Technical assessment of a mobile CT scanner for image‐guided brachytherapy

Abstract Purpose The imaging performance and dose of a mobile CT scanner (Brainlab Airo®, Munich, Germany) is evaluated, with particular consideration to assessment of technique protocols for image‐guided brachytherapy. Method Dose measurements were performed using a 100‐mm‐length pencil chamber at the center and periphery of 16‐ and 32‐cm‐diameter CTDI phantoms. Hounsfield unit (HU) accuracy and linearity were assessed using materials of specified electron density (Gammex RMI, Madison, WI), and image uniformity, noise, and noise‐power spectrum (NPS) were evaluated in a 20‐cm‐diameter water phantom as well as an American College of Radiology (ACR) CT accreditation phantom (Model 464, Sun Nuclear, Melbourne, FL). Spatial resolution (modulation transfer function, MTF) was assessed with an edge‐spread phantom and visually assessed with respect to line‐pair patterns in the ACR phantom and in structures of interest in anthropomorphic phantoms. Images were also obtained on a diagnostic CT scanner (Big Bore CT simulator, Philips, Amsterdam, Netherlands) for qualitative and quantitative comparison. The manufacturer’s metal artifact reduction (MAR) algorithm was assessed in an anthropomorphic body phantom containing surgical instrumentation. Performance in application to brachytherapy was assessed with a set of anthropomorphic brachytherapy phantoms — for example, a vaginal cylinder and interstitial ring and tandem. Result Nominal dose for helical and axial modes, respectively, was 56.4 and 78.9 mGy for the head protocol and 17.8 and 24.9 mGy for the body protocol. A high degree of HU accuracy and linearity was observed for both axial and helical scan modes. Image nonuniformity (e.g., cupping artifact) in the transverse (x,y) plane was less than 5 HU, but stitching artifacts (~5 HU) in the longitudinal (z) direction were observed in axial scan mode. Helical and axial modes demonstrated comparable spatial resolution of ~5 lp/cm, with the MTF reduced to 10% at ~0.38 mm−1. Contrast‐to‐noise ratio was suitable to soft‐tissue visualization (e.g., fat and muscle), but windmill artifacts were observed in helical mode in relation to high‐frequency bone and metal. The MAR algorithm provided modest improvement to image quality. Overall, image quality appeared suitable to relevant clinical tasks in intracavitary and interstitial (e.g., gynecological) brachytherapy, including visualization of soft‐tissue structures in proximity to the applicators. Conclusion The technical assessment highlighted key characteristics of dose and imaging performance pertinent to incorporation of the mobile CT scanner in clinical procedures, helping to inform clinical deployment and technique protocol selection in brachytherapy. For this and other possible applications, the work helps to identify protocols that could reduce radiation dose and/or improve image quality. The work also identified areas for future improvement, including reduction of stitching, windmill, and metal artifacts.

and interstitial (e.g., gynecological) brachytherapy, including visualization of soft-tissue structures in proximity to the applicators.

