First clinical experience with a novel, mobile cone-beam CT system for treatment quality assurance in brachytherapy

Computed tomography (CT) is a well-established imaging modality and integral part of the imaging workflow in several medical specialties [1‐9]. In recent years, particularly cone-beam CT (CBCT) has found increasing applications for interventional [1] and intraoperative [2, 3] purposes. One factor for this is that corresponding mobile devices as C‑arms [1‐4] or O‑arms [5] allow working flexible at the patient, even in imaging position.

Potential for CBCT in brachytherapy is undoubted. Due to the modality’s good high-contrast capability, resolution, and geometric accuracy, CBCT allows in general distinct visualizations of inserted applicators [6]. As examples, Peters et al. [7] performed CBCT-based adaptive planning of seed implantations into the prostate. Al-Halabi et al. [8] reported respective brachytherapy planning for cervical cancer. Besides treatment planning, there is increased need also for improvements in treatment quality assurance (QA), both during the surgical procedure and throughout the therapy course. This is crucial for therapy success, since interfractional applicator position changes might have significant effects on the dose delivered to target volumes and organs at risk (OARs) [10‐14]. Image-guided real-time checks of applicator positions appear essential to account for patient-specific clinical requirements and to ensure accurate applicator arrangements. CBCT represents a suitable modality for the aforementioned imaging tasks.

CBCT scanners for interventional environments usually feature mobile designs [15‐18]. Concurrently, they show significant image quality limitations compared to conventional CT, especially regarding image artifacts [15‐18]. The clinical use of respective devices may require more effort than three-dimensional (3D) C‑arms [19], due to larger space requirements. In this respect, a compact device design is considered important for the smooth integration in interventional workflows. Furthermore, to enable patient examinations in the lithotomy position required for intraoperative brachytherapy, a large gantry clearance is crucial. One way to combine these requirements is the usage of vast flat-panel detectors (FPDs). These enable examinations of large anatomical regions of interests (ROIs) with one single gantry rotation and non-moving patient table. In addition to CBCT, radiography and fluoroscopy can be performed with FPDs in general. The combination of these imaging modalities particularly provides the possibility of 2D–3D image registration. This has shown to enable rapid assessments of surgical outcomes such as implant stability [20, 21].

On-site CBCT enables the cross-sectional visualization of the patient anatomy with applicators in situ directly at the brachytherapy ward. This allows both treatment planning as well as intraoperative and interfractional treatment QA measures. For this purpose, we rolled out the innovative CBCT system ImagingRing m (medPhoton, Salzburg, Austria) as the first site worldwide. The clinical introduction started in February 2021 and focused on patients with breast cancer and gynecologic malignancies. We present the first clinical images and experiences.

For six breast patients, a control-CBCT to planning-CT registration and reconstruction of catheter tracks was performed according to Kallis et al. [25]. In none of these cases, treatment adaption became necessary. All patients underwent tumor resection prior to brachytherapy and therefore had surgical clips marking the tumor bed. These were clearly identifiable on the CBCT scans (Fig. 3). A reasonable median CNR of 7.7 (range 5.9–9.2; Table 2) between catheters and surrounding tissue was obtained. Hence, the image quality was sufficient for treatment QA in each case. This was achieved at reasonable dose levels, with a median X‑ray exposure of 5.1 mGy (weighted cone-beam dose index CBDI_w, range 2.7–8.3 mGy). One breast patient did not comply with the given breathing commands, which led to severe motion artifacts. Another patient could not be examined due to technical problems.

The CBCT scans of gynecologic patients allowed in general distinct visualizations of implanted applicators, due to their high contrast to the tissue (Fig. 4). The courses of vaginal and intrauterine applicators were well identifiable for the majority of patients. In each of these cases, applicator position control could be performed. In none of these cases, relevant positional changes, shifts, or slippage were identified. The deviations of the applicator and needle distances measured on the control-CBCT to the ones documented on the planning-CT are provided in Table 3 for each patient. The applicator–transfer tube assignment was feasible by following the courses of the inserted X‑ray markers (Fig. 4, PG3). However, the exact position verification of implanted needles was not always possible. This was specifically the case when needles and particularly needle tips were in close proximity to sigma or rectum filled with intestinal gas. Here, streaking and shadow artifacts originating from these air–tissue transitions substantially impaired the image quality (Fig. 4). Under these circumstances, the CBCT scans were unsuitable for a complete treatment QA. This was the case for 5 of the 18 examined patients (Table 3).

