Gantry Classification and Design for Hadron Therapy

GANTRIES FOR HADRON THERAPYMay 2006P.J. Bryant – CERN, Geneva

What is a gantry? A more detailed classification… Passive spreading gantries. Divergent-beam voxel-scanning gantries. Parallel-beam voxel- scanning gantries. Exo-centric gantries. Novel designs. Geometry is not enough… Gantry matching requirements. Matching methods… Symmetric beam. Round beam. Rotator method. Equal sigmas method. Patents. GSI therapy complex. PIMMS. Gantry construction. Superconductivity. Some controversy. Riesenrad gantry film. Conclusions. Contents

If you have a therapy centre, then you will want a gantry or maybe two!

What is a gantry? • A gantry directs the beam onto the patient at whatever angle is required by the treatment plan. Ideally, the full 2p should be available about the gantry axis. • Ideally one should also be able to rotate the patient, so as to access the full 4p solid angle. • The requested treatment field is 40  40 cm2. • The requested beam penetration 27 cm. • There are two broad classifications : • Iso-centric gantries. • Exo-centric gantries. Iso-centric Exo-centric

A more detailed classification… • Beam diverges after last dipole: • Passive spreading gantries (protons). • Divergent-beam voxel-scanning gantries (protons). • Beam diverges within gantry lattice: • Parallel-beam voxel-scanning gantries (light-ions). • Exo-centric “Riesenrad” gantry (light ions). • There is also an extended category of novel designs.

Passive spreading gantries Loma Linda Corkscrew • The first passive spreading gantry built was the “corkscrew” gantry at Loma Linda. • Today the accepted conventional design is the “conical” gantry as demonstrated by IBA. To understand these gantries it is useful to look at the passive spreading technique. IBA Conical

Double-scatterer system for protons • First scatterer significantly increases angular divergence. • Second scatterer is shaped to scatter the dense centre to the edges while letting the edges pass largely unaffected. • ~60% of the beam will belost. • Scatterers will be a high Z material to favour scattering (copper). Double scatterer Collimator Quasi-uniform beam (within 2%) over 20  20 cm2

Proton beam preparation before the scatterers • Adjust the beam energy (Bragg-peak) to the maximum tumour depth. • Stepwise energy modulation to define the slices in the tumour. • Fast modulation by a rotating propeller to create SOBP. • Static modulation by a ‘ridge’ filter may be used to replace propeller. • Low-Z materials preferred for less scattering (plexiglass).

Passive spreading gantries continued • The Twiss functions and beam emittances at the entry to the gantry are not critical because the beam is strongly scattered after the last dipole largely destroying the memory of the initial beam. • Similarly, the alignment of the incoming beam is not critical because the beam will be spread out and collimated to the correct shape and position just before the patient. • The magnets in the gantry can have normal apertures (i.e. weight, cost) because the beam is spread out after the last dipole. • This type of beam delivery and gantry is only used for protons. Light ions would fragment in the scatterers and the beam would be heavily polluted by neutrons.

Divergent-beam voxel-scanning gantries PSI Eccentric • PSI have a working voxel scanning system for protons from a cyclotron. • IBA are promising a replacement nozzle for their passive-spreading proton gantries that will perform voxel scanning. (The ‘Nozzle’ contains the spreading or scanning system and the collimator). IBA Conical For voxel scanning, the tumuor is divided into layers and each layer is divided into pixels. Each pixel has a width, height and thickness of a few mm, i.e. it is a volume pixel or voxel. The beam energy and energy spread are set to match the layer depth and thickness and the beam size is set to the voxel size.