Conclusion:
The technical assessment highlighted key characteristics of dose and imaging performance pertinent to incorporation of the mobile CT scanner in clinical procedures, helping to inform clinical deployment and technique protocol selection in brachytherapy. For this and other possible applications, the work helps to identify protocols that could reduce radiation dose and/or improve image quality. The work also identified areas for future improvement, including reduction of stitching, windmill, and metal artifacts. Their work showed spatial resolution up to 4 lp/cm in the head fieldof-view (FOV) for the mobile scanner, compared to 7 lp/cm resolution from a Siemens Sensation 64-slice MDCT scanner for the same FOV. 1 The Airo was also found to exhibit higher radiation dose than the Sensation 64 for comparable technique factors (50% increase for head, 85% increase for body phantom), and ring-like artifacts were noted as contributors to increased low-frequency noise in the image NPS. 1 The system was further evaluated for application in spine surgery by Hecht et al., 2 who assessed the accuracy and workflow for navigated spinal instrumentation, reporting a screw placement accuracy rate of 95.9%. 3 Similarly, Scarone et al. 3  For application in image-guided proton therapy, the study reported by Oliver et al. 4 included comparison of mobile CT performance to a CBCT system and two MDCT systems. The limiting spatial resolution was reported at 2.1 lp/cm for the Airo, compared to 4.0 lp/cm for the Brilliance Big Bore CT simulator (Philips, Amsterdam, Netherlands) and 3.7 lp/cm for the EDGE CBCT system (Varian, Palo Alto, CA). Compared to the Philips CT simulator, the Airo was found to exhibit a 60% higher dose for head protocols and 8% higher dose for abdomen protocols. Despite the lower spatial resolution, localization accuracy was within 0.6°and 0.5 mm, which was concluded to be sufficient for therapy guidance. 4 The work reported below is distinct from previous publications in several important respects. A technical assessment guiding the selection of technique protocols for the Airo has yet to be described.
Specifically, the results shown below provide a thorough evaluation of imaging performance for both helical and axial modes, and the effect of various technical performance characteristics on image quality in a range of pertinent anatomical sites is evaluated. Accordingly, the work identifies distinct sources of helical and axial mode image artifacts that may be significant for some imaging tasksviz., helical mode windmill sampling artifacts (especially pronounced about high-contrast, high-frequency structures, such as bones or metal instrumentation) and axial mode stitching artifacts that present nonuniformity in the longitudinal direction. We also investigate the effect of centering errors (i.e., patient misaligned from isocenter) on image uniformity, owing to the effect of such errors on bowtie filter calibration. The performance of the manufacturer's metal artifact reduction (MAR) algorithm is investigated, and emphasis throughout is primarily on soft-tissue visualization tasks pertinent to soft-tissue interventions (e.g., liver lesions, prostate, and cervix). Finally, we focus the studies upon the growing scope of clinical application of this mobile CT scanner in brachytherapya context within which its performance characteristics have yet to be evaluated and interpreted with respect to relevant imaging tasks.
Brachytherapy is an important modality for both definitive and adjuvant treatment of cervical, endometrial, and prostate cancers, [5][6][7] and significant developments over the last two decades have increased the use of 3D image guidance in brachytherapy. 8 Current evidence suggests that 3D image-guided brachytherapy improves local control compared with conventional brachytherapy; 9 however, the integration of imaging and treatment delivery in brachytherapy is often limited by logistical and space constraints. Patient transfer between the imaging and treatment areas is often the only viable solution and increases the potential for motion of the applicator with respect to target and normal tissues and creates uncertainty in treatment delivery. 10 Uncertainty can be partially mitigated by using intraoperative ultrasound to assist with placement of the brachytherapy applicator or by implementing fixation mechanisms to restrict external patient movement. Limitations of the ultrasound approach include image quality / interpretation, image artifacts arising from the brachytherapy applicator, and challenges in modifying the treatment plan. 11 Limitations of the fixation mechanisms include the inability to prevent internal organ deformation (e.g., bowel, bladder) between imaging and dose delivery.
The incorporation of a CT system in the treatment room would enable the integration of applicator insertion, imaging, and treatment delivery into a single location without the need to transfer the patient. Additionally, the improved workflow efficiency that an integrated treatment room provides could improve patient safety for procedures involving anesthesia. 12 Previous reports have described integrated image-guided brachytherapy suites. For example, the Advanced Multimodality Image Guided Operating (AMIGO) suite integrates CT, positron emission tomography (PET), magnetic resonance imaging (MRI), and ultrasound. 12 While the AMIGO suite uses an impressive array of imaging modalities at the time of implantation, patients must then be transferred to radiation oncology where there is adequate shielding for brachytherapy treatment. An integrated CBCT brachytherapy suite is described at the MAASTRO clinic in the Netherlands; 13 however, the quality of CBCT is known to be inferior to MDCT, challenged in particular with respect to soft-tissue visualization. In this work, we describe the use of a mobile CT unit in a shielded brachytherapy suite, informed by a technical assessment of the CT scanner.
The technical assessment reported below examines the imaging performance and radiation dose for the Airo mobile CT scanner, including a variety of manufacturer-specified protocols available at the time of writing and differences between helical and axial scan modes. Imaging performance was quantitatively evaluated in terms of CT number accuracy, uniformity, spatial resolution, noise, NPS, and contrast-to-noise ratio (CNR) in simulated soft-tissue structures. extremities, and lower extremities. For each of these, x-ray tube potential is fixed at 120 kV, and although 80 kV and 100 kV protocols are accessible via the nonclinical/engineering interface, the current deployment only supported air calibration for 120 kV. Rather than common protocol variations for "small/ large" body habitus or "adult/ pediatric" subjects, the x-ray tube current automatically scales according to the weight (kg) of the patient (entered manually via the control interface).
The scanner features a detector with N row = 32 rows, with each detector element of size d x = 0.5 mm in the lateral direction and d z = Image reconstruction is based on filtered backprojection (FBP) with adjustable filters referred to as "soft," "standard," and "sharp." At the time of writing, the system is implemented such that the filter must be specified prior to performing the scan (and cannot be For consistent terminology below, the term "axial" is used in reference to axial scan mode (cf., helical scan mode), and the term "transverse" is used in reference to an (x, y) slice of the image reconstruction (cf., sagittal (y, z) or coronal (x, z) slices). (CTDI w ) as:

