delivered to 95% of the planning target volume (PTV) (D95%) reduced significantly for all cases (11.53 ± 6.00%) likely due to anterior shifts of prostate/SVs relative to full bladder conditions. D15% to the bladder increased consistently in all subjects (42.27 ± 40.52%). Changes in D15% to the rectum were patient‐specific, ranging from −23.93% to 22.28% (−0.76 ± 15.30%). Conclusions Variations in the bladder and rectal volume can significantly dislocate the prostate and OARs, which can negatively impact the dose delivered to these organs. This warrants proper preparation of patients during treatment and imaging sessions, especially when imaging required longer scan times such as MR protocols.


| INTRODUCTION
Prostate cancer is the most common type of cancer in men, with over 160 000 cases reported in 2017 in the United States. 1 The current treatment-planning workflow involves using computed tomography simulation (CT-SIM) as the primary planning modality. However, magnetic resonance imaging (MRI) has been shown to show superior accuracy to CT for identifying the prostate gland, the prostatic apex, and areas of high tumor burden. [2][3][4][5][6] By performing an MRI to CT rigid registration, the prostate can be delineated on MRI and then transferred to CT for subsequent planning. This co-registration may introduce uncertainties of 2-3 mm for prostate cancer. [7][8][9] Recently, as a means to circumvent this uncertainty and streamline the clinical workflow, MR-only planning has emerged in the clinic. For the male pelvis, two MR-only packages are currently clinically available for prostate cancer treatment planning with synthetic CTs [synCTs, or CTs generated from MRI input(s)]. One FDA-approved software package, the Philips Magnetic Resonance for Calculating Attenuation (MR-CAT), is based on a dual echo three-dimensional (3D) mDixon fast field echo sequence with synCTs generated on the scanner. 10,11 In a recent study by Farjam et al., 12 13 Comparison of the synCT and CT-based dose plans for prostate showed <1% difference in the mean absorbed dose to the PTV for the MR-CAT 14 and 0.0 ± 0.2% for the SDA methods. 13 It has been shown that substantial variations in the bladder volume occur during the course of treatment. 15,16 Importantly, the variations in the bladder filling adversely impact the dose delivered to the prostate over a standard course of radiotherapy for the prostate. 17,18 Huang et al. 19 used daily cone-beam computer tomography (CBCT) images to measure target/organ volumes and dosimetric differences in 28 prostate cancer patients and found mean percentage volume differences of 44% within the bladder volumes in the treatment plan which led to percentage dose difference of 2 ± 2% in the prostate.
As MRI emerges as a primary treatment planning modality option for prostate cancer, 20 it becomes important to quantify the potential uncertainties introduced in an MR-only workflow due to variable physiological status that may confound accurate dosimetry and highprecision radiation therapy. One of the main issues that arises with MRI is the longer scanning times as compared to CT, which may lead to higher variations in the bladder and rectal volumes. It is currently unknown how robust MR-only treatment planning is to internal anatomy changes nor how the dosimetry may be impacted. This work characterizes the temporal, spatial, and dosimetric intra-fractional changes between the prostate, seminal vesicles (SVs), and other organs at risk (OARs) in response to bladder filling conditions for MR-only prostate cancer radiation therapy planning.

2.B | Imaging protocol
T2-weighted turbo spin echo images were acquired since it is the most commonly used image for delineation of organs in the pelvis. 22 The imaging protocol also consisted of dual echo 3D FFE (Fast Field Echo) mDIXON sequences 11   . The mDixon scan is designed to yield high-geometric accuracy by using short echo times and high bandwidth. The advantage of using the two echoes in the mDIXON approach is to allow water, fat, and in-phase images to be derived within the same acquisition by using the frequency shift of the fat and water protons. 11 These images are inputted into the MR-CAT software to produce the Synthetic CT image used for treatment planning. 10 Briefly, MR-CAT automatically segments the external anatomy from background air using the water and in-phase images. Next, a model-based segmentation method is used to segment bone from the external contour based on training datasets. 11 Soft tissue is defined as voxels within the body volume and outside the segmented bone. 10,11,23 An intensity-based classification is then used to segment adipose and muscle within the soft tissue using the water and fat images. Finally, the bone voxels are divided into compact and spongy bone based on the voxel intensity of the in-phase image. In summary, MR-CAT categorizes the contents of the MR images into five classes (air, fat, water-rich tissue, spongy bone, and compact bone) and assigns to each voxel a bulk Hounsfield Unit value based on its classification, and the final synCT image is generated for treatment planning. 11 One current limitation of MR-CAT is that it does not account for rectal gas in the image classification. To fully elucidate the dosimetric impact of bladder and rectal status changes, the intestinal gas with each rectal contour was automatically thresholded and assigned a CT value of −350 HU based on values obtained from the literature. 24

2.C | Volumetric and geometric analysis
The prostate, SVs, bladder, and rectum were delineated on the T2-

2.D | Dosimetric analysis
Using MR-CAT images of the last time point (i.e., full bladder, which is consistent with our clinical practice), volumetric modulated arc radiotherapy (VMAT) plans were generated using two full arc beams with 6 MV photons. The treatment planning was designed to deliver 79.2 Gy to the PTV using RTOG 0815 dose constraints as a guideline. 25

2.E | Statistical analysis
Repeated measures mixed models containing fixed (time points) and random (subjects) effects were used to assess the significance of changes of dose at different bladder states (initial, middle, and final) while using the multiple/repeated measures on the same subject. If the effect of the bladder states was significant (P < 0.05), pairwise comparisons of three time points using the overall mean square error (MSE) was calculated.
To investigate the associations between organ displacement and bladder and rectum volumes, multilevel modeling methods 29 were used with the intercept coefficient as the random effect and the slopes (effect of changes of the bladder/rectum) as the fixed effect.

