Dose reduction potential in cone-beam CT imaging of upper extremity joints with a twin robotic x-ray system

Cone-beam computed tomography is a powerful tool for 3D imaging of the appendicular skeleton, facilitating detailed visualization of bone microarchitecture. This study evaluated various combinations of acquisition and reconstruction parameters for the cone-beam CT mode of a twin robotic x-ray system in cadaveric wrist and elbow scans, aiming to define the best possible trade-off between image quality and radiation dose. Images were acquired with different combinations of tube voltage and tube current–time product, resulting in five scan protocols with varying volume CT dose indices: full-dose (FD; 17.4 mGy), low-dose (LD; 4.5 mGy), ultra-low-dose (ULD; 1.15 mGy), modulated low-dose (mLD; 0.6 mGy) and modulated ultra-low-dose (mULD; 0.29 mGy). Each set of projection data was reconstructed with three convolution kernels (very sharp [Ur77], sharp [Br69], intermediate [Br62]). Five radiologists subjectively assessed the image quality of cortical bone, cancellous bone and soft tissue using seven-point scales. Irrespective of the reconstruction kernel, overall image quality of every FD, LD and ULD scan was deemed suitable for diagnostic use in contrast to mLD (very sharp/sharp/intermediate: 60/55/70%) and mULD (0/3/5%). Superior depiction of cortical and cancellous bone was achieved in FDUr77 and LDUr77 examinations (p < 0.001) with LDUr77 scans also providing favorable bone visualization compared to FDBr69 and FDBr62 (p < 0.001). Fleiss’ kappa was 0.618 (0.594–0.641; p < 0.001), indicating substantial interrater reliability. In this study, we demonstrate that considerable dose reduction can be realized while maintaining diagnostic image quality in upper extremity joint scans with the cone-beam CT mode of a twin robotic x-ray system. Application of sharper convolution kernels for image reconstruction facilitates superior display of bone microarchitecture.

www.nature.com/scientificreports/ elbow region are more common in younger populations 4 . Plain radiographs usually represent the primary imaging method if traumatic injuries of the upper limb are suspected 5 . Radiograms offer a compromise between fast scan time, ubiquitous availability, and low radiation dose on the one side, and limited 2D image information on the other side 6 . Considering that trauma mechanisms can be vastly heterogeneous, ranging from simple falls on the outstretched hand from standing height to high-velocity injuries in sports activities and motor vehicle accidents 2,7 , 2D imaging may be insufficient for trauma evaluation in some patients. Especially more severe injuries, e.g., multi-fragmentary, intraarticular and displaced fractures of the wrist and elbow require additional 3D assessment for presurgical planning 8 and image fusion with intraoperative fluoroscopy 9,10 . In addition, computed tomography scans allow for reliable exclusion or detection of radiographically occult fractures in case of divergent clinical presentation 11 . Even for certain chronic conditions or postoperative follow-up imaging, CT can provide valuable diagnostic information with the typical drawback of increased radiation dose 12,13 . Despite the predominance of multidetector CT in most radiology departments, the value of another technical approach to computed tomography is increasingly recognized for skeletal imaging tasks: Combining a flat-panel detector build with pyramid-shaped beam geometry, cone-beam computed tomography (CBCT) has become a powerful tool for 3D imaging in trauma patients. Detailed visualization of bone microarchitecture 14 , superior positioning options and potential for reduced radiation dose are among the major advantages compared to conventional multidetector CT 15 . Having been considered an integral part of maxillofacial imaging for a long time [16][17][18][19] , access to dedicated scanners for the appendicular skeleton has led to increasing relevance of CBCT in musculoskeletal imaging during the last decade 20 .
The twin robotic x-ray system in this study represents a novel approach to the CBCT formula that uses two telescopic arms for 2D and 3D image acquisition instead of a conventional gantry-based approach. Whilst the benefits of the CBCT option for fracture detection have previously been analyzed, data regarding the actual dose reduction potential of this multi-use system is lacking. Therefore, in this work, various combinations of acquisition and reconstruction parameters were evaluated for the CBCT mode of the twin robotic x-ray system in cadaveric wrist and elbow scans, aiming to define the best possible trade-off between image quality and radiation dose.