Parallel-beam voxel-scanning gantries • Parallel-beam scanning reduces the surface dose given to a patient by divergent-beam scanning. • Parallel-beam scanning gantries solve the problem of lack of space in divergent-beam gantries for the scanning magnets for ion beams that have a higher magnetic rigidity. • Gantries specially designed to give parallel scanning are the “cylindrical” gantries. • The disadvantage of the cylindrical gantry is the large aperture needed in the final dipole, which increases size, weight and power consumption. GSI Cylindrical

Voxel scanning • Highest precision,but the small beam size makes tumour movement a serious limitation. • Well suited to light ions that scatter less and therefore preserve the small beam sizes. • Synchrotrons offer the best flexibility. • From full-volume passive spreading through to voxel scanning, there has been a reduction in the elementary volumes that are irradiated and a corresponding increase of about 3 orders of magnitude in the speed required from the on-line dosimetry system to maintain the treatment time and accuracy. This makes voxel scanning the highest technology variant.

Wobbling of an enlarged beam spot • Consider the above scheme, especially for ions. • Full passive spreading of ion beams is not recommended as the beam fragments and the impurities have different penetrations and RBE values. • In the above, the scatterer/absorber produces an enlarged beam transversely with a momentum spread, typically 2 cm  2 cm  1 cm (spread-penetration). • The enlarged spot (‘blob’) is rapidly ‘wobbled’ in a circular motion across the collimator to give a uniform irradiation field.

Some comparisons • To avoid “hot” and “cold” spots in the treatment field, voxel scanning requires sub-millimetre precision for the positioning and the size of the beam spot. Similarly, it imposes the same precision on the immobilisation of the tumour. • Passive spreading avoids “hot” and “cold” spots within the treatment area. The collimator before the patient “screens” the effect of upstream movements of gantry elements and the scatterer “screens” the patient from changes in the upstream beam parameters from for instance gantry angle changes. • Movements of the tumour appears as a blurred edge to the treatment volume. • Thus voxel-scanning gantries must be more rigid for all gantry angles and the beam matching for changing gantry angles is more critical. • Divergent scanning gantries are limited to protons by the underlying technological limits on the scanning elements and the gantry size.

Some more comparisons • Cylindrical gantries for ion voxel scanning are of the order of 700 t making the specific load on the rollers consequential. • All the iso-centric gantries roll on large support rings with diameters up to 12 m. If a ring is damaged, which occurs, it is virtually impossible to replace the ring without completely re-building the gantry. • Due to the constraints of weight and size, all the iso-centric gantries have a limited patient space at the iso-centre. • In comparison,exo-centric gantries are lighter, less power consuming, the support rings are smaller and can be replaced by commercially available roller bearings for turrets and the patient space is effectively unlimited. (IBA have solved the problem of minor damage to the rings, by providing a second precision surface on the inside of the ring. This surface is used to guide a mobile grinder that clamps on the ring using the inner surface as reference.)

Exo-centric gantries • This category is the “engineer’s” solution that places the heavy equipment to be rotated (magnets, counter-balance weight) on the axis and the light equipment (patient, couch, robot arm) off axis, BUT the medical community does not like this solution. • Perhaps the first publication is by R.L. Martin.

“Riesenrad” exo-centric gantry • The “Riesenrad” is named after the famous “wheel” in Vienna. • The heavy dipole magnet (~70 t) is kept on axis where it can be more easily balanced and supported. • The patient is in a spacious room which must provide a firm support, but need not be positioned to high precision. • The patient’s couch is then aligned with respect to the dipole by a robot arm and a photogrammetric system. • The key to this gantry is how to match the beam.

Novel gantries • The criteria applied by the early gantry designers are not always clear. Reducing the the axial length seems to be the most frequent aim and surprisingly they seemed insensitive to weight. • The block of 4 drawings below are taken from the EULIMA study (~1991). • In general, no optical principles were published for matching the exotic gantries apart from the dipole layout.

Novel gantries continued • The “Planar Gantry” is proposed by M.M. Kats ITEP, Moscow. • The moving structure is eliminated, but the space around the patient is still limited and the total bending is about twice that in the accelerator. • Two fixed beam lines (one horizontal and one at 60 deg.) is simpler and with rotation and limited tilting of a supine or sitting patient nearly all requirements can be reasonable met.

