In this clinical update, we describe the development and the expected clinical impact of the UMC Utrecht MRI linac (MRL) system. This combination of a radiotherapy accelerator with a diagnostic quality 1.5 T MRI scanner provides real-time imaging at the moment of treatment, thus providing unprecedented targeting accuracy. Typically, a series of radiotherapy fractions lasts several weeks. A treatment plan, which defines the machine settings and dose distribution, must be made for this series. Such a treatment plan is based on an anatomical ‘snapshot’ at the beginning of the treatment and is used for the whole treatment series, disregarding the daily uncertainties in patient and tumour position, tumour movements (e.g. breathing and peristalsis), tumour regression, etc. To guarantee the dose coverage of the tumour, large radiation fields are chosen. It is clear that large tumour positioning uncertainties result in large volumes and thus dose-limiting normal tissue involvement. To minimize this normal tissue involvement, considerable effort has been devoted to the development of daily position verification techniques. In Utrecht, we developed the use of small gold fiducial markers which can be implanted inside the prostate to act as a surrogate for the prostate position 1. The introduction of cone beam CT at the linear accelerator was a major breakthrough 2 (Fig. 1). For bony structure-related tumours and those with large CT contrast, for example lung tumours, better position verification became possible. In this process, the radiation fields became smaller, greatly reducing the normal tissue involvement. Originally, the normal tissue volume treated with a high dose was about 10–100 times the tumour volume. With the development of better position verification techniques, this ratio reduced to about five to ten times more normal tissue. Radiotherapy started to change from a fractionated treatment towards a more local treatment of the tumour process (Fig. 2). Great progress was made in the treatment of prostate cancer (gold fiducials) and of peripheral lung tumours (cone beam CT contrast) 3. In this process, the research focus moved from fractionation regimens to create therapeutic gains between tumour and normal tissue, towards imaging, intensity modulation and position verification. However, it was also clear that for most tumour locations, including pancreas, kidney, oesophagus, liver, colon and rectum, the cone beam CT imaging was of such a poor quality that this radiotherapy transition was unsuccessful (Fig. 1). Considering the importance of obtaining higher tumour doses, whilst reducing the dose to normal tissues, it is clear that there may be an important role for MRI in online treatment guidance. MRI provides superb soft tissue contrast, and possibly in real-time at the moment of treatment. The idea of combining a radiotherapy accelerator with a diagnostic quality MRI system for online guidance of radiotherapy delivery was first presented by Lagendijk and Bakker 4 at the 2000 ESTRO conference in Istanbul. To produce a truly integrated system, the radiotherapy accelerator must rotate around the MRI system with the patient inside (Fig. 3a). It was clear from the beginning that to obtain the best image quality, we had to use a diagnostic quality 1.5 T MRI. Having a system which is compatible with the diagnostics systems also provides access to the MRI knowledge platform at radiology. The basic technical problem that we encountered was the magnetic interaction between the two systems: the magnetic field of the MRI distorts the accelerator, whilst the metal of the accelerator distorts the homogeneity of the magnetic field inside the MRI. The original idea was to passively shield the accelerator tube, making the load to the MRI homogeneous by circumferential shimming of the accelerator gantry. The design implied shooting through the cryostat, making a homogeneous window and putting aside the superconducting coils. The major breakthrough was the adaption of the concept of active shielding. The modification of the active shielding of the magnet produced a zero magnetic field in a toroid around the magnet 5. Thus, we were able to move away from the concept of passive shielding of the accelerator, making the accelerator and MRI design more manageable. The design now consisted of a 1.5 T magnet surrounded by a 6 MV accelerator, with active shielding producing a zero magnetic field zone in a toroid around the magnet (Fig. 3a). The sensitive accelerator components are positioned in this area. The beam shoots through a homogeneous part of the cryostat, with the superconducting wires removed from this area. In addition, the gradient coil is split (Fig. 3a). With an MRI accelerator, the patient is irradiated whilst inside a homogeneous 1.5 T magnetic