Material and methods
Cadaveric phantoms. For this experimental study, we received permission from the institutional review board of the University of Würzburg, Germany (IRB number 20200506 01). Two formalin-fixed cadaveric specimens were obtained from the anatomical institute of the local university. Body donors had voluntarily donated their bodies to the anatomical institute for study and scientific purposes. Thus, the ethics committee of the University of Würzburg waived additional written informed consent. Bilateral wrist and elbow examinations were performed for both cadaveric specimens in supine position using the twin robotic x-ray system's tableside scan trajectory. Scanning process for upper extremity imaging is displayed in Fig. 1 by a staff member (informed consent for publication was obtained). All procedures were conducted in accordance with the relevant guidelines and regulations.

Cone-beam CT scans.
For image acquisition, the multifunctional radiography system (Multitom Rax, Siemens Healthcare GmbH, Erlangen, Germany) uses two telescopic arms connected to ceiling rails within the x-ray suite. One of the arms holds a quadratic flat-panel detector (edge length 426 mm), while the other carries the x-ray tube. Different motion trajectories for both arms allow for 2D and 3D imaging in specified positions. CBCT scans are performed with an asymmetric source-to-image distance of 1150 mm, an input field of 213 × 213 mm 2 and a 2D matrix of 1440 × 1440 un-binned pixels in high-resolution mode, which results in an effective pixel size of 149 µm. The current software version VF 11 (Siemens Healthcare GmbH) supports acquisition of 26 projection images per second with a sweep angle of 200 degrees, hence resulting in a total scan time of 14 s.
Before each CBCT scan, system settings can be manually adjusted to achieve target dose levels. Using a clinically established full-dose protocol (FD) as reference, four other scan protocols with lower radiation doses were  Table 1. Dose-area products for all examinations were obtained from the automatically created scan report. To calculate volume CT dose indices (CTDI vol ), multiplication by a linear scaling factor that we computed in advance for every combination of tube current-time product and tube voltage was needed. Therefore, dose-length product measurements were performed in the five chambers of a conventional 16 cm PMMA dosimetry phantom (IEC 60601-2-44:2009 compliant). After standard weighting schemes were applied to obtain volume dose-length product values (DLP vol(16 cm) ), these values were divided by the field of view in z-direction (equal to the beam width) to compute the CTDI vol (16 cm) . Finally, dividing the CTDI vol(16 cm) by the DAP resulted in the required scaling factor. Image analysis. After image reconstruction, all datasets were analyzed in randomized and blinded fashion with standard picture archiving and communication software (Merlin, Phoenix-PACS, Freiburg im Breisgau, Germany) by five independent radiologists with 8, 7, 6, 5 and 4 years of experience in musculoskeletal imaging. First, each observer was tasked to evaluate whether the reviewed scan was sufficient for diagnostic use. Second, readers assessed overall image quality and gave dedicated image quality ratings for cortical bone, cancellous bone and soft tissue using a seven-point Likert scale (7 = excellent, 6 = very good, 5 = good, 4 = satisfactory, 3 = fair, 2 = poor, 1 = very poor image quality). Comparing the effect of each convolution kernel on the display of bone microarchitecture, visualization of cortical and cancellous bone was best in very sharp reconstructions of full-dose and low-dose examinations (p < 0.001), with LD Ur77 scans also providing favorable bone depiction compared to FD Br69 and FD Br62 (p < 0.001). Accordingly, sharper ULD Ur77 (p = 0.005) and ULD Br69 (p = 0.009) reconstructions were deemed more suitable for differentiation of trabecula and fatty marrow than softer ULD Br62 studies. The effect of the different convolution kernels on image sharpness is exemplified in Fig. 4. Quality of cortex display in ultra-low-dose scans was independent of image processing (p = 0.764). For soft tissue analysis, no difference between kernels was identified in full-dose (p = 0.302) and low-dose image acquisition (p = 0.129). In ultra-low-dose studies, both ULD Br69 (p = 0.008) and ULD Br62 (p < 0.001) provided superior soft tissue assessability compared to ULD Ur77 with no difference between the two softer kernels (p = 0.739). Detailed image quality evaluation for bone and soft tissue is summarized in Table 3.

