Research participants:
J.J.W. Lagendijk,
B.W. Raaymakers,
A.J.E. Raaijmakers,
E.M. Kerkhof, R.W. van der Put, Sjoerd Crijns, Mette Stam, Federica Benedosso, Jan Kok, Niels de Graaff, J. Overweg (Philips), K. Brown (Elekta), John Allen (Elekta)
Research projects:
Dutch Cancer Society: UU 2005-3406 To investigate the clinical rationale of daily on-line MRI anatomy observation in image guided radiotherapy
STW: A hybrid MRI radiotherapy system
Summary of the project: Precise, soft-tissue based, on-line position verification and treatment monitoring is a prerequisite for image guided radiotherapy (IGRT). In order to obtain on-line MRI guidance for radiotherapy we have constructed a 1.5 T MRI scanner integrated with a 6 MV radiotherapy accelerator. Basically, the design is a modified 1.5 T Philips Achieva MRI scanner with a small, single energy (6 MV) accelerator rotating around it, see figure 1 for an artist impression of the concept. For the prototype we have assembled a static system as shown in figure 2 and 3. The system is built in collaboration with Elekta, Crawley, UK and Philips Research, Hamburg, Germany and Philips Healthcare, Best, The Netherlands.
Figure 1: schematic design of the MRI accelerator combination. A cut-away view of the MRI is shown with around it the gantry with the accelerator and its peripherals. The light blue toroid indicates the low magnetic field zone outside the MRI created by adapting the active shielding in order to decouple the MRI and the accelerator.
Figure 2: Schematic of the set-up of the prototype as currently installed in Utrecht. The accelerator is in a fixed position lateral to the MRI.
Figure 3: Photograph of the 1.5T MRI accelerator in the UMC Utrecht. The accelerator is mounted on the wooden stand, by placing two copper RF cages at either side of the magnet, the accelerator and the MRI are also RF wise decoupled.
The main aim for the prototype is to give the proof of principle of simultaneous irradiation and MR imaging with millimeter resolution. To achieve this both the radiation specifications, the imaging specifications and the clinical implications of having on-line MR images for treatment guidance will be investigated. The proof of concept is given by simultaneous irradiation and MRI imaging of a piece of pork chop, as shown in figure 4. No interference between the MRI and the accelerator is found. Then, since we have a full Philips sequence library at our disposal we can also run MRI performance tests on volunteers, the system yields diagnostic quality 1.5 T MRI images, see figure 5.
Figure 4: Proof of concept of simultaneous irradiation and MRI imaging. A T2 weighted Turbo Spin Echo sequence of a piece of pork chop with and without radiation beam on. The images are identical, proving the independent behaviour of the MRI and the accelerator, see Raaymakers et al (2009) for details.
Figure 5: Brain scans using a T1 Spin Echo sequence (a), a T2 Turbo Spin Echo sequence and the MIP of a phase contrast enhanced angiogram.
With respect to the radiation specifications the beam characteristics after transmission through the MRI system will be quantified. Also the dose deposition in the presence of a magnetic field will be investigated. The photon beam is not affected by the magnetic field, however, the secondary electrons do experience the Lorentz force and therefore the dose distribution is affected. Especially around air-cavities, since there secondary electrons are forced into circular trajectories in air due to the Lorentz force. In figure 6 an example of an IMRT plan with and without the presence of a magnetic field shows that similar dose depositions are feasible. Geant4 Monte Carlo simulations are used to investigate this.
Figure 6: IMRT plan optimised for a oropharynx tumour with and without the presence of a magnetic field. The resulting dose distribution is the same for both cases, see Raaijmakers et al (2007) for details.
The magnetic field also impacts the dose response on for instance ionisation chambers. Using Geant4 Monte Carlo simulations we investigate the behaviour of various dosimeters in the presence of a magnetic field. For the Farmer NE2571 chamber it became clear that the orientation of the chamber with respect to the incident beam and the magnetic field is determining the dose response, see figure 7.
Figure 7: Impact of a magnetic field on the dose response of a NE2571 ionisation chamber. The chamber was positioned in a Bruker magnet that was put next to a clinical accelerator. The dose response was measured and simulated for different orientations and increasing magnetic field strength, see Meijsing et al (2009) for details.
With respect to the MR imaging specifications the geometric accuracy of the MR images is crucial since these images are used to guide the radiation beam. Current correction protocols to compensate for B0- and gradient- errors yield 1-2 mm accuracy. These protocols need to be improved to yield sub-mm accuracy. Also susceptibility errors need to be taken into account, as well as the coupling between the MRI coordinate system and the accelerator coordinate system.
Another issue the clinical implementation of using on-line MR images, i.e MRI based Image Guided RadioTherapy (IGRT). The first challenge is to interpret these, i.e. find the location and shape of the tumour as well as of the surrounding organs. Our current idea is to use as many data as possible from the treatment planning phase, this turns the automatic segmentation problem into an automatic tracking problem where the contours from the previous day(s) are used to generate the current contour, for instance for the cervix, see figure 8. The next challenge is to decide what to do with these data. Ultimately the system may be used for daily treatment optimisation by providing the imaging information for on-line treatment planning.
Figure 8: Contour propagation for the CTV around a cervix carcinoma. The typical variation in the contour is shown by weekly MRI with their manual delineation. Various methods, based on either rigid registration, non-rigid-registration and user-assisted registration were compared to determine which method is most suitable for generating a new contour on daily MRI, for details see Van der Put et al (2009) (accepted for publication in PMB).
The availability of real-time high quality soft-tissue contrast images from the treatment table facilitate also the exploration of new indications for radiotherapy. We are investigating body stereoactic applications for sites currently not or sparsely treated with radiotherapy but which might become feasible with real-time MRI guided Radiotherapy. An example is the study after the feasibility of hypofractionated irradiation of renal carcinoma, see figure 9.
Figure 9: Feasibility study for irradiating renal carcinoma. MRI provides high quality soft tissue contrast images, in this case with a scan time of 0.4 second per frame. When recording this, the probability distribution of the kidney can be constructed, showing motion of the order of centimetres with breathing. By using an exhale breath-hold technique while using MRI to verify the reproducibility and stability a highly conformal dose distribution can be given. This opens the door towards new radiotherapy indications.
On-line MRI may also provide treatment monitoring and treatment response assessment required for further biological optimisation. Fortunately these issues are not unique for MR IGRT, and the way is being paved by for instance cone beam guided radiotherapy and tomotherapy.