Novel gantries continued There are some other “mobile-magnet” geometries that have been patented”, e.g. by Prof. G. Kraft...

Novel gantries continued Another idea is to reduce the overall gantry diameter and weight by inclining the final beam at 60 degree (the “alternative gantry” proposed by Marius Pavlovic and patented by GSI). Note: If two fixed lines are used instead of a gantry then the combination of horizontal + 60 deg. gives access to more solid angle than horizontal + vertical because rotation about the vertical beam brings no gain. If tilting the patient is allowed, then the access becomes very good.

Novel gantries continued • The “S.C. pipe gantry” that guides and focuses a beam like a hose pipe guides water, suggested by G. Benincasa. • The dream of having a flexible beam guide may have started with S. Van der Meer, “The Beam Guide”, CERN 62-16, 1962. Van der Meer studied a coaxial system with the outer conductor at infinity and mentioned the use of a superconductor. • A PhD student, A. Maier, tried to improve the focusing by shaping the X-section of the conductor, adding the return conductor and stepping the conductors to get focusing and defocusing regions, but a practical scheme could not be found. • In fact, the interest in making charged-particle pipes is quite widespread. The next step is to add some iron, which leads to the “Pipetron” an extruded single or double channel magnet with one S.C. cable.

Geometry is not enough... • Up to this point we have concentrated on the dipole geometry of gantries without explaining how the dispersion and focusing could be made to work. In fact, many early gantry proposals ignored this point completely. • To match a gantry ones needs to take care of several aspects: • The rotational optics (see next slides). • In some cases, it is necessary to consider the transverse beam distributions, i.e. to distinguish between beams from a resonant slow extraction in a synchrotron and a beam from say a cyclotron. • For voxel scanning, it is necessary to design the optics of the line and the gantry as an integrated whole e.g. beam size control may be in the line or in the gantry according to the method applied.

Gantry matching requirements • The shape and size of the beam spot at the patient must be totally independent of the gantry angle. • There must be no correlation between momentum and position across the beam spot. • For the purposes of scanning, the optics inside the gantry must be independent of gantry angle.

Matching methods… • Symmetric beam method with zero dispersion (exact) • The beam must have zero dispersion and be rotationally symmetric i.e. the same distribution (gaussian or KV) with equal Twiss functions and equal emittances in both planes at the entry to the gantry. • The gantry must be designed with a closed dispersion bump in the plane of bending. • Round-beam method with zero dispersion (partial) • The beam must have zero dispersion, the same distribution (gaussian or KV) in both planes with the condition Exbx=Ezbz at the entry to the gantry. It would also be desirable but not absolutely necessary to have ax=az=0. • The gantry must be designed with phase advances of multiples of p in both planes (i.e. 1:1 or 1:n matrices) and a closed dispersion bump in the plane of bending. • The problem in this case is that the optics inside the gantry changes with rotation angle. However, it is often possible to “freeze” the last section so that the scanning is unaffected.

Matching methods continued • Rotator method (exact and completely general) • This method will rigorously map all Twiss functions and the dispersion functions into the gantry coordinate system independent of the gantry angle. • The gantry must be designed to give zero dispersion at the exit, but note that it can be finite at the entry. This will be illustrated later with the “Riesenrad”. • This method is essential for slow extracted beams that have extremely unequal emittances and control their beam sizes in unconventional ways. • The rotator appears in a Loma Linda patent, but is not explained!

Rotator method • Maps the beam 1:1 to the gantry independent of the angle. • This is the rotator solution and it maps the dispersion function and the Twiss functions rigorously to the gantry coordinate system.

