The Mobius AIRO mobile CT for image‐guided proton therapy: Characterization & commissioning

Abstract Purpose The purpose of this study was to characterize the Mobius AIRO Mobile CT System for localization and image‐guided proton therapy. This is the first known application of the AIRO for proton therapy. Methods Five CT images of a Catphan®504 phantom were acquired on the AIRO Mobile CT System, Varian EDGE radiosurgery system cone beam CT (CBCT), Philips Brilliance Big Bore 16 slice CT simulator, and Siemens SOMATOM Definition AS 20 slice CT simulator. DoseLAB software v.6.6 was utilized for image quality analysis. Modulation transfer function, scaling discrepancy, geometric distortion, spatial resolution, overall uniformity, minimum uniformity, contrast, high CNR, and maximum HU deviation were acquired. Low CNR was acquired manually using the CTP515 module. Localization accuracy and CT Dose Index were measured and compared to reported values on each imaging device. For treatment delivery systems (Edge and Mevion), the localization accuracy of the 3D imaging systems were compared to 2D imaging systems on each system. Results The AIRO spatial resolution was 0.21 lp mm−1 compared with 0.40 lp mm−1 for the Philips CT Simulator, 0.37 lp mm−1 for the Edge CBCT, and 0.35 lp mm−1 for the Siemens CT Simulator. AIRO/Siemens and AIRO/Philips differences exceeded 100% for scaling discrepancy (191.2% and 145.8%). The AIRO exhibited higher dose (>27 mGy) than the Philips CT Simulator. Localization accuracy (based on the MIMI phantom) was 0.6° and 0.5 mm. Localization accuracy (based on Stereophan) demonstrated maximum AIRO‐kV/kV shift differences of 0.1 mm in the x‐direction, 0.1 mm in the y‐direction, and 0.2 mm in the z‐direction. Conclusions The localization accuracy of AIRO was determined to be within 0.6° and 0.5 mm despite its slightly lower image quality overall compared to other CT imaging systems at our institution. Based on our study, the Mobile AIRO CT system can be utilized accurately and reliably for image‐guided proton therapy.


| INTRODUCTION
Advances in in-room computed tomography (CT) scanners and conebeam technology have led to the proliferation of CT localization for image guided radiation therapy (IGRT). [1][2][3][4][5][6][7] These imagers provide improved treatment accuracy over conventional orthogonal imaging allowing for increased precision in radiation delivery. Patient localization accuracy is particularly important in proton beam radiation therapy due to the sharp dose fall-off compared to conventional x-ray therapy. Unfortunately, due to the size and geometry of proton therapy units, imaging has largely been limited to orthogonal kV/kV x-ray systems. 8 Recently, people have reported on the use of inroom CT scanners for proton beam radiation therapy localization. [9][10][11] Some proton vendors are developing technologies for CBCT (Proteus â ONE, IBA, Belgium, HITACHI, Tokyo, Japan). To date, standalone in-room CT scanners utilized have had large footprints and are either not amenable to or cumbersome to use in the more compact proton therapy centers, such as the S250 (Mevion Medical Systems, Littleton, MA, USA). 12 For these reasons, a small-footprint mobile CT scanner, commonly used for image-guided surgery, could be of great value for 3D image-guided proton therapy (IGPT).
The AIRO Mobile CT System (Mobius Imaging LLC, Shirley, MA, USA) is a large bore (107 cm) helical 32 slice CT scanner historically utilized for intra-operative imaging for spinal surgeries. Our institution is the first to acquire and clinically implement the AIRO Mobile CT System (AIRO) for IGPT. The AIRO's small footprint (W 9 L 9 H: 1.94 9 1.54 9 1.90 m) occupies 1.28 m 2 of treatment floor space (in scan mode). When not in use, the AIRO is stored in the maze (Fig. 1) to avoid radiation-induced damage to its sensitive electronics. The AIRO's motor-controlled castors provide effortless transport from the storage location (in the maze hallway) to the scanning location.
The AIRO's image quality characteristics have been reported by Weir et al. for surgical applications. 13 Our study assessed the performance characteristics of the AIRO at our institution compared with IGRT systems used for intensity-modulated radiation therapy (IMRT) and CT simulators at our institution for radiotherapy applications. Our current IGPT workflow involves CT simulation using a Philips Big Bore 16 Slice CT Simulator with routine orthogonal kV/kV image pairs (flatpanel detectors) preceding each treatment fraction. The AIRO will be utilized for target localization and inter-fraction adaptive treatment assessment. To our knowledge, our facility is the first compact proton therapy system to use a mobile CT scanner for IGPT.

| MATERIALS AND METHODS
Mevion Medical Systems has developed software, Verity TM 3D, that uses CT images from any scanner to perform 3D image registrations for proton therapy. Due to the ease of motion and small footprint, we have selected the AIRO CT scanner for IGPT.
The AIRO is sold commercially for surgical purposes by BrainLab (Munich, Germany). Mevion has developed an infrared camera tracking system to interface with the CT scanner to facilitate use of CT images for IGPT. This system references the 3D image set to the in-room coordinates and planning CT via a reference frame, infrared camera, and CT scan. This reference frame attaches to the treatment couch and includes infrared markers to determine initial room coordinates as well as ceramic markers to reference the CT image to room coordinates. An additional infrared marker is rigidly attached to the gantry so that changes in the reference frame location in relation to the treatment room can be detected as the robotic couch moves between the initial treatment setup location and the CT imaging location. The AIRO unit was specifically adapted to use inside the proton vault by removing the integrated stand that supports the weight of surgical gurneys in order to avoid interference with the robotic couch ( Fig. 1). Image acquisition was performed with various protocols. On the AIRO, images were acquired with soft, standard, and sharp reconstruction kernels with the pre-set clinical head protocol. The AIRO's pre-set clinical protocols for head, thorax, abdomen, pelvis, shoulder are listed in Table 1. The sharp, standard, and soft kernels are noise filtering Gaussian smoothing kernels with r = 0.00, 0.68, and 0.92, respectively.