2.B | Dose measurements
The volume CTDI (CTDI vol ) was given by: where pitch is 1.4 for helical mode.

2.C | Imaging performance
Imaging performance was assessed in "Service" mode, allowing full control over the tube current (5-250 mA) and beam energy (80, 100, or 120 kV). The phantoms detailed below were used to measure image uniformity, noise, spatial resolution, etc., and each was marked with small plastic beads or tape to facilitate repositioningfor example, to repeat scans with different reconstruction filters.

2.C.1 | CT number accuracy and linearity
To evaluate the HU accuracy of the system, a 33-cm diameter cylin-

2.C.2 | Uniformity
Image uniformity was assessed in a 20-cm-diameter cylindrical water phantom scanned with a nominal technique of 120 kV, 211 mAs axial mode (149 mAs eff helical mode) using soft, standard, and sharp reconstruction filters. The phantom was scanned centered at isocenter (with 25.6 cm reconstruction FOV; a x = a y = 0.5 mm; a z = 1.0 mm) and offset laterally by 5 cm (with 30 cm reconstruction FOV; a x = a y = 0.56 mm; a z = 1.0 mm). Scans were also acquired at 80 kV and 100 kV, recognizing that the system only allowed air calibration at 120 kV at the time of writing. Nonuniformity (t cup ) was evaluated as the difference in average CT number measured at the center and periphery of the phantom. In the offset scan acquisition, nonuniformity was evaluated as the difference in average CT number measured at the anterior (near isocenter) and posterior periphery of the phantom.

2.C.3 | Spatial resolution
Spatial resolution was assessed using an edge-spread phantom (2.8cm-diameter acrylic rod in air) scanned in both helical and axial modes at 120 kV, 211 mAs axial (149 mAs eff helical) with a 25.6-cm FOV (a x = a y = 0.5 mm; a z = 1.0 mm) and reconstructed with soft, standard, and sharp filters. The oversampled edge-spread function, ESF, was computed from images of the edge, and the numerical derivative of the ESF was computed (yielding the oversampled linespread function, LSF) and normalized to unity area, from which the MTF was computed by Fourier transform. 15,16 The spatial resolution of the system was also assessed qualita- where N x;y;z are the size of each ROI (65 × 65 × 65 voxels), < > denotes the ensemble average of 48 ROIs, and F is the 3D discrete Fourier transform. The ROIs were taken at fixed distance 4.0 cm from center and detrended by a first-order hyperplane to yield zero- Scans were also acquired on the Philips Big Bore CT simulator (same parameters as Sec. II.C.iii, reconstructed with a standard filter) of the same 20-cm-diameter water phantom, and NPS was evaluated as described above. For comparison, a helical scan was acquired using the Airo system at 120 kV, 67.8 mAs eff , and reconstructed with 25.6-cm FOV and a standard filter. The dose values (CTDI vol ) for the Philips and Airo scans were 9.9 mGy and 10.1 mGy, respectively.

2.C.5 | Low contrast resolution
The low contrast resolution was assessed using a custom 16-cm-diameter cylindrical polyethylene phantom containing 2.8-cm-diameter electron density inserts (Model 467, Gammex RMI, Madison, WI) to simulate soft tissues. Scans were performed at 120 kV in axial mode with mAs varying from 10 mAs to 480 mAs with 25.6-cm FOV. The contrast-to-noise ratio (CNR) was analyzed in the simulated adipose insert (AP6, ρ w e = 0.93) relative to polyethylene background. The CNR was calculated as 18 : The low contrast resolution was also assessed qualitatively using