3.A | Volumetric and geometric analysis
Subjects had an average bladder volume increase of 342.4 ± 284% between initial and final time points (137.26 ± 113.12 cc to 417.2 ± 262.1 cc) and the corresponding change in rectal volume was -6.9 ± 37.7% (102.8 ± 77.2 cc to 102.3 ± 57.9 cc). Figure 2 shows the 3D rendered volumes of the prostate and OARs for sub-  analysis revealed a 20.2% reduction in the D95% dose to the PTV and 22% increase of the D15% dose to the bladder (D15%(TP1) = 80.13 Gy) which is deemed not clinically acceptable. 28 However, the mean dose to the rectum decreased by 11.98% (D15% (TP1) = 2.3 Gy). Table 2 lists the minimum dose and D95% to the PTV as well as the D15%, D25%, and D35% to the bladder and rectum, as mea-  Although previous studies seeking to find an optimal bladder and rectal state 30,31 for prostate radiotherapy have not found significant differences in the intra-fractional prostate displacement between plans that were designed for patients with full and empty bladders, they have not investigated displacement and change in the dose to the prostate between extreme bladder states. Our results showed that changes in the bladder volume can lead to large, systematic displacements in the prostate and SVs. The major displacements are observed in the A-P and S-I directions in both organs.

3.B | Dosimetric analysis
While the prostate is shifted anteriorly in most cases as the bladder volume is reduced, in some cases, posterior shift is observed. These findings match the results of a recent study based on analysis of CBCT and four-dimensional (4D) trans-perineal ultrasound (4D TPUS) measurements of 60 patients that showed intra-fractional motion of the prostate in the A-P and S-I directions. 32 In this study, reduction of A-P motion of the prostate was observed when the planned bladder volume was greater than 200 ml. Also, when the daily bladder volume was within the third quartiles of the planned CT volumes, the A-P and S-I intra-fraction displacement of the prostate was reduced.
The major contributor to vector displacements of the prostate and SVs is the change in bladder volume; however, rectal status/volume can also contribute to the range of displacements of these organs, which can be dominant along different axes. This can be observed in subject 7 (Fig. 2)   T A B L E 1 Centroid displacement of the prostate and seminal vesicles between initial and final time points for the cohort. Δx, Δy, and Δz represent displacement of the organ centroids in the LR, AP, and SI directions, respectively. The last two rows reflect the number of subjects that had an organ center of mass displacement of >2 mm along each axis or as the total vector displacement.

Prostate
SVs A previously published CT-based treatment planning study evaluated the dosimetric impact of full and empty bladders. 18 The present study builds upon this previous work by incorporating MRI across 3-5 time points per subject. MRI has been shown to enable more accurate and more consistent delineations. 34 Further, Moiseenko et al. found that bladder filling status had limited dosimetric impact on the prostate and rectal doses; however at that time, treatment planning was conducted using a four-field box technique. 18 Our work implements much more conformal treatment planning using arc therapy which showed that when the bladder volume changes from full to empty, PTV coverage was adversely affected. Finally, while CT is the gold standard for treatment planning, the present work is the first to evaluate the performance and dosimetric impact of synthetic CT across varied subject anatomies.
Previous studies have shown that using different table-top configurations used in MRI (i.e., flat and curved couches) may lead to changes in the relative location of pelvic organs. 35 When performing MRI scans with the patient in treatment position (i.e., using a flat tabletop similar to the treatment couch vs. a curved diagnostic couch), more accurate rigid registrations between MR images and CT images for prostate RT planning has been observed. 36 It has also been shown that during MR-SIM, the weight of the flexible anterior body coils may contribute to changes in the position of pelvic organs. 34 In this study, serial imaging data were acquired using a single setup for each subject (i.e., subjects did not leave the MR  17 Using CBCT images of 19 subjects, they found that a 10% increase in bladder volume leads to 5.6% reduction of mean dose to the prostate. They did not find significant variations of the rectal volume. These findings contradict a previous report that although confirming displacement of the prostate in the A-P direction after voiding the bladder, found no correlation between prostate shifts with bladder and rectal volume. 18 These results may be due to that study only evaluating two bladder states (full and empty), and imaging/contouring was done using CT images. Also, the reference for prostate motion was based on external fiducial markers.
Considering that contouring the organs might introduce some uncertainty in their size and position, we eliminated the possible inter-observer error by having only one physician delineate the organs in all subjects, based on the protocol guidelines of RTOG 0815. 25 This minimized the differences between the organ contours for each subject at different time points to ensure isolation of volume differences and their effects on the organs. T A B L E 2 Dose volume histogram (DVH) metrics for the planning target volume (PTV), bladder and rectum for the nine subjects. For each subject, the top row represents the dose at the bladder empty state and the bottom row the dose at the bladder full state. Mean and standard deviation of the dose in the initial and final time points across all subjects has been calculated and the bottom row represents the significance of the difference between these doses in these two time points. Bold-italized values indicate doses that are outside the accepted range.