Although the IRm was principally suitable for validating applicator positions, the device showed distinct weaknesses in the imaging of gynecologic patients. As mentioned, severe artifacts occurred at air–tissue transitions. In addition, directional noise originating from pelvic bones and hardening artifacts in the applicators’ vicinity further impaired image quality (Fig. 4). A clear delineation of organic structures (especially of cervix, uterus, and intestine) as well as the identification of vessels and ureters was not possible in general, due to insufficient contrast differences (Fig. 4). This assessment was quantitatively confirmed by the low CNR (median 2.1, range 0.2–4.8; Table 2) measured exemplarily between uterus and bladder. Thus, despite a high median CBDI_w of the examinations of 26 mGy (range 9–50 mGy), individual soft tissues could not be differentiated sufficiently. Note, that the examined patients had a median BMI of 23 (range 20–36; Table 2) and thus represented in general normal weighted patients [26].

Regarding the workflow, the gantry’s large diameter allowed working free and flexible at the patient, e.g., for connecting afterloader transfer tubes. The vast FPD combined with the VDW facilitated the acquisition of FOVs sufficiently dimensioned for applicator arrangement control, as well as exact FOV adjustments even for non-isocentric patient positions. Due to its motorized mobility, the system can be maneuvered time-efficiently into imaging position regardless of the exact patient table position within the room. The total time from patient positioning (including repositioning of immobile patients from bedside) to the examination end was <5 minutes for all patients. In each case, the CBCT scans were available fully reconstructed in <50 seconds after acquisition completion. Hence, the fast assessment of all images could always be performed directly on the HMI.

By means of the IRm, we created a time-efficient workflow for the QA of breast and gynecologic brachytherapy. The versatile battery-powered mobility allowed fast maneuvers of the device. The reconstructed images were quickly available, which enabled direct and fast assessments regarding decisions on potential therapy adaptions. In this respect, a mobile CBCT system on-ward strongly facilitates the brachytherapy workflow, since neither patients nor staff have to be sent to more distant CT scanners. This leads particularly to the technical feasibility of performing a treatment QA also for gynecologic patients, which was previously not practicable due to the high effort involved (e.g. transport to distant CT scanners). Such a high effort in particular also precludes the acquisition of magnetic resonance imaging (MRI) scans at off-site locations. On-site MRI for interventional purposes, which would be most desirable due to the superior soft tissue delineability [27‐29], is logistically still not implementable at many brachytherapy sites at all.

For none of the examined patients, the necessity for therapy adaption was diagnosed. Despite this positive experience, performing a treatment QA is considered important for therapy success. For breast patients, Kallis et al. [25] showed that treatment adaptions are indicated in 4% of all cases. For gynecologic brachytherapy, other authors already found substantial variations of the dose delivered to target volumes and OARs throughout the treatment course [10‐14]. These were related to both inter- and intrafractional anatomical variations (e.g., organ and patient motion), which may lead to altered applicator positions with significant dosimetric impact [10‐14]. It is one very important precondition of adaptive brachytherapy to detect such applicator arrangement changes and to adapt the treatment instantaneously to the altered conditions. In this respect, the implementation of automated procedures for treatment plan verification and immediate re-planning (in case of plan invalidity) is most desirable. Hence, this is the subject of ongoing research [10, 30, 31]. In summary, it is important for the precise therapy of patients to perform a treatment QA. CBCT represents a functional and smooth modality for this purpose.