field. Using a Bruker electromagnet (field strength up to 1.2 T) and a conventional accelerator, the effect of the magnetic field on the dose distribution was investigated. The most important aspect is the electron return effect (ERE). This effect, where electrons return to the tissue at the beam exit point and produce a hot spot in the tissue, is a basic principle of physics (Lorentz force on moving electrons). This occurrence of hot spots inside tissue around air cavities has to be countered, especially in lung and head and neck cancers 6. Opposing beams proved to cancel the effect in first order. The ERE exists at every magnetic field strength and implies that the Monte Carlo code is needed for dose calculation and care must be taken in the design of treatments in which tissue air boundaries are present. In March 2007, we started to prepare one of our treatment rooms for the MRL prototype and had to bring the large magnet into a treatment room which has 2 m thick concrete walls. Finally, we sawed through the concrete walls to widen the entrance. In February 2009, the installation of the prototype was initiated, using a dedicated Magnex 1.5 T magnet and an Elekta compact accelerator (Elekta Inc, Crawley, UK). On 4 March 2009, the simple prototype, a stationary accelerator on a wooden table shooting mid-plane into the MRI system, worked (Fig. 3b). Images of pork chop were produced (Fig. 3d) to demonstrate that the radiation had no impact on image quality (Fig. 3c). Identical images with the beam on and off showed that the system indeed worked independently and that we were able to create images of high diagnostic quality with the radiation beam on 7, 8. This success resulted in the breakthrough decision by Elekta to develop the next-generation prototype and to build a full gantry-based MRI linear accelerator. Only close collaboration between the radiology and radiotherapy departments would ensure that the MRL project was successful. At the UMC Utrecht, we established a new institute: the Centre for Image Sciences (CIS) (www.umcutrecht.nl/cis). This new institute merged all research on MRI-guided oncological interventions, including MRI-guided external beam radiotherapy, MRI-guided brachytherapy, high field (7 T) tumour characterization, MRI-guided HIFU and MRI-guided Holmium embolization techniques. We decided to focus, with the MRI linac, not only on regular radiotherapy applications but also on those complex tumour sites, such as the kidney, liver, pancreas, rectum, oesophagus and colon, where present radiotherapy techniques fail due to problems with tumour visualization and tumour movement. The motto of the CIS is ‘surgery without a knife’, indicating our expectation if the MRL system becomes the success we anticipate. At present, over 140 PhD students are working at the CIS, 80% on projects related in some way to MRI, and more than 40 on MRL-related projects. Within CIS radiotherapy, we have the availability of three treatment rooms for MRL, two dedicated Philips radiotherapy MRI simulators (wide bore 1.5 and 3.0 T; Ingenia, Philips, Amsterdam, The Netherlands), a dedicated 1.5 T Achieva MRI scanner (Philips, Amsterdam, The Netherlands) for HIFU and a dedicated 1.5 T Philips Ingenia system (Philips, Amsterdam, The Netherlands) for HDR brachytherapy, making radiotherapy truly MRI guided. Whilst we were completing the setting up of our new CIS building, Elekta was devoting huge effort to the commercial design of the system. The installation of the first clinical grade prototype in June 2014 began with the lowering of the gantry parts through the roof of our shelter (Fig. 4), followed by the magnet and the accelerator. On 15 October 2014, we were able to produce the first images with the clinical grade prototype with the beam on, showing Ingenia 1.5 T MRI quality with the beam on. At present, we are preparing the installation of the clinical version of the MRL system. Great effort is being made to prepare the clinical commissioning of the system and design the clinical procedures. The Phase I safety and feasibility study will be conducted with a patient with bone metastases. Bone metastases are stable, easy to visualize on MRI and also easily visible on the integrated megavoltage imager. As such the MRI-based targeting accuracy can be verified with the megavoltage imager and the safety can be assessed. In conventional radiotherapy, a treatment plan is produced at the beginning of the treatment series, repositioning patients daily. From the beginning, it was decided that this new system required online treatment planning (Fig. 5). There is no rationale to correct for translations only, using a moving table top, whilst the online MRI provides information about translations, rotations, deformations, tumour regression, movements, etc. We had to