Appearance of rotator in a Loma Linda patent • The “rotator” quadrupoles are clearly labelled, but the patent does not explain the theory or function of the device (omission of a non-trivial step should invalidate the patent). It is not sure that the patent writer realised that the “rotator” rotates . • The “rotator” was invented by Lee Teng of Fermilab, but he did not publish the design. The detailed derivation appears in an internal Loma Linda report and his private laboratory notebook. • The Rotator design shown is clearly a FODO design which is not ideal as it exhibits large beam changes inside the lattice during rotation.

Matching methods continued • Equal sigma method (partial) • This method is used in the GSI ion gantry. • The method has been patented by GSI (EP 1 041 579 A1). • The validity of this method has not been fully demonstrated. • The problem can be understood intuitively by first considering the beam spot size in terms of sigma and then in terms of the FWHH, for example. Although the sigma values may be equal, the FWHH values will be different in the two planes because of the different beam distributions. Thus, the beam spot will be distorted and its orientation will depend on the particular optics. • This distortion may, or may not, be acceptable after smoothing by scattering in the patient’s body.

Equal sigma method • The beam is represented by its sigma matrix: • The sigma matrix translates as: • Diagonal terms gives beam widths, e.g.:

Equal sigma method continued… • The angular dependence in the beam widths can be removed. For example, if the gantry matrix is arranged to give r1,1=0 and the incoming beam is adjusted to give s(0)2,2=s(0)4,4, then • The vertical plane and the correlation terms can be similarly treated. • Despite it not being possible to patent equations, this method has been patented, See European Patent Application EP 1 041 579 A1. If one reads carefully, “round” is defined as equal sigmas, which does not necessarily mean rotationally symmetric.

Patents • You are probably familiar with the web site: http://ep.espacenet.com • But maybe you are not familiar with: http://ofi.epoline.org/view/GetDossier • Try entering for GSI’s therapy system : EP 0 986 070 A1 EP 0 986 071 A2

GSI therapy complex • GSI have recently revamped the patent for their therapy complex. The “gist” is quoted below: “The new set of claims 1 to 22 concentrates on the inventive idea to use a special designed delivery system for high energy particles from an accelerator system to the isocenter of a treatment field. The claimed invention is based on the split of the high energy beam transport system into a high energy beam transport line and gantry. The high energy beam transport line delivers a beam to a coupling point of the gantry with special beam properties, so that the gantry is able to guide the beam for all angles to the treatment field having a circular beam without the need of adjusting the high energy beam line.” • The above is based on the “Equal sigma” method, but the patent writerdoes not understand what “round” really means. • Note. This is the opposite of the PIMMS method (next slide) that leaves the gantry unchanged and puts the adaptive elements in the line. GSI gantry matched by the equal-sigmas method

PIMMS (Proton Ion Medical Machine Study) • PIMMS, CERN 2000-006, Volume II is a design for a synchrotron-based, light-ion therapy centre with various gantries. • The transfer lines and gantries were designed as an integrated system. The lines are assembled from basic modules with 1:1 or 1:n transfer matrices. The phase shifter and betatron stepper are placed upstream and can act for all gantries. The rotator can usually be placed upstream, but for the “Riesenrad” it is needed next to the gantry.

Horizontal beam size control • The extracted segment (in blue) is called the ‘bar of charge’. • The bar of charge is about 10 mm long and has a very small angular spread. • Horizontally the distribution is quasi rectangular and vertically it is gaussian. • Fitting an ellipse to this narrow bar is impractical. • Instead the bar is regarded as a diameter of a larger unfilled ellipse. • The bar turns at the rate of the phase advance. • A phase shifter can be used to turn the bar and change the projected beam size.

Vertical beam size control Classic beam size control The stepper is a 1:n telescope module that keeps all beam parameters constant except the vertical betatron amplitude function. • The vertical beam distribution is the usual gaussian and the beam size is controlled in the classic way by varying the betatron amplitude function. In the PIMMS design, the phase shifter and beta stepper are combined in the lattice module illustrated. This module can keep all parameters constant while varying mx and bz in any desired combination.

