First prospective clinical evaluation of feasibility and patient acceptance of magnetic resonance-guided radiotherapy in Germany

All patients underwent MR simulation directly at the treatment machine. Thereby, not only were MR images for treatment planning generated, but tolerability of immobilization and placement of receiver coils, patient compliance to breathing instructions, and ability to perform inspiration breath-hold were also assessed. Depending on the treatment region, 3D simulation MR images were acquired either in free breathing or in inspiration breath-hold, followed by planar cine-MRI in a sagittal plane to evaluate target motion characteristics. The pulse sequence used for both was always a trueFISP sequence [14], which is the only pulse sequence currently used clinically on the system. The acquisition time of 3D simulation MR images ranged from 17 s in breath-hold to about 3 min for pelvic scans in free breathing, with an in-plane resolution of 1.5 × 1.5 mm² and slice thickness of either 1.5 mm or 3 mm. MR simulation was carried out without administration of contrast agent. Subsequently, native CT simulation scans were performed on the same day for each patient with identical immobilization devices and simulation dummy coils, usually reproducing the same breathing state as in the MR simulation. Additional contrast-enhanced CT scans were acquired if no contraindications were present (e.g., renal dysfunction or allergies against contrast media). For treatment planning, the integrated treatment planning system (TPS) of the MRIdian Linac was used. Following simulation, MRI and CT datasets were transferred to the TPS and deformably registered using the vendor-supplied deformation algorithm, which for multimodal deformable image registration iteratively tries to minimize a dissimilarity measure computed from mutual information. When available, diagnostic MR images were additionally imported and rigidly registered with focus on the region of the GTV. Gross tumor volumes (GTV) comprised the sum of macroscopic tumor delineated in all co-registered modalities. Depending on the region to be treated, clinical target volumes (CTV) were expanded from the GTV using margins of 1 mm (for example pelvic lymph nodes) to 5 mm (for example hepatic metastases). An isotropic planning target volume (PTV) margin of 4 mm was added in order to account for technical inaccuracies.

For all patients, the simulation MRI datasets were chosen as primary image sets for treatment planning, in order to facilitate same-modality image registration during daily treatment. Electron density information for dose calculation was derived from the registered CT scans. Step-and-shoot treatment plans were generated using the dedicated TPS, where dose calculation was always performed via Monte Carlo dose calculation taking into account the static magnetic field.

Daily image guidance was performed for each fraction by onboard 3D MRI using identical settings (field of view, duration, pulse sequence, breathing instructions) as during MR simulation. Soft tissue-based registration with the reference MR scan was applied, usually registering directly on the GTV, and the couch shifted accordingly.

Real-time MR gating was used for all patients for whom respiratory movement of the target was detected during MR simulation. All respiratory-gated treatments were performed in inspiration breath-hold. A gating target (region of interest, ROI) was defined (either the tumor/resection cavity or a surrogate, e.g., vessel, bronchus in close proximity), and the gating boundary was selected as a numerical margin added to the target ranging between 3 and 4 mm at the discretion of the treating physician. Due to uncertainties in deformable registration and image noise possibly causing errors in contour tracking [17], a small percentage of target outside of the predefined boundary without triggering a beam-off event was allowed (threshold-ROI%) [17]. Threshold-ROI% was set to 3% for most of the cases and to values up to 7% in isolated special cases. The system software automatically stopped radiation delivery when this threshold was exceeded. If intrafractional changes in target position were detected in the cine-MR, no two-dimensional table shifts were performed. Instead, the treatment was always interrupted and a repeat volumetric MRI scan was performed to allow a 3D positional correction.

During gated delivery, patients were provided with visual guidance via an in-room screen displaying the live sagittal cine-MR image with an overlay of gating target and boundary, thus enabling them to steer their repeated breath-holds to the right position. This idea was adopted from other published in-room screen solutions for MRgRT [17, 20]. Patients were able to observe this monitor during the whole treatment with the help of a mirror fixed to the bore (see Fig. 1). If necessary, patients received assisting breathing commands in order to find the optimal breath-hold. In addition to respiratory gating, the gating functionality of the system was also used for target tracking on patients in whom the target did not move with respiration, in order to ensure that no other movements happened during treatment [21].

