The transformation of radiation oncology using real-time magnetic resonance guidance: A review

Abstract

Radiation therapy (RT) is an essential component of effective cancer care and is used across nearly all cancer types. The delivery of RT is becoming more precise through rapid advances in both computing and imaging. The direct integration of magnetic resonance imag-ing (MRI) with linear accelerators represents an exciting development with the potential todramatically impact cancer research and treatment. These impacts extend beyond improved imaging and dose deposition. Real-time MRI-guided RT is actively transforming the work flows and capabilities of virtually every aspect of RT. It has the opportunity to change entirely the delivery methods and response assessments of numerous malignancies. This review intends to approach the topic of MRI-based RT guidance from a vendor neutral and international perspective. It also aims to provide an introduction to this topic targeted towards oncologists without a speciality focus in RT. Speciality implications, areas for physician education and research opportunities are identified as they are associated with MRI-guided RT. The uniquely disruptive implications of MRI-guided RT are discussed and placed in context. We further aim to describe and outline important future changes to the speciality of radiation oncology that will occur with MRI-guided RT. The impacts on RT caused by MRI guidance include target identification, RT planning, quality assurance, treatment delivery, training, clinical workflow, tumour response assessment and treatment scheduling. In addition, entirely novel research areas that may be enabled by MRI guidance are identified for future investigation.

1. Introduction

Clinical evidence supports the use of radiation therapy (RT) in 50% of all cancer patients. With global cancer cases to reach 24.9M by 2035, further advances in RT are important to improve cancer outcomes and to minimise side effects [1]. Image-guided RT (IGRT) has represented an important advance in RT for well over a decade [2]. IGRT has been widely adopted by the radiation oncology community and is used in the majority of radiation treatments [3,4]. Radiation oncologists are amply trained in the acquisition and interpretation of computed tomography (CT) images used for IGRT. Contemporary IGRT allows for increasingly precise target localisation along with tumour and normal structure position verification in three dimensions [2].

Magnetic resonance imaging (MRI) offers superior soft tissue discrimination, increased sensitivity for tumour detection and dynamic biological and functional data about tumours and normal tissues. MRI has been used for well over a decade to help define and direct RT volumes in many cancer sites. MRI’s role in RT has largely remained limited to the initial radiation planning stages, that is, before treatment begins. MRI devices have recently merged with radiation delivery devices (linear accelerators) to form an entirely novel RT paradigm categorised as MR-guided RT (or ‘MRgRT’). MRgRT possesses the ability to acquire an MRI immediately before, during and after the patient is treated with RT, all with the patient in the same treatment position. In this review, we summarise the current approaches to MRgRT and highlight its implications and opportunities on oncology at large.

2. Methods

An experienced cohort of radiation oncologists and medical physicists were assembled, bringing together users of the two commercially available MRgRT systems. Each author had intimate familiarity with the logistics and technical challenges associated with MRgRT in the two currently available commercial units. A literature search was conducted via PubMed using the key words ‘MR/MRI-Guided RT (766 items reviewed)’, ‘MR Linac (248 items reviewed)’, ‘ViewRay (54 items reviewed) ‘, and search results were curated. We restricted our search to English-language articles published from 1995 to 2019 and those that focused on single unit MRgRT devices. Abstract only publications were excluded. Research articles focused on combining MR images with RT treatment were also excluded as they were extensive in nature and considered beyond the scope of this review. The literature search was restricted solely to series focused on the combination MR-linear accelerator-based devices; articles examining in-room MR solutions or ‘MR on rails’ were also excluded.

3. Discussion and observations

The current state of MRgRT is rapidly evolving and expanding. This technology is positioned to dramatically impact the treatment of thousands of cancer patients annually. It will also rapidly introduce entirely novel research questions and opportunities. MR guidance has some similarities, but also crucial differences, to other recent technological developments in the delivery of RT. Fig. 1 presents a general overview of workflow differences introduced with MR guidance.

Fig. 1.

Differences between CT and MR-based RT. RT, radiation therapy; CT, computed tomography; MR, magnetic resonance.

RT has undergone several transformations over the past fifteen years. The routine use of CT simulation required a primary knowledge expansion on the part of radiation oncologists who had to familiarise themselves with cross-sectional and three dimensional anatomy as depicted on CT-based imaging. Since the introduction of CT-based planning, other major technological advances in the form of intensity-modulated RT (IMRT) and proton therapy have been introduced. The introduction of IMRT represented a considerable improvement in radiation dose sculpting and normal tissue avoidance. Protons and heavy ions represent further gains in dosimetric conformality. MRgRT is unique in its contribution of extremely high quality imaging at the time of treatment. To help illustrate the impact of MRgRT, we have divided the discussion into several categories outlining the current and anticipated changes to the discipline of RT following the introduction of MRgRT.