Discussion
In this multi-observer study, we evaluated the image quality provided by 15 combinations of acquisition protocols and reconstruction parameters for the CBCT scan mode of a twin robotic x-ray system in cadaveric wrist and elbow imaging. We were able to show that dose reduction of more than 90% is feasible for certain imaging tasks, while maintaining diagnostic image quality with a dedicated ultra-low-dose scan protocol. Reducing the radiation dose even further by implementing detector-based dose modulation resulted in a considerable amount of non-diagnostic studies. Therefore, we postulate that the ultra-low-dose protocol with reduced tube voltage represents the best trade-off between required dose and image assessability for upper extremity scans.
A recent meta-analysis by Nardi et al. identified an average dose of 7.1 µSv for CBCT scans of the appendicular skeleton, stating considerable dose reduction potential in comparison to multidetector CT imaging 22 . Compared to conventional radiography, however, radiation dose of CBCT remains significantly higher in most studies 23 . While scan protocols without a focus on dose reduction may have the potential to overcome the image quality of state-of-the-art multidetector CT scanners, we believe that CBCT can also challenge radiography as the primary method of fracture diagnosis 24 . Due to distance to the radiation-sensitive body trunk, effective dose of extremity imaging is generally lower than for other body regions 25 . Providing diagnostic 3D image information with approximately four times the effective dose of standard 2D radiograms, the ultra-low-dose protocol in this study appears promising for primary use in particular imaging tasks, e.g., carpal bone 14,23 or radial head evaluation 26 . Furthermore, pediatric radiology could benefit from the reduced radiation dose in cases that require 3D assessment to ascertain the injury pattern, e.g., in distal humerus fractures 26 . To investigate which scan protocol is appropriate for different pathologies should be part of future research. From a technical point of view, the maintained image quality in ultra-low-dose scans may foremost be attributed to the small effective pixel size of 149 µm realized by the combination of asymmetric acquisition geometry (the detector rotates in a Table 2. Overall image quality assessment. Synopsis of overall image quality evaluation in cadaveric wrist and elbow scans by five radiologists for each combination of convolution kernel and dose protocol. Median rating values and share of examinations with at least "satisfactory" image quality (score ≥ 4) are reported. Percentage of studies that were deemed "suitable for diagnostic use in patient studies" is displayed. www.nature.com/scientificreports/ smaller radius around the isocenter than the x-ray tube to balance the effect of the focal spot size) and un-binned readout of the flat-panel detector in high-resolution mode.
With very sharp reconstructions of low-dose examinations providing favorable image quality over full-dose scans with softer reformatting, additional dose reduction potential can be realized through image postprocessing. In particular, the superior display of cancellous bone in studies reconstructed with the two sharper convolution kernels may aid fracture assessment due to better discrimination of trabecula and fatty bone marrow. In contrast, the inferior visualization of bone microarchitecture provided by the intermediate convolution kernel might be attributed to the fact that the realized modulation transfer function of this kernel is far below the system's resolution limit without a defined kernel.
Limitations. The two formalin-fixed cadaveric specimens were selected without knowledge on body donor age, time of fixation and bone density of the examined wrist and elbow region. Thus, preexisting osteopenia and

Conclusion
With the cone-beam CT scan mode of a twin robotic x-ray system, dose reduction of more than 90% can be realized for certain upper extremity joint examinations, while maintaining diagnostic image quality. Application of sharper convolution kernels for image reconstruction facilitates superior display of bone microarchitecture, potentially aiding in detection of microfractures in future patient studies.  Table 3. Detailed image quality assessment for bone and soft tissue. Dedicated image quality evaluation of cortical bone, cancellous bone and soft tissue by five radiologists for each combination of convolution kernel and dose protocol in cadaveric wrist and elbow scans. Median rating values and share of examinations with at least "satisfactory" image quality (score ≥ 4) are reported.