MRI in radiotherapyRAD Magazine, 39, 453, 19-20

by Dr Gary LineySenior MRI physicist

Dr Lois HollowaySenior radiation physicist

Ingham Institute for Medical Research,Liverpool and Macarthur Cancer therapy cen-

tres, Sydney; Institute of Medical Physics,University of Sydney; Centre for Medical

Radiation Physics, University of Wollongong;SWSCS, University of New South Wales,

Sydney, Australia

IntroductionOver the preceding 20 years or so there has beenan ever-growing interest in utilising MRI in theradiation treatment of cancer. This short articleis a personal commentary into the backgroundand challenges faced with its implementation intoclinical practice and also a view of the currentstate of play and future direction in this field.

As the delivery of radiotherapy has becomeever more conformal and targeted, the concept of3D (and 4D) image guidance has grown in impor-tance. When considering current imaging modal-ities, CT is the primary imaging modality. Incombination with CT, PET and/or MRI may alsobe utilised. CT is the long established gold stan-dard to aid 3D planning as it offers excellent geo-metric integrity and a direct relationship betweenHounsfield units (CT number) and electron den-sity, a critical parameter used in the dose calcu-lations. However, as well as requiring the use ofionising radiation, the modality generates quitelimited soft tissue contrast, making the distinc-tion between tumour and normal tissue sub-opti-mal. As delivery precision improves, the abilityto increase the therapeutic ratio through reducedtreatment margins and potential dose escalationwill rely on confident tissue delineation from highquality images. PET has been utilised to estab-lish the extent of disease involvement throughfunctional imaging but has poor spatial resolu-tion, and will not be discussed further here. MRIprovides high quality soft tissue imaging as wellas high spatial resolution.The pros and consMRI is a good candidate for treatment simulation as it pro-vides excellent soft tissue contrast that can be varied con-siderably, enabling particular tissue characteristics to behighlighted. A typical MRI examination will consist of sev-eral sequences, each acquired in various favourable planesand with many different contrast weightings. Furthermore,MRI is able to provide functional imaging which may bebeneficial in more advanced treatment regimens.1 Theseinclude, but are not limited to MR spectroscopy, which canprovide metabolic mapping of tumour profiles through detec-tion of chemically distinct shifts in resonant frequency, anddiffusion weighting which can track restricted motion in

tumours at the cellular level. Both contrast and non-con-trast enhanced techniques can be used to probe tumour per-fusion. Still other quantitative parameters (for example R2*)have been shown to correlate with oxygenation levels andthereby reveal hypoxia and radio-resistant sub volumes oftumour. Each of these techniques has the potential to com-plement the more routine anatomical images and providethe ultimate description of the intended treatment volume.Figure 1 shows this concept in a brain tumour – the limi-tation of CT here is obvious compared to the extra informa-tion provided by the MRI examination that providesopportunities to ‘target, boost and avoid’.

For all its advantages, MRI has two distinct drawbacks,the first being the inherent geometric distortion in theimages. Many of the earlier researchers were able to showalarming geometric inaccuracies on open low field systemsthat have long stayed in the minds of the radiotherapy com-munity. MRI is prone to both system and patient relatedeffects, the former due to non-uniformities in the main fieldand non-linearities in the imaging gradients, becoming worseas a function of distance from the isocentre of the magnet.The latter type of effect is introduced via the patient interms of fat-water shifts and susceptibility effects. Most (butnot all) of these effects can be mitigated by careful sequenceand parameter selection and/or correction schemes providedby the vendor. Nevertheless, it is important to characteriseany MRI system being used.2

The second drawback stems from the lack of inherentelectron density information. At present the current prac-tice is still to fuse to CT for this purpose. However, manygroups are looking into ‘MR only’ planning; segmentationand bulk density assignment is a popular approach3 but oth-ers are looking into providing this indirectly from mappingCT atlas data4 and directly from an MR signal surrogate.5

Related to this is the desire to generate so-called pseudoDRRs from MRI directly so that patients may be set up onlinac beds with only an MR planning scan.Current practiceMany studies have shown reduced user variability throughimproved delineation.6 MRI is now routinely used in manysites to assist in the contouring of PTV and OARs. Brain isusually the easiest site to begin with owing to its small cen-tral rigid volume that is easy to register without the needfor extra localisation equipment. This site also lends itself tothe full myriad of functional techniques (figure 1). Theother sites of interest pose their own unique requirementsbut, commonly, the main issue is the compromise betweenmaintaining a treatment position and the use of a suitableRF coil. Head and neck imaging should always be performedin the treatment position owing to the wide variation inneck flexation between planning CT and a diagnostic MRI.

Breast imaging, performed diagnostically well in theprone position with dedicated RF coils, may also be acquiredin the treated (supine) position with good success.7 It is asyet unclear which position will win the day, as both areviable options.

FIGURE 1CT (left) and various MRI images of a brain tumour.Here, a T2-weighted image has been acquiredtogether with an MRS spectroscopy scan (dark voxelsindicate suspicious metabolic profile) and fMRI scanhighlighting the motor cortex.