3.1. Differences between MRgRT and other technological advances in RT

While IMRT and proton therapy have resulted in important improvements in radiation dose deposition, neither has addressed two central problems facing RT. The specific challenges that remain inadequately addressed include (1) high doses of radiation delivered to normal organs in very close proximity to the treated tumour (which is often unavoidable secondary to the need for a planning target volume [PTV] expansion to account for daily set up uncertainties) and (2) personalisation of RT via active monitoring of biologic tumour response. Total dose deposition and conformality of dose have been an understandable focus of radiation oncology technological advancement for decades. This has motivated the development of IMRT, IGRT and proton therapy. Important to realise is that radiotherapy planning involves delivering dose not only to the tumour volume but also to a rim of surrounding normal tissue to take into account systematic and random errors such as calibration uncertainty and organ motion. While the high-dose conformality with IMRT and proton therapy have enabled dose escalation, the radiation oncology community has seen that dose escalation alone may fail to improve outcomes [5]. This might be limited by the need for a PTV expanding into some critical local structures. Indeed, CT-based strategies for dose escalation are often compromised by the large uncertainty of tumour and normal structures with low soft-tissue contrast on CT images. Moreover, current CT-based methods suffer from an incapability of monitoring tumour and normal structure movement during radiation delivery. MR guidance directly addresses and improves upon these issues. The ability to determine the location of the tumour and adjacent normal organ/tissue boundaries, together with the ability to account for intratreatment and intertreatment motion, will reduce radiation dose to normal organs, thereby widening the therapeutic window. A summary of major technologic developments in radiation oncology and their potential contributions to clinical oncology research can be seen in Table 1.

Table 1.

Technological advances in RT.

Technological development	Changes to RT	Clinical trials enabled By technological development	Limitations of technological development
CT-guided intensity-modulated RT (IMRT) [30]	Significant improvements in radiation dose conformality and reduction in radiation doses to critical normal structures Ability to avoid or reduce doses to local structures	Reduction in doses of radiation to critical organs at risk (e.g. parotid glands in head and neck cancer) [31] Dose escalation strategies	Expensive Insufficient soft-tissue contrast, limiting treatment adaptation Limited ‘beam-on’ tumour monitoring during treatment delivery
Proton therapy	Reduction in radiation dose to normal structures Ability to completely eliminate dose to some normal structures	Trials focused on clinical improvements enabled by reductions in moderate radiation doses to normal tissues [32,33]	Very expensive Insufficient soft-tissue contrast in image guidance Inability to reduce highest radiation doses to normal organs Limited treatment adaptation
MR-guided intensity-modulated RT	Substantial improvement in soft tissue imaging during treatment Online treatment adaptation based on MR-defined ‘anatomy of the day’ Beam-on imaging with MR may enable considerable reductions in high doses to normal organs Detection of radiation response of tumour and normal structures	Anatomically and biologically adaptive RT, based on changes seen on daily MRI during a treatment course (e.g. daily DWI) Intra-treatment, or ‘beam-on’, monitoring of normal organ and tumour movement NTCP reduction with dose reduction, particularly of local structures in close proximity to the tumour	Expensive Risks and complexity associated with the use of MRI New training needed Novel effects of MRI on radiation dose distributions Limited patient eligibility (not an option for those with contraindications to MR-based imaging)

MR, magnetic resonance; DWI, diffusion weighted image; NTCP, normal tissue complication probability.