For breast patients, the IRm’s image quality fulfilled clinical requirements at reasonable dose levels. The obtained CNR between catheters and tissue of at median 7.7 was approximately in the range of planning-CTs (median CNR of about 9, measured as described in the “Assessment of structure differentiability” section). Thus, the CBCT scans enabled distinct catheter reconstructions. Treatment QA for breast patients was therefore feasible by using the IRm scans without major drawbacks compared to conventional CT images. However, the current scan times of the IRm of 16–18 seconds are a disadvantage compared to conventional CT (acquisition time in our clinic: <6 seconds). This was also evident from the fact that one examined patient was not able to hold her breath. In particular, based on our experience, not all patients are able to follow correspondingly long breathing commands immediately after surgery. However, this would be required for the acquisition of planning-CTs. To avoid potential motion or breathing artifacts, it is therefore desired to reduce the IRm’s acquisition time. Currently, this associated with enhanced artifacts and decreased circularity of catheters due to partial undersampling. The scan time reduction with undiminished image quality is therefore subject of ongoing work.

The CBCT scans of gynecologic patients were acquired with a median CBDI_w of 26 mGy. This exceeds the dose of comparable devices (CBDI_w about 20 mGy) [32] and also conventional CT systems (CTDI_vol ≤15 mGy) [33, 34]. Despite this increased radiation exposure, distinct differentiations of pelvic organs and tissues were not achievable with the IRm. Target structures and OARs could only be inadequately separated. This was quantitatively confirmed by the CNRs calculated exemplarily between uterus and bladder. With a median CNR of only 2.1, the scans did generally not fulfill Rose’s criterion (Rose’s criterion states that a structure must show a CNR of at least 3 to 5 to be detectable) [35]. This was clear evidence for insufficient low-contrast differences. Please note again that the examined patients represented with a median BMI of 23 in general normal weighted patients, as reported in the “Results” section [26]. The inadequate CNR is therefore not directly attributable to extreme patient characteristics (such as it might be the case for very obese patients) accompanied by strong object scatter. Rather, it indicates a lack of CT-number contrast in the device’s imaging performance. For comparison, planning-CTs yield a respective median CNR of about 6 (again measured as described above). Furthermore, as mentioned, the IRm showed severe artifacts in the imaging of gynecologic patients.

Based on these observations, we conclude that the routine use of the IRm for the purpose of differentiating pelvic soft tissues is currently not justifiable. In particular, the IRm can currently also not support the intraoperative, image-guided insertion of applicators. This is because (due to the lack of organ differentiability) no reliable prediction can be made in which direction and how far applicators and needles can be inserted without risking injuries of corresponding OARs. In this respect, the imaging performance of the IRm is inferior to conventional CT. However, as reported, the device can be used for the verification of the implant stability in-situ and showed to be in principle applicable for the treatment QA of gynecological patients. Although for some patients (Table 3) image artifacts substantially impaired the precise detection of individual needle tips, the IRm enables imaging for treatment verification directly on the ward. This is a procedure that was previously very impractical in our hospital due to the high effort involved, as described above. We think that the advantage of such an on-site treatment verification outweighs the disadvantage of a suboptimal image quality and can have a large benefit for individual patients.

From our perspective, intraoperative CBCT-guided applicator insertion might represent an important step for adaptive gynecologic brachytherapy at many brachytherapy wards. However, to support this by the IRm, significant improvements of its imaging performance are imperative. This relates in particular to improvements in image contrast and soft tissue delineation. Nonetheless, with CBCT it appears feasible to detect the position of applicators precisely, especially around the fundus uteri or deep in the pelvis (in regions for which transrectal sonography is not or only limited applicable). For this, CBCT imaging is appropriate and beneficial. The principle of mobile, non-isocentric imaging also enables patient examinations directly in the surgical theater. This offers a possibility to improve the accuracy of implantations and consequently of treatment quality. A large gantry, as provided by the IRm, further enables imaging in the lithotomy position. This procedure was previously only possible to a limited extent with other gantry-based CBCT-CT/systems. Hence, we think that mobile CBCT imaging with the IRm has in principle high potential for image-guided adaptive brachytherapy of gynecologic malignancies. However, as mentioned, improvements of the device’s imaging performance are imperative.

By means of a mobile device for on-site CBCT imaging, we implemented a time-efficient workflow for the QA of breast and gynecologic brachytherapy. However, significant performance enhancements of the IRm are strictly required for treatment planning imaging and, most desirable, intraoperative image-guidance.