A clear indication for the inclusion of MRI is in prostateplanning of hip prostheses. Severe high Z-artefacts observedin CT are replaced by more localised susceptibility effectsaround the implant in MRI. Prostate planning may be donewith fields of view big enough to encompass registrationlandmarks but not so big as to reveal distortions. More prob-lematic is the potential for movement of the gland betweenCT and MRI, despite consistent bowl and bladder prepara-tion. Gold seeds that show up on both MR and CT may bea useful way forward here. Continuing advances in bothimaging speed and motion compensation schemes meansmore challenging sites such as lung may now be considereda realistic goal for MR planning. Often the poor relation inradiation treatment, brachytherapy has seen many groupsactively pursuing MRI guided techniques.8 MR compatibleapplicators can be made more visible using catheters orchanging the imaging sequences used. Often the planningsystems do not correct for tissue heterogeneity, meaning it isa particularly attractive site to develop quickly into ‘MR-only’.HardwareManufacturers of MRI scanners are all actively pursuingradiotherapy specific products and road-maps. A majoradvance saw the introduction of a 70cm wide bore system,creating the room needed for immobilisation equipment to beused in situ. All the major vendors now provide flat tabletopcouches, often with an integrated RF coil underneath, tosimulate treatment position and facilitate the fusion withCT planning. Other RF coils that fit around immobilisationmasks notably at head and neck sites are also commerciallyavailable (figure 2). Each vendor now offers this equipmentfor both 1.5T and 3.0T magnets. For many centres, thischoice may ultimately be driven by cost, although the caseat higher field strength for an MR-simulation systemremains unproven and as yet there is no clear evidence as tothe best field strength. Lasers as used for radiotherapytreatment positioning are now also being installed with MRIscanners which are used for radiotherapy purposes.

The full potential of MRI is yet to be exploited in radio-therapy. Some groups have begun to map patterns of failurewith MRI data retrospectively. However, more clinical trialsare required to establish how these multi-functional datasetscan be put to best use prospectively. Until dedicated MRsystems are installed as the ubiquitous planning scanners,many centres will be limited to the current practice of beg-ging and borrowing limited time on radiology machines thatare already full to capacity.FutureAlthough this article has only considered planning aspects,the full role of MRI is actually to complement every part ofthe radiotherapy chain, from planning through to verifica-tion and follow-up. MR gel dosimetry is a technique that

utilises radiosensitive materials that can be irradiated andthen ‘read out’ by virtue of the changes in magnetic proper-ties (figure 3). By careful calibration, a true 3D dosimetricanalysis may be performed with good accuracy.9 The lack ofionising radiation means MRI is the ideal modality torepeatedly image the patient during the course of treatmentand also for longer term follow-up.10 The combination of theanatomical and physiological imaging described earlier isideal to verify the success or otherwise of treatments andpotentially adapt fractions accordingly. Perhaps the ultimatecombination of MRI and radiotherapy is the design of hybridMR-linac systems, of which there are now three at variousprototyping stages throughout the world. Here in Sydneythe research bunker has been completed ready for the instal-lation of the southern hemisphere’s only version (figure 4).A fourth system combining Cobalt sources (instead of a lin-ear accelerator) with MRI guidance is already commerciallyavailable. With these systems comes the promise of real-time image guidance and adaption of tailor-made treatmentplans on the fly. These exciting developments are sure topush both radiotherapy and MRI fields forward over thecoming years.

FIGURE 2Set-up for a dedicated head and neck MRI planningscan. A flat tabletop is used together with two lateralRF surface coils placed around the immobilisationmask.

FIGURE 3Screenshot from in-house software analysing a treat-ment plan (top left) against MRI dosimetry data (topright) with the two volumes overlayed (bottom).

FIGURE 4Gary Liney outside the MR-linac research bunker atLiverpool Hospital in Sydney, Australia.

AcknowledgementsThanks to staff at the Centre for MR Investigations at HullRoyal Infirmary and Radiation Physics, Castle Hill Hospital,Hull, and Mohammad Al’Sad at the Department ofComputing, Manchester University (mohammad.alsad@manchester.ac.uk) for the screenshot shown in figure 3.

References1, Groenendaal G, M R Moman, J G Korporaal, P J van Diest, M van

Vulpen, M E P Philippens, U A van der Heide. Validation of functional imag-ing with pathology for tumor delineation in the prostate. Radiotherapy andOncology 2010;94:145-150.

2, J N H Brunt. Computed tomography-magnetic resonance image regis-tration in radiotherapy treatment planning. Clinical Oncology 201;22:688-697.

3, Lambert J, Greer P B, Menk F et al. MRI-guided prostate radiationtherapy planning: Investigation of dosimetric accuracy of MRI-based doseplanning. Radiotherapy and Oncology 2011;98:330-334.

4, Greer et al. An atlas-based electron density mapping method for mag-netic resonance imaging (MRI)-alone treatment planning and adaptive MRI-based prostate radiation therapy. International Journal of Radiation OncologyBiology Physics 2012.

5, Jonsson J H et al. Treatment planning using MRI data: an analysis ofthe dose calculation accuracy for different treatment regions. RadiationOncology 2010;5:62.

6, Rasch C, Steenbakkers R, van Herk M: Target definition in prostate,head, and neck. Semin Radiat Oncol 2005;15:136-145.

7, Giezen M E, Kouwenhoven A N, Scholten E G et al. MRI versus CT-based volume delineation of lumpectomy cavity in supine position in breast-conserving therapy: An exploratory study. International Journal of RadiationOncology Biology Physics 2012;82:1332-1340.

8, Haie-Meder C et al. Recommendations from gynaecological (GYN) GEC-ESTRO working group (I): concepts and terms in 3D image based 3D treat-ment planning in cervix cancer brachytherapy with emphasis on MRIassessment of GTV and CTV. Radiotherapy and Oncology 2005;74(3):235-245.

9, Baldock C, De Deene Y, Doran S et al. Polymer gel dosimetry [Review].Phys Med Biol 55(5), R1-R63 (2010).

10, van de Bunt L, Jurgenliemk-Schulz IM, de Kort GA et al. Motion anddeformation of the target volumes during IMRT for cervical cancer: what mar-gins do we need? Radiother Oncol 2008;88:233-240