3.2. Existing technological implementations of MRgRT

At the time of writing this article, there are two commercially available MRgRT technologies. These devices are manufactured by ViewRay (ViewRay Technologies Inc, Oakwood Village, Ohio) and Elekta (Elekta AB, Stockholm, Sweden). There are also at least two devices that are in active development: one is by an Australian-based development group [6] and the second is the Aurora-RT system (MagnetTx Oncology Solutions, Edmonton, Alberta, Canada) [7]. These devices are gaining rapid and wide market traction (eg. 36 Elekta Unity systems have been sold and approximately 51 from ViewRay). Key differences between devices are presented in Table 2. The discussion will focus on the two commercially available devices, namely the ViewRay MRIdian and Elekta Unity MR Linac systems. To date, ViewRay has produced two different systems consisting of their first device, a split 0.35 T magnetic resonance scanner with a ring gantry and 3 multileaf collimator-equipped ⁶⁰Co heads (no longer in production) followed by their second device, capable of 6 MV photon production combined, again, with a 0.35 T MR [8]. The ViewRay MRI-cobalt device has been US Food and Drug Administration (FDA) approved since May 22, 2012, and Viewray MRIdian linear accelerator has been approved since February 24, 2017 and the Elekta Unity system received FDA approval on December 5th, 2018. The Elekta Unity system is a 1.5 T MR produced by Philips combined with a 7 MV linear accelerator produced by Elekta [9,10]. There have been over 5000 patients treated with these systems, with multiple published series describing the initial clinical experience and case mix [12,23,34–48]. Details regarding each of these systems are presented in Table 2.

Table 2.

Types of MR Linac technologies.

Commercial name	Manufacture	MRI field strength	Bore size	Beam strength	Clinical outcome publicationsa
Commercially available
ViewRay [8]	ViewRay Technologies Inc,	0.35 T	70 cm	Co-60 source	[12,23,34–48]
- Co-60	Oakwood Village, Ohio			6 MV
- Linac
Elekta	Elekta AB, Stockholm, Sweden	1.5 T	70 cm	7 MV	[24,50,51]
Unity [49,50]
In development
Australian MRI Linac System [6]	Australian MRI-Linac Program	1 T	82 cm	6 MV	N/A
Aurora-RT System [52]	MagnetTx, Edmonton, Alberta, Canada	0.6 T	60 cm	6 MV	N/A

MV, megavoltage; cm, centimetres; Co-60, Cobalt-60.

^aClinical outcome publications involved the treatment of patients with reported outcomes.

3.3. Implications of MR guidance on radiation treatment volumes

As mentioned previously, at the core of RT is the PTV [11]. This PTV margin is needed to account for interfractional and intrafractional variability of setup of the tumour and normal structures. Typically ranging from 3 mm to 15 mm, the PTV structure extends considerably into adjacent normal organs or tissues that do not contain tumour. The need to incorporate a PTV margin with the associated exposure of adjacent normal structures currently limits the dose and fractionation schemes to what is tolerated by normal tissues rather than what is required to achieve tumour control. A fundamental change that will likely be seen with MR guidance is that using MRI immediately before and during a treatment delivery will enable accurate delineation and monitoring of tumour and normal structures at every treatment. This will result in much smaller irradiated volumes. For example, the elimination of PTV margins in the treatment of breast cancer has resulted in a 52% reduction in the PTV volume, which likely has important implications for cosmetic outcomes [12]. This will result in lower doses of radiation to normal structures very close to the tumour. Fig. 2 visually illustrates differences, and potential benefits, of smaller PTV expansions.

Fig. 2.

Examples of differences in expansions potentially enabled with real-time MR guidance. MRI, magnetic resonance imaging; PTV, planning target volume.

While exciting and promising, this change also has important implications on physician and other radiation caregiver time. Such an MRI guidance strategy requires a radiation oncologist, medical physicist and radiation therapist to spend additional time at the machine to adapt treatment delivery on a daily basis, as tumour and normal organ changes are detected [13]. Fig. 1 highlights some of these key differences in workflow that may be presented with MRgRT. Radiation oncologists, medical physicists and radiation therapists will have to adapt their schedules to account for the additional time necessary to monitor the tumour and normal organs now seen continuously with MRgRT.

3.4. Training and education in MRgRT for radiation oncologists

Routine use of MRI will require ‘up-skilling’ radiation oncology professionals to acquire an enhanced MRI knowledge base and skill set. Although MR has benefits at the time of simulation, it is not routinely used across the world. To optimise response evaluation, regional organ definitions and data provided by various MRI sequences, these sequences will need to be understood by radiation oncologists. Radiation oncology societies will need to work with MR societies to develop educational programs to ensure radiation oncologists, medical physicists, dosimetrists and radiation therapists/radiographers are adequately trained in MR utilisation, assessment and safety. In addition, the images acquired during the process of radiation treatment planning and delivery may be of value to other specialities, particularly in medical systems in which MR resources are scarce. An infrastructure through which treatment images could be easily shared with oncology colleagues may also be necessary for radiation oncology departments to consider.