2.A | Image quality characterization and comparison to simulators and IGRT systems
On the EDGE, images were acquired using the following pre-set clinical techniques: head, thorax, pelvis, pelvis obese, and image gently ( Table 1). On the Philips CT simulator, standard clinical protocols were used (120 kVp, 450 mA) with 1.5 mm, 2.0 mm, and 3.0 mm slice thickness. On the Siemens CT simulator images were acquired with standard brain (120 kV, 450 mAs), high res brain (120 kV, 580 mAs), abdomen (120 kV, 250 mAs), and thorax  (Table 2). 14 The mean and standard deviation of five or more scans are reported. The low contrast module (CTP515) was assessed on the AIRO, Philips, and Siemens scans. Low CNR was calculated using the 15.0 mm 1% Supra-slice target. A region-of-interest (ROI) was dropped onto the Supra-slice target and another on a uniform region of the phantom.

2.B | Localization accuracy
A Stereophan phantom (Sun Nuclear Corporation, Melbourne, FL, USA) was used to measure the localization accuracy of the Mevion-AIRO system. The universal spacer insert and CT/MRI insert were inserted into the cylinder cavity. The phantom was leveled with the precision leveling stand. Treatment isocenter was placed at the BB target corresponding to the laser alignment marks on the outside of the phantom. The robotic couch was re-positioned to acquire the AIRO images following the Mevion localization workflow. Once the images were acquired, the robotic couch was moved back to the initial treatment isocenter based on the robotic couch's coordinates.
The software then performed a CT registration correlating the reference frame's ceramic fiducials to its infrared markers (relative to treatment isocenter). The images were then manually registered to the 3D planning CT. Successively, a kV/kV image pair was acquired and registered to the digitally reconstructed radiograph (DRR). Localization accuracy was defined as the difference between suggested shifts from the kV/kV pair versus the AIRO CT image set, where the kV/kV image was considered the gold standard. In addition, end-toend tests were performed to illustrate both time and workflow for patient setup and target localization.  The AIRO image quality results for various kernels are listed in sharp/soft and standard/soft kernels. Figure 2 illustrates that the T A B L E 2 Image quality tests and calculation methods.

Modulation Transfer Function (MTF) a
The modulation of multiple ROIs in the phantom with various line bar patterns. As spatial frequency increases, ROI's modulation decreases. The modulation is normalized to its highest value and plotted.

HU90ÀHU10 HU90þHU10
Where HU 90 s the 90 th percentile CT number in the ROI and HU 10 M % ð Þ s the 10 th percentile CT number in the ROI.
Scaling Discrepancy a The maximum possible error in a measurement of distance.
Mð%Þ À 100% j j Â L Where M ð%Þ is the magnification of the image and L is the length of the image's longest side.
Geometric Distortion a Verifies that correct distance measurement occurs in all regions of the CT image.
Overall Uniformity a 1 À Max90 ÀMin10 Max90 þMin10 Â 100% Where Max 90 is the maximum 90 th percentile CT number in all uniformity regions and Min 10 is the minimum 10 th percentile CT number in all uniformity regions Minimum Uniformity a The lowest uniformity of all ROIs. Uniformity is calculated as: Where HU 90 is the 90 th percentile of the pixel CT number in the ROI and HU 10 is the 10 th percentile of the CT number in the ROI.
Maximum HU Deviation a The maximum deviation between mean HU value of an ROI subtracted from its defined reference HU absolute value of all calculated HU deviation on an image.

Contrast a
HU2 ÀHU1 HU2 þHU1 Â 100% Where HU 2 is the mean CT number in the region with greater signal and HU 1 is the mean CT number in the region with lesser signal.
High Contrast-to-Noise Ratio a HighContrast Noise Where r 1 is the standard deviation of the high contrast ROI, r 2 is the standard deviation of background, HU 1 is the mean CT number of the high contrast ROI and HU 2 is the mean CT number of background.
Low Contrast-to-Noise Ratio 13 LowContrast Noise ¼ HUROIÀHUb Where r b is the standard deviation of the background ROI, HU ROI is the mean CT number inside the low contrast ROI and HU b is the mean CT number of the background ROI.   Conversely, the Philips CT simulator exhibited superior MTF and spatial resolution compared to all modalities (Table 4 and Fig. 3  These results included robotic couch motion from the CT scanner imaging position to the standard kV imaging isocenter.

3.B | Localization accuracy of Mevion-AIRO system
T A B L E 3 AIRO image quality results for varying reconstruction kernels.

Sharp Standard Soft
Scaling discrepancy (mm): The localization accuracy tests also included a workflow and time analysis to determine the optimal location of the CT scanner in the treatment room and to gauge the time required for a CT scan. The resultant workflow required one therapist to bring the CT scanner into the room while an additional therapist performed the initial patient setup. An additional 5-10 min were needed to acquire a CT scan.

3.C | CTDI vol
Measured CTDI vol for the AIRO, and Philips CT Simulator are illustrated in are required to bring the scanner into the treatment room. Future work will involve a time study for patient localization and evaluation scanning as well as characterization of the CT images as they relate to stopping power and CT number constancy for use in adaptive proton therapy.

CONCLUSI ONS
This work evaluated various image quality, localization accuracy, and dosimetric metrics of the AIRO and compared them to other common modalities. Based on our findings, we recommend that the AIRO Mobile CT System can be applied clinically for safe and accurate image-guided proton therapy.

CONF LICT OF I NTEREST
There are no conflicts of interest.