design a new planning system which would be able to generate the required dose distribution and its related MLC sequencing in seconds, acting on the dynamic online anatomy. In 2013, huge progress was made in online treatment planning 9. The clinical MRL system must be able to provide submillimetre stereotactic accuracy for every location in the body, whether stationary or moving. The online and real-time MRI brings certainty to the treatment process and as such enables dose painting based on knowledge of tumour presence and tumour characteristics. The MRL system may greatly contribute to improve the quality of regular radiotherapy applications such as for prostate, head and neck and lung cancer treatments. However, a breakthrough is expected for those tumours which are not well visualized by cone beam CT such as tumours of the oesophagus, pancreas, kidneys, rectum and lymph nodes, as well as metastases. Patients with these tumours may benefit from better dose painting, increasing the tumour dose whilst reducing the normal tissue involvement. It is clear that the best diagnostic quality MRI capable of showing the required tissue information is needed to control treatment. In particular, normal tissue structures can be visualized well and can thus be avoided in the radiation fields. From the beginning of the project, it was clear that the MRL system may reshape the field of radiotherapy. If ‘seeing what you treat’ is possible, large treatment fields with major normal tissue involvement would no longer be acceptable. It was also clear that the radiotherapy field is not prepared; MRI knowledge is minimal and online procedures must be developed. The introduction of such a novel technology needs an international knowledge platform. We decided together with Elekta to create an international consortium of leading radiotherapy departments. We are extremely proud that the consortium now consists of the MD Anderson Cancer Center, Royal Marsden-ICR, The Christie, NKI-AvL, Sunnybrook and MCW. We will hope to see in the coming years that collaborative clinical studies will guide the introduction of this new system. It may be an interesting exercise to compare the capabilities of the MRL system with those of the present proton therapy systems 10. In most proton therapy applications, due to the stopping of the beam, Bragg peak, the lower integral normal tissue dose is the benefit aimed for. This lowering of the integral dose, however, can also be obtained with better targeting and thus smaller radiation fields. Volumes grow with the third power of the radius. Dirk Verellen 11 showed with his well-known comparison of an orange and the volume of its shell that most of the volume is in the shell. Thus, better targeting implies reduced volume, reduced integral dose and reduced high-dose volume. The proton field has to react on the development of the MRL system. MRI guidance must be developed for proton-based radiotherapy. MRI-guided proton (MRP) therapy is not a simple procedure, whereas the MRL system relies only on targeting, the MRP systems also have to define the exact depth in tissue in real-time to exploit the Bragg peak. If this depth is not measured accurately enough, the MRL may outperform MRP for most high precision applications. As with brachytherapy, proton therapy must become MRI guided or it will probably not play a significant role in modern radiotherapy. Online MRI guidance may start a paradigm shift in radiotherapy: the central position becomes MRI, not the knowledge of fractionated radiation and radiobiology. Soft tissue MRI for guidance will be extended in a later phase with functional information about the tumour obtained with advanced imaging with MRI, PET and SPECT. The use of endogenous contrasts will assist tumour characterization, better delineation and treatment response assessment 12. As a consequence of MRI therapy guidance, radiotherapy becomes more of an interventional radiology process. Close collaboration is needed between the radiation oncologist, radiologist, pathologist, medical physicist and surgeon. Such a multidisciplinary team will guide the care of cancer patients with local disease. Huge steps have already been taken. Remaining issues to consider are the development of real 4D treatment planning, 4D MRI technology, MRI training of the radiotherapy community and the definition of clinical procedures for practically executing these treatments and training radiation technologists, radiation oncologists and radiologists. The UMC Utrecht has a co-development agreement with Elekta Inc. and Philips. Several research projects are sponsored by Elekta and/or Philips. The authors do not have any affiliation with these industry partners.
No takes yet. Share an insight, caveat, or question.
Lagendijk et al. (2016) studied this question.