3.5. Reimbursement associated with MR guidance

Additional physician and medical physicist time and effort will be a component of the MRgRT workflow. This includes, but is not limited to, time spent at the treatment machine in delineating tumour and normal structures, adapting treatment plans based on the daily MRI, observing real-time tumour and normal structure changes and evaluating treatment response from functional imaging [14]. Professional society government/payer relation groups involved with developing reimbursement codes may need to consider if additional billing codes are needed (and if they are justified) to account for this type of treatment delivery. Until such reimbursement codes are clarified, guidance as to the use of existing codes will be necessary to appropriately account for the time, effort and risk involved in the delivery of MRgRT. It is imperative that radiation oncology research groups collect data to establish the value of these added efforts through high-quality, peer-reviewed, prospective research by showing measurable clinical improvements in cancer specific outcomes and/or reduction of toxicity as compared to existing radiotherapy delivery technologies. To justify this additional time, effort, and cost, value must be proven. For this purpose, efforts are underway through two active research consortia to collect such data to prove the value-added of these technologies [15,16].

3.6. Unique challenges associated with MR-guided RT

RT deposits radiation dose via secondary charged particles, primarily electrons. The exposure of these electrons to a strong magnetic field changes the manner in which radiation dose is deposited. One well described effect, entitled the electron return effect, presents such an example of this challenge [17]. This effect is the result of electrons moving in a circular pattern in the presence of a magnetic field, as opposed to a more linear path in the absence of a magnetic field. This effect results in a complicated radiation dose effect, particularly at tissue–air interfaces. Fortunately, advanced treatment planning software can model this dose effect, and this can be accounted for in the process of RT planning [18]. It should also be considered that dose calculation in the presence of B-fields is a novel and important challenge. MR images are also subject to geometric distortion which can alter the appearance of normal anatomy or the target. Such distortion must be carefully considered during a treatment course [19,20]. The impact of immobilising patients in a confined space for extended lengths of time for image acquisition and treatment delivery must also be carefully considered. Claustrophobia is a common concern, and the impact on patient-reported quality of life has been examined; in total, approximately 5% of patients seem to have found this treatment unacceptably long [21]. Moreover, robust quality assurance methods of radiation treatment plans may require an entirely novel approach, given the influence of the magnetic field [22]. There is also a necessity for compromises in the functionality of both the MRI device, along with the radiation delivery device when combining these units given the technical complexity of combining these devices. Such compromises may lead to longer treatment times. During this treatment time, there may be an increase need to account for the movement of normal organs. In addition, the process of online adaptive replanning will introduce entirely novel challenges, with regard to physician time and workflow. Finally, MRgRT will present unique challenges with increasing need for improvements in mechanism for automation, archiving imaging, deformable image registration, exquisite Rad Onc attention to ferrous materials and dose accumulation using MRI.