Scanning with the spot shape from a slow extraction • The spot in PIMMS is NOT round. Horizontally the distribution is quasi-rectangular and vertically it is gaussian. • It is important that the spot always has the same orientation in the gantry frame, so that the spot moves with the sharp-edged distribution in the direction of motion and the gaussian tails overlap from row to row. • The unevenness in the spill in time can be smoothed (within limits) by a feed-forward that acts on the scan velocity.

“Riesenrad” gantry matching • Since the “Riesenrad” has only one dipole it excites dispersion, but cannot close the dispersion bump to zero within the gantry itself. • In the PIMMS design, the bend from the main extraction line excites the dispersion function and the gantry is used to close the bump, thus delivering zero dispersion to the patient. This is possible because the rotator turns the dispersion function to match the gantry’s coordinate system.

Gantry construction – “Riesenrad” The central cage supports the three scanning magnets (1.5 t) and the large 90° dipole (62 t). The total weight is ~127 t, of which 23 t are due to the counterweight. The design of the central cage is driven by the desire to minimise sagging of the dipole no matter what gantry position is considered.

Gantry construction continued • Top graph shows the mechanical movement of the exo-centre due to elastic deformations of the gantry while rotating –p to +p. • Lower graph shows the shift of the exo-centre in the transverse plane due to the optical errors caused by the movements of the magnets during rotation.

Superconductivity • It is frequently suggested that S.C. magnets should be used to reduce gantry weight and size. • In practice, it is not trivial: • To build large bending angle S.C. dipoles, especially with large apertures. • To either use flexible cryogenic lines that roll on and off a drum on the gantry axis, or to mount the helium liquefier on the gantry. • To ensure the continuous flow of cryogens at all gantry angles. • To ramp S.C. magnets quickly. • If the S.C. dipole is iron-free, then the stray field is problematic for the detectors. • If the S.C. dipole has an iron yoke, then the weight is not so strongly reduced. • There are also security risks: • Quenching could frighten patients when the release valve opens. • There is a risk of oxygen deficiency if large quantities of helium are released. • There is a risk of cold burns (especially for the lungs) if large volumes of vapour are released. • A reduction in magnetic rigidity or treatment field would bring a more direct advantage.

Some controversy… On the other hand… • Severe engineering problems render the ‘ideal’ gantry expensive. • Some argue that: • 90% of patients require a treatment field of no more than 10 x 20 cm2. • A penetration depth of 20 cm is largely sufficient. • An exo-centric gantry is superior. • Sufficient angular access can be obtained with horizontal and 60 degree fixed beam lines by turning, tilting and sitting the patient. On the one hand… • Iso-centric gantries are the preferred solution of the medical community. • And there is considerable reluctance to relax the specifications i.e. • Treatment field: 40 x 40 cm2. • Magnetic rigidity corresponding to penetration up to 27 cm.

Conclusions • The design of the gantry depends on: • The method chosen for the rotational optics matching. • The type of beam delivery: passive spreading, wobbling or voxel scanning. • The transverse beam distributions: slow extracted beams or fast extracted beams. • Whether the beam is spread before or after the last dipole. • The magnetic rigidity: protons or light ions. • Check the patent situation for the design chosen. • Be critical of the size of the treatment field and the maximum beam penetration, as these two parameters are very expensive.

Post scriptum • The CD-ROM contains the full CERN Report 2000-006 of the Proton Ion Medical Machine Study (PIMMS). • The optics program used for the design can be copied from the CD-ROM and the lattice files for the various gantries can be found in the folder “Winagile/Lattices/Optical/Extrline/ for the individual modules and “Winagile/Lattices/Engineer/Extrline/ for the gantries set in their transfer lines. • The AutoCad concept design drawings of the elements can be found in the Hardware folder.

Loma Linda IBA IBA

Gantry Classification and Design for Hadron Therapy