Treatment planning

The task

The heart of the GSI carbon ion radiotherapy project (as well as of the follow-ups at HIT and Marburg ) is the magnetic raster scanner. Together with the active energy variation supplied by GSI's SIS synchrotron it allows completely active three-dimensional dose shaping. 

Principle of active 3D dose shaping with the raster scanner. Lateral deflection by two dipole magnets, depth coverage by energy variation of the beam.

Intensity modulated scanner patterns in "last", some middle and "first" layer, respectively. Preirradiation leads to a lower number of particles in the proximal slices.

Treatment planning for our rasterscan system means determination of the necessary energies, positions and number of particles for the ion beam spots which, when overlaid, should give the prescribed biological dose distribution in three dimensions. The desired dose distribution is prescribed by the physicians and should conform to the target volume as much as possible, under the constraint that healthy tissue should be spared as much as possible. The task is equivalent to a nonlinear (due to RBE dependence on dose ) optimization problem with about 50000 degrees of freedom for our largest tumours. As shown in the above figure, this results in intensity modulated scan patterns even for single fields.
A visualization of the capabilities of such an active scanning system is shown in the figure below. 

A spherical volume of 6 cm diameter between 9 and 15 cm depth in water was irradiated with a conformal field of carbon ions. CR-39 nuclear track detectors were placed in water in steps of a few mm. After etching and proper illumination the radiation damage shows the precise delivery of dose in the target volume.

Ingredients

A prerequisite for treatment planning is the availability of computational models describing the interaction of ions with living matter with sufficient accuracy. The physical beam model comprises particle energy loss, energy loss straggling and nuclear fragmentation. Improved versions include multiple scattering as well.
The radiobiological model is based on LEM and predicts the biological dose response of various tissues in complex radiation fields composed of particles of all nuclear charges and energies.
Both models were (and are further) developed at GSI biophysics, and are integrated into our treatment planning computer program TRiP98.
Various algorithms of different speed and accuracy have been implemented to calculate the three-dimensional distribution of absorbed and biological dose, as well as other measurable quantities like cell survival, X-ray film blackening and response of thermoluminescence detectors. Various algorithms of different complexity and speed have been implemented to solve the nonlinear optimization problem.
It is not only a good idea but a necessity to verify the overall performance of treatment planning against measurements. Apart from our regular Quality Assurance (QA) procedures carried out by our colleagues from DKFZ. we have done so with our "head phantom" which measures cell survival in two dimensions and thus tests the whole treatment planning chain from modelling to actual irradiation with the raster scanner.

Experimental setup to measure cell survival distributions. Cell-covered slabs of a special plastic material are submerged into medium. The PMMA (aka Plexi) vessel has about 20cm diameter and thus resembles a human head.  
 

Cell survival distributions as a means to verify a model plan: an H-shaped target volume is exposed to two opposing fields. The area inbetween the arms of the H can be viewed as organs at risk.Colour comparison between measurement and TRiP98 prediction shows good overall agreement. (The numbers 0..1 correspond to 0..100% survival).
The lower half of the figure shows one-dimensional cuts along the lines indicated in the upper left figure. Good quantitative agreement between experiment and planning prediction clearly demonstrates the effect of target cell killing and healthy tissue sparing. 

Application & Results

In clinical practice treatment planning starts with the imaging of the tumour region and its surroundings by means of computer tomography (CT) and magnetic resonance (MR) imaging. This includes the delineation of target volume and organs at risk (OAR). The clinical steps are performed by our collaborators at Uni and DKFZ. With these inputs, 3D CT data and volume descriptions. TRiP98 optimizes the biological dose distribution and generates the appropriate raster scan data, i.e. particle irradiation patterns, by means of inverse planning. In most cases two approximately opposing fields already suffice to obtain satisfactory dose distributions. Sometimes three, and very rarely four fields are used. In general increasing the number of fields is not a good idea because this increases positioning and irradiation time, complicates QA procedures and exposes a larger volume of healthy tissue to the risk of secondary cancer.

For tumours at critical locations, i.e. in the head and neck region, comparison with "conventional" (photon) treatment plans consistently shows better dose conformation and better healthy tissue sparing with a lower number of fields.

At the time of this writing, more then 320 patients have been successfully planned and treated, mostly in the head and neck region, some also with sacral chordoma, and experimental treatment of prostate cancer has also begun. For more information on the clinical aspects look here.

Outlook & Developments

Development will continue on various fields. Multifield optimization will be improved in capability and speed. The models of beam interaction with living matter will be refined and extended to include other ion species as well, such as helium and oxygen. And finally three-dimensional planning will be augmented by the fourth dimension so that also moving targets can be planned and treated, i.e. tumours at locations affected by breathing or heart beat.

References

See Biophysics' publication list