Patient-reported acceptance of the whole treatment procedure was documented using an in-house developed patient-reported outcome questionnaire (PRO‑Q; see Table 1), which was completed after the first fraction, weekly during the treatment, and after the last fraction. The PRO‑Q consisted of questions regarding potential MR-related experiences and complaints (e.g., noise, bore size, fixation with coils). For patients undergoing respiratory gated treatments with audiovisual feedback, the perception of their active role was additionally evaluated. Items were scored using a five-point scale.

In addition to the experience reported by the patients, the staff on the system (therapists and physicians) were questioned about overall patient compliance for each patient after the first and last fraction. They were asked to score the overall compliance of the patient with the treatment procedure on a scale from 1 (very uncomplicated) to 10 (very complicated).

From April 2018 to April 2019, 43 patients were treated on the MRIdian Linac system, with a total of 428 fractions. Patients had a mean age of 64 years (range 32–87) at the beginning of treatment, were mainly male (58%), and had a median Karnofsky performance score of 80%. Patient characteristics and treated tumor sites are illustrated in Table 2. The most frequently treated anatomic sites were abdominal nodal metastases and liver lesions. MRgRT was selected for these patients due to a variety of reasons, including superior soft tissue contrast or gated dose delivery and mostly for a combination of these factors. For five patients with bony lesions, a benefit of soft tissue contrast was assumed because the lesions were only limitedly visible on CT scans and had been diagnosed with positron-emission tomography (PET) or scintigraphy before. Three patients with centrally located pulmonary lesions were also treated on the MRIdian Linac, as daily MRI facilitated better distinction of the target volumes from the mediastinum. Stereotactic body radiotherapy (SBRT) was applied in 20 patients (47%). Total applied doses ranged from 4 to 66 Gy, with single doses ranging from 2 to 15 Gy. The mean number of fractions per patient was nine (range 2–33). For another 23 patients, MRgRT was initially foreseen but could not be performed due to different reasons (see Table 3).

Completed questionnaires were available for 34 patients. In total, patients rated MRgRT as positive or at least tolerable, with mean scores of 1.0–3.6 in the 14 main questions (see Table 4). No statistically significant changes were detected between the first fraction and the end of treatment for all assessed questions (p ≥ 0.05).

However, several patients (65%) reported some degree of potential MR-related complaints at least once (score ≥4). Patients mainly complained about the temperature in the room (24%) and of some particular body parts (27%). Furthermore, 18% of the patients experienced paresthesia during treatment and 12% rated the positioning as well as having to lie still for at least half an hour during treatment negatively (score ≥4). The size of the MRI bore and the duration of treatment were classified as at least narrow or long, respectively, by 6% of the patients. Despite the routine use of headphones, 3% of the patients scored the noise of the MRI as disturbing.

The sub-section of the questionnaire related to breath-hold-gated dose delivery with audiovisual feedback was completed by 22 patients. No patient reported severe difficulties in controlling the target by holding his/her breath. Additionally, no patient answered that watching his/her tumor on the monitor during treatment was considerably disturbing. All patients seemed to appreciate their active role in controlling the duration of treatment.

Patients tolerated the treatment very well with only mild acute toxicity. In detail, patients reported fatigue CTCAE grade I (n = 19), nausea CTCAE grade I (n = 12), coughing CTCAE grade I (n = 7), flatulence CTCAE grade I (n = 6), diarrhea CTCAE grade I (n = 4), dyspnea CTCAE grade I (n = 2), dysphagia CTCAE grade I (n = 2), dyspepsia CTCAE grade I (n = 7), and pain in the thoracic wall CTCAE grade I (n = 2). Except for 4 patients with fatigue CTCAE grade II, no acute toxicity ≥ CTCAE grade II was detected. 11 patients did not describe any aggravation of pre-existing symptoms or the occurrence of new symptoms after RT.