3.7. Future directions and routine MR-based tumour response monitoring

The current clinical experiences of the MRgRT systems have been focused on daily plan changes based on geometric changes in the organs at risk [23,24]. These have formed the basis of multiple ongoing prospective studies, especially in pancreatic cancer (NCT01972919, NCT02283372, NCT03621644). However, the collection and application of these MRI-based data may enable the delivery of the most effective dose to individual tumour biology rather than the highest dose of radiation that can be delivered to a given tumour based on histology, stage and location. Reduction in the PTV and improvements in tumour targeting will represent important, if linear, steps forward towards better and more conformal treatments for patients treated with MRgRT. While they are critically important, such innovation by itself would not represent a true transformation of RT. The ability to routinely acquire daily, functional MRI and subsequently act on those data, directly in the treatment setting, presents an entirely new paradigm for the speciality. Imaging-based biomarkers will represent the future of cancer treatment delivery, and numerous candidate biomarkers have been extensively discussed and published [25]. Radiation oncology is well positioned to embrace this shift, and MR guidance is optimally suited to enable response-based, personalised RT. Indeed, the future of a biological, image-guided adaptive RT (BIGART) is very exciting. There are multiple established, MR-based response metrics in patients undergoing treatment for cancer. A summary of a variety of these metrics can be seen in Table 3. Most of these have been examined in patients being treated with chemotherapy, and some have been examined in patients undergoing RT (Table 3). To date, the routine acquisition of diagnostic quality MRIs during a course of RT has been prohibitively expensive and inconvenient for patients. Time on an MRI scanner is a highly limited resource at nearly all hospitals. The introduction of the MR Linac, with the capability to acquire many MRI sequences routinely during daily treatment, introduces a novel method of data collection. Rapidly, centres will start observing changes on MRI that occur in both the tumour and normal tissues during a treatment course. These changes, uniquely seen on MRI, can then be tested as early biomarkers correlated with cancer-specific outcomes. Contrast agents during treatment may also hold promise for investigation [26]. The potential for early changes in MR (e.g. diffusion-weighted imaging) to be correlated with long-term outcomes has been shown in multiple tumours during a course of treatment with radiation, summarised in Table 3. These findings need robust prospective validation, which routine use of MRI guided RT may provide. In addition to the use of current functional MR sequences, the MR Linac workflow allows for quantitative feature extraction using radiomic approaches to develop imaging-based biomarkers. This concept of daily response assessment may introduce an entirely novel dose prescription method into RT. Specifically, rather than treating to a prespecified ‘historic’ dose, a patient’s treatment dose could be determined by an imaging biomarker goal. One could envision targeting a specific apparent diffusion coefficient level (for example) as the goal for a patient’s treatment. This would represent a distinct paradigm change from the historic method of prescribing dose founded on a population basis. The future of dose prescription in RT could become a biologically adaptive, imaging biomarker–driven dose for a specific patient. Moreover, MR-based functional imaging could represent radiosensitive or radioresistant subvolumes of disease that may benefit from differential dosing strategies. A visual representation of this potential ‘threshold’ based dosing strategy is seen in Fig. 3. Consortium-based collaboration, validation and qualification will be absolutely critical to validate the clinical significance of the imaging metrics that are acquired during a course of radiation. The implications of a shift of this nature are extremely promising for patients. One of the most frequently asked patient questions during a radiation treatment course, ‘is this working?’, could be answered by the radiation oncologist with a much greater degree of confidence. Table 3 summarises the potential imaging-based response characteristics. It becomes feasible to visualise a future of RT driven by tumour-specific imaging signatures of treatment response. This could enable personalisation of RT dosing strategies or the potential for routine, functional adaptive dose-painting strategies [27].

Table 3.

Potential imaging biomarkers of treatment response to be acquired using an MR Linac.

Candidate imaging biomarkers	Candidate imaging metric	Tumours site with diagnostic significance	Example clinical series showing clinical significance of imaging changes during radiation therapy
Diffusion-weighted imaging [53]	ADC	Primary brain Rectal adenocarcinoma Head and Neck Pancreatic adenocarcinoma Cervical Prostate Liver	Rectal adenocarcinoma [54,55] High-grade glioma [56] Pancreatic adenocarcinoma [57] Cervical cancer [58] Prostate [59] Liver [60]
IVIM [61]	Perfusion fraction Dslow Dfast	Primary brain Rectal adenocarcinoma Head and neck Pancreatic adenocarcinoma Cervical Prostate Liver Oesophageal	Rectal cancer [62] Head and neck [63,64] Cervical cancer [65,66] Bone metastases [67] Oesophageal [68]
DCE-MRI [69],a	KTrans ve vp Blood flow IAUC	Primary brain Secondary brain Bone metastases Rectal adenocarcinoma Head and neck Pancreatic adenocarcinoma Cervical cancer Prostate Pancreatic adenocarcinoma	Head and neck [70] Prostate [71] Secondary brain tumours [72] High-grade glioma [73] Liver [74]
Relaxometry [75]	T1 map T2 map BOLD	Primary brain Secondary brain Pancreatic adenocarcinoma	N/A
CEST [76]	Metabolite maps	Primary brain Secondary brain	High-grade glioma [77–79]

IVIM, intravoxel incoherent motion; MRI, magnetic resonance imaging; DCE, dynamic contrast-enhanced; CEST, chemical exchange saturation transfer; ADC, apparent diffusion coefficient; BOLD, blood oxygen level dependent; IAUC, initial area under the curve.

^aLimited by the need for exogenous contrast agents.

Fig. 3.

Hypothesised novel method of radiation dosing with routine MR guidance. RT, RT; MR, magnetic resonance.

4. Conclusions

MR-guided RT is an exciting and rapidly advancing area of cancer research, accelerating with both computational and hardware advances. MR guidance is positioned to transform many aspects of RT as we currently know them. Even more novel integration of MRI into treatment delivery devices is also under development, such as MR-guided proton therapy [28,29]. Oncology research teams should prepare for innovative clinical trials involving personalisation and adaption of RT to a level that has simply not been seen thus far within the speciality of radiation oncology.