A new workflow of the on-line 1.5-T MR-guided adaptive radiation therapy

In December of 2021, radiation therapy utilizing the 1.5-T MR-Linac system (Elekta AB, Stockholm, Sweden) was initiated. This innovative system merges a 7MV flattening filter free (FFF) Elekta linear accelerator with a Philips 1.5-T MRI scanner (Philips Healthcare, Amsterdam, the Netherlands) [1‐4]. This cutting-edge MR image-guided radiation therapy (IGRT) allows for a timely and appropriate treatment re-planning to be conducted while the patient is situated on the treatment couch [5, 6]. This facilitates an immediate adaptive radiotherapy approach based on both the position of the tumor and normal tissue during daily treatment, using near-real-time beam-on cine MR imaging during treatment. This procedure, termed MR-guided on-line adaptive radiation therapy (on-line MRgART), necessitates meticulous off-line treatment planning and a seamless on-line process on the day of the treatment. However, the actual on-line MRgART method varies across institutions, with no standardized methodology having been established thus far [7‐14]. In this paper, we present the establishment of a novel workflow for on-line MRgART at our institution, as well as an evaluation of the feasibility of this new workflow in regard to meeting dose constraints.

On-line MRgART using Elekta Unity (Fig. 1) was initiated in December 2021. The basic concept and structure of this treatment machine has been described by Raaymakers et al. [4]. The system is composed of a 1.5-T MRI magnet, surrounded by a large rotating gantry of linear accelerators, on which a magnetron, a tri-pole electron gun, and an S-band standing wave accelerating tube are located. The magnetic field strength of the donut-shaped space containing the electron gun is designed to be almost zero while maintaining the uniformity of the main magnetic field so that electrons emitted from the gun reach the acceleration tube. The therapeutic beam is delivered through a slit in the center of the magnet. Basic specifications are outlined in Table 1. The 1.5-T MR-Linac allows for MR imaging in off-line treatment planning, on-line treatment planning, and after treatment irradiation completion. Near-real-time two-dimensional cine MR images at 5 frames per second can be obtained in three cross sections prior to and during the treatment beam delivery, also known as beam-on imaging.

At the end of September 2022, a total of 50 patients consecutively received MRgART using this treatment procedure. The number of radiation fractions administered to each patient varied between 2 and 8, with a median of 5. Ninety percent of the patients (45 out of 50) were treated in 5 fractions. The patients had various types of cancer, including prostate cancer (20 patients), hepatocellular carcinoma (10 patients), pancreatic cancer (6 patients), lymph node oligo-metastasis (5 patients), renal cancer (3 patients), bone metastasis (3 patients), liver metastasis from colon cancer (2 patients), and mediastinal tumor (1 patient).

Out of a total of 247 fractions administered to the 50 patients, most of the fractions (95%) involved adaptive re-planning using ATS procedure, with only 12 fractions using ATP procedure. For all patients, the time required from pre-treatment MR image acquisition to the start of actual irradiation after on-line adaptive re-planning in total fractionated irradiations was recorded. The median of the average time for ATS re-planning in all 50 cases was 17 min, ranging from 3 to 58 min. For each disease site with three or more cases, the median time required for ATS re-planning was 46 min for pancreatic cancer, 18 min for hepatocellular carcinoma, 17 min for prostate cancer, 14 min for lymph node oligo-metastasis, 17 min for renal cancer, and 7 min for bone metastasis (Table 3). On the other hand, the median time required for re-planning by ATP was 2 min, with a range of 1 to 7 min.

Of all 247 fractions, there were interruptions in a total of 25 on-line MRgART procedures; patients had to interrupt their treatment due to urination or other discomfort in 7 fractions, and patients had to get off the couch due to significant changes in organ position in 15 fractions. Also, in three fractions, the treatment was interrupted due to the troubles with the equipment. However, finally, the treatment was completed in all fractions.

In the feasibility assessment of dose constraints in five prostate cancer patients, all dose constraints were successfully met in all 5 patients using ATS-based re-planning (Plan 1). In contrast, a total of 14 dose constraints among the 5 patients could not be met using Plan 2, in which only the irradiation position was adjusted after contour modification (Table 4). The main dose constraints that could not be met were D1cc of the PTV, Dmax of the rectal wall, and Dmax of the bladder wall. Figure 3 shows a case in which the dose constraints of the rectum wall and bladder wall could not be met with a virtual plan (Plan 2) but could be met with an ATS-based adaptive re-planning.