2.D | Brachytherapy applications
To assess performance in application to brachytherapy, two custom phantoms were constructed incorporating commonly used gynecological brachytherapy devices (Fig. 1). The first was a stainless steel (channel) and plastic (segments) vaginal cylinder (Nucletron, Veenendaal, The Netherlands) used to deliver intracavitary high dose-rate (HDR) brachytherapy along the wall of the vaginal canal 19 as shown in [ Fig. 1(b)]. Relevant imaging tasks in placement of the vaginal cylinder include the ability to assess the applicator contact to the surrounding soft tissues and to identify the presence of air pockets. 19 The second was an MRI/CT-compatible interstitial ring and CHERNAVSKY ET AL.  Table 1 summarizes technique factors and dosimetry for various protocols. The axial mode CTDI vol for the head (16 cm) and body (32 cm) protocols was 78.9 mGy and 24.9 mGy, respectively, consistent (within~5-10%) with measurements reported by Oliver, et al. 2 The CTDI vol reported on the console based on manufacturer specifications was systematically~20% lower than the measured values -64.7 mGy and 20.9 mGy, respectively. The CTDI vol for helical mode is scaled down from the dose in axial mode by the pitch -56.4 mGy and 17.8 mGy for the head and body, respectively.

3.B | Accuracy and uniformity
The CT number linearity followed a bilinear trend with electron density as shown in [ Fig. 2(a)], 21

Mode
Diameter (   toms discussed below (and in Fig. 9). These effects may result from imperfect detector air calibration at the longitudinal edges of the detector adjacent to the collimator [evident as stitching artifacts in Fig. 3(d)].

3.D | Noise
As shown in Fig. 6, the NPS for helical mode is greater than that in axial mode for similar mAs/mAs eff , exhibiting a slightly increased low-to mid-frequency noise characteristic [ Fig. 6(a)

3.E | Low contrast resolution
As shown in [ Fig. 8(a)], the CNR followed the expected square-root relationship with dose, consistent with Fig. 5, and contrast (i.e., mean difference in HU) was independent of dose. The CNR also improved as expected with smoother reconstruction filters. Figure 8b illustrates the detectability of a low-contrast insert (6 HU contrast to background) for the~25 mm diameter cylinder in helical mode, compared to the~6 mm diameter cylinder in axial mode, consistent with the increased noise levels observed in helical mode (as in Figs. 5, 6).

3.F | Image quality and artifacts in anthropomorphic phantoms
The artifacts and image characteristics described in Figs pitch. 23 The ability to visualize the (spherical) liver nodules in the transverse view was comparable between helical and axial scan modes. Figure 10 illustrates the performance of the MAR algorithm available on the system at the time of deployment, illustrating the reduction in streak artifact for both helical and axial modes. For helical mode, streaks arise from the high-frequency structure associated with the metal screw [ Fig. 10(A-a,b)] due to a combination of beamhardening, photon starvation, and helical z-interpolation/sampling effects. [22][23][24] The MAR algorithm treats the former (beam-hardening and photon starvation) but not the latter, so the helical mode image  Fig. 10(B-a,b)]. The ability to delineate the boundaries of the screw (e.g., identify pedicle breach) was still challenging even with MAR, especially in helical mode where windmill artifacts arising from the screw presented strong nonuniformity.

3.G | Brachytherapy Applications
As illustrated in Figs   The advantages of a mobile CT scanner (e.g., compared to realtime ultrasound or a diagnostic MDCT located in a separate imaging suite) for brachytherapy guidance were also evident from this work.
Chief among these is the ability to image and treat the patient without repositioning, the small footprint of the scanner, and the feasibility of moving the scanner out of the room. Likely improvements to workflow are also anticipated (but not measured directly in the current work) by reducing time requirements in patient transfer to the brachytherapy suite. Moreover, the scanner offers the clinician the ability to verify applicator position immediately prior to treatment delivery without disturbing the treatment position.
A limitation of the current work is that the phantom measurements did not probe image quality factors related to patient motion.
The studies therefore did not assess potential benefits to image quality associated with faster scan speed in helical mode. Future work includes evaluation of clinical workflow with the system and potential improvements to brachytherapy treatment outcomes. One may also anticipate future improvements to the system from the manufacturer, including lower-dose protocols, an increased range of technique factors (e.g., selection of kV and pitch), the ability to reconstruct a dataset with any reconstruction filter, improved longitudinal image uniformity (reduced stitching artifacts), and improved MAR methods.

ACKNOWLEDG MENTS
The authors thank colleagues in the I-STAR Laboratory and the Carnegie Center for Surgical Innovation at Johns Hopkins University, as

CONF LICT OF I NTEREST
No conflict of interest.