On-line MRgART is a novel technique that employs near-real-time MRI to precisely guide the administration of radiation therapy. This state-of-the-art approach provides visualization of tumors and surrounding normal tissue during radiation treatment, thereby enhancing treatment accuracy and minimizing damage to normal tissue. Compared to traditional CT-based image-guided radiation therapy, the benefits of 1.5-T MRgART are numerous. First, radiation therapy is delivered with superior accuracy owing to the high contrast resolution of 1.5-T MR images, which provide detailed information about the location, size, and shape of the tumor. Second, MRgART enables near-real-time visualization of tumors and surrounding tissue during radiation treatment, thereby allowing for the direct monitoring, capture, and control of irradiation without the need for additional radiation exposure or invasive insertion of metal markers for image guidance. This allows for immediate adjustments to the radiation delivery, ensuring that the tumor receives the maximum dose while minimizing exposure to normal tissue. Finally, adaptive re-planning based on MR images immediately prior to irradiation can recreate optimal dose distributions based on the patient's individual anatomy and response to treatment [17‐19]. These advantages are expected to result in more precise dose delivery to the target, reduced radiation exposure of risk organs, and improved DVH parameters. This is particularly relevant for tumors located near critical organs or structures. Online MRgART thus expands and promotes the use of increased fractional doses, fewer fractionations, and hypo-fractionated irradiation in the treatment of various types of cancer. For pancreatic cancer, prostate cancer, lymph node recurrence, etc., excluding brain tumors and/or head and neck tumors, the number of fractions is typically five, according to the 2020 Consortium Meeting [18].

The Monaco for Unity treatment planning device has two key functions: ATP procedure and ATS procedure. In ATP, the MR image obtained immediately before treatment is matched with the planned MR image via rigid registration, the shift value of the irradiated area is calculated, planning and optimization are performed again, and dosimetry of the target and risk organs are evaluated. In ATS, the contour shape is compared between the MR image obtained immediately before treatment and the treatment planning MR image, and then re-planning is performed after correction of anatomical shapes. The contour information of the tumor and normal tissue depicted in the MR image obtained immediately before treatment is newly adopted, and planning and optimization are performed again using the electron density obtained from CT for each contour. While ATP is a re-planning process that primarily corrects the position, ATS constitutes a complete re-planning process that utilizes the new contour information, and therefore requires more time [5, 11, 13, 14].

We have implemented an institutional workflow for on-line MRgART. We deem this approach feasible because of its adherence to dose constraints, as demonstrated in the feasibility study. MR-based IGRT allows accurate delineation of tumor and risk organs due to its superior contrast resolution compared to CT. Certainly, re-planning using MR images taken just prior to treatment is more time-consuming and labor-intensive than CT-based radiotherapy. However, this can be mitigated by pre-drawing the tumor and organ contours during MR simulation with the treatment device itself. This permits delineation of the contours on new MR images on the day of treatment with minor modifications based on the MR of the reference plan. In most cases, ATS was required for re-planning, although employing the deformable image registration (DIR) technique necessitated only minor manual contour corrections. Compared with other reports [11, 14], the unique feature of our workflow is that the MR-Linac itself, rather than other MRIs such as simulator or diagnostic MRIs, is used to take the treatment planning MRI and create the reference plan. The ability to use images taken with the same bed and coils on the same device to compare a patient's anatomy between pre-treatment MR images and those of the reference plan is significant. It reduces the time required for contour correction and is expected to have minimal effect on DVH and other parameters due to the good consistency of MR images. The time required for modifying organ contours was shorter for liver cancer, renal cancer, and lymph node oligo-metastasis due to the limited number of gastrointestinal tracts bordering the tumor. However, the time required for on-line re-planning for pancreatic cancer was relatively longer as the pancreas is surrounded on three sides by the stomach, small intestine, and large intestine.

The workflow for on-line MRgART we have established seems to contribute to the improved throughput of the daily treatment procedures owing to the effective cooperation of the medical team and resultantly to the learning curves of the staff members.

In conclusion, we have established a workflow for on-line MRgART using the newly implemented Elekta Unity 1.5-T MR-Linac system. The initial evaluation of the dose distribution and DVH demonstrated that the established workflow was feasible. As the clinical practice of on-line MRgART has only recently begun, we anticipate further improvements and evolution in future.