Uli Weber

Treatment planning for heavy-ion radiotherapy: physical beam model and dose optimization

We describe a novel code system, TRiP, dedicated to the planning of radiotherapy with energetic ions, in particular 12C. The software is designed to cooperate with three-dimensional active dose shaping devices like the GSI raster scan system. This unique beam delivery system allows us to select any combination from a list of 253 individual beam energies, 7 different beam spot sizes and 15 intensity levels. The software includes a beam model adapted to and verified for carbon ions. Inverse planning techniques are implemented in order to obtain a uniform target dose distribution from clinical input data, i.e. CT images and patient contours. This implies the automatic generation of intensity modulated fields of heavy ions with as many as 40000 raster points, where each point corresponds to a specific beam position, energy and particle fluence. This set of data is directly passed to the beam delivery and control system. The treatment planning code has been in clinical use since the start of the GSI pilot project in December 1997. Forty-eight patients have been successfully planned and treated.

Design and construction of a ripple filter for a smoothed depth dose distribution in conformal particle therapy

The ripple filter was designed to broaden the Bragg maximum of carbon beams for the raster-scan technique, a special type of tumour-conformal ion beam treatment. In this technique the target volume is divided into individual layers that are treated sequentially by varying the energy from the accelerator stepwise. Because the unmodified Bragg maximum has a small half-width, below 1 mm for small energies (< 160 MeV u(-1)), homogeneous irradiation at small penetration depths of 2-6 cm can only be obtained by using a large number of energy steps. If the energy step is too large, ripples are produced in the superimposed depth dose distribution. The ripple filter widens a Bragg peak to a Gaussian peak with a half-width of more than 2 mm. This helps to smooth the extended Bragg peak and to reduce the number of energy steps required by a factor of two to three, leading to significantly shorter overall irradiation times and a higher particle fluence per layer. The ripple filter consists of a 2 mm thick Plexiglass (PMMA) plate with a periodic structure of fine grooves. It can be mounted 60 cm upstream of the patient as a stationary device, because the fine structure of the grooves is completely washed out by the lateral scattering of the beam.

Comparison of carbon ions versus protons.

At present, beam ion beam therapy has started to spread worldwide. In Europe and Asia, combined carbon/proton facilities are favored, but in the US, only proton centers are under construction. This development is partially due to the different funding procedures and partially due to the more complex physical and especially biologic features of the heavy ions. In this article, the basic properties of both ions are presented, and their features for therapy are outlined. This refers to the dose conformity, the general precision of the treatment, and the ability to monitor via in-beam positron emission tomography the ions range inside the patient. Then the very complex biologic features are treated, and, finally, the treatment plans are compared.

Acute Response of Pig Skin to Irradiation with 12C-ions or 200 kV X-rays

The acute response of pig skin to treatment with high energy carbon ions (plateau region) at the Gesellschaft für Schwerionenforschung (GSI, Darmstadt, Germany) was compared with changes after 200 kV x-irradiation. Carbon doses isoeffective to the x-ray doses were computed with a recently established model for calculation of the biological effect of heavy ions. Clinical changes and physiological symptoms (blood flow, erythema, trans-epidermal water loss, skin hydration) were scored. The parameters analyzed were maximum and mean values of each symptom during days 24 to 70 after irradiation, and the quantal endpoints for the establishment of dose effect curves were the median values of these. With exception of the maximum change in the red blood cell concentration (p < 0.02) no significant differences could be found in the response to x-rays and RBE-corrected heavy ions. These results indicate that the model is valid for the calculation of biological effects of 12C-ions (plateau region) and may at least for epidermis be applied to treatment planning.

Fragmentation of 120 and 200 MeV u−1 4He ions in water and PMMA targets

Recently, the use of 4He particles in cancer radiotherapy has been reconsidered as they potentially represent a good compromise between protons and 12C ions. The first step to achieve this goal is the development of a dedicated treatment planning system, for which basic physics information such as the characterization of the beam lateral scattering and fragmentation cross sections are required. In the present work, the attenuation of 4He primary particles and the build-up of secondary charged fragments at various depths in water and polymethyl methacrylate were investigated experimentally for 120 and 200 MeV u-1 beams delivered by the synchrotron at the Heidelberg Ion-Beam Therapy Center, Heidelberg. Species and isotope identification was accomplished combining energy loss and time-of-flight measurements. Differential yields and energy spectra of all fragments types were recorded between 0° and 20° with respect to the primary beam direction.

Optimization of the stopping-power-ratio to Hounsfield-value calibration curve in proton and heavy ion therapy

For CT-based dose calculation in ion therapy a link between the attenuation coefficients of photons and the stopping-power of particles has to be provided. There are two commonly known approaches to establish such a calibration curve, the stoichiometric calibration and direct measurements with tissue substitutes or animal samples. Both methods were investigated and compared. As input for the stoichiometric calibration the data from ICRP-report 23 were compared to newly available data from ICRP-report 110. By employing the newer data no relevant difference could be observed. The differences between the two acquisition methods (direct measurement and stoichiometric calibration) were systematically analyzed and quantified. The most relevant change was caused by the exchange of carbon and oxygen content in the substitutes in comparison to the data of the ICRP-reports and results in a general overshoot of the Bragg peak. The consequence of the differences between the calibration curves was investigated with treatment planning studies and iso-range surfaces. Range differences up to 6mm in treatment plans of the head were observed. Additionally two improvements are suggested which increase the accuracy of the calibration curve.

Monte Carlo simulations of new 2D ripple filters for particle therapy facilities

Abstract At particle therapy facilities with pencil beam scanning, the implementation of a ripple filter (RiFi) broadens the Bragg peak (BP), which leads to fewer energy steps from the accelerator required to obtain an homogeneous dose coverage of the planned target volume (PTV). At the Universitätsklinikum Gießen und Marburg, Germany, a new second generation RiFi has been developed with two-dimensional groove structures. In this work we evaluate this new RiFi design. Methods: The Monte Carlo (MC) code SHIELD-HIT12A is used to determine the RiFi-induced inhomogeneities in the dose distribution for various ion types, initial particle energies and distances from the RiFi to the phantom surface as well as in the depth of the phantom. The beam delivery and monitor system (BAMS) used at Marburg, the Heidelberg Ionentherapiezentrum (HIT), Universitätsklinikum Heidelberg, Germany and the GSI Helmholtzzentrum für Schwerionenforschung, Darmstadt, Germany is modeled and simulated. To evaluate the PTV dose coverage performance of the new RiFi design, the heavy ion treatment planning system TRiP98 is used for dose optimization. SHIELD-HIT12A is used to prepare the facility-specific physical dose kernels needed by TRiP, and for recalculating the physical dose distribution after TRiP optimization. Results: At short distances from the RiFi to the phantom surface fine structures in the dose distribution are observed. For various RiFis, ion types and initial particle energies the distance dmax at which maximum dose inhomogeneity occurs is found and an expression for dmax is deduced. The distance d0.01 at which the dose inhomogeneity is less than 1% is estimated and used as a threshold distance at which dose distributions are considered homogeneous. The MC data are found to agree with analytical expressions for dmax and d0.01; both are inversely related to the angular distribution. Increasing scatter from the beam delivery and monitoring system results in reduced dmax and d0.01. Furthermore, dmax and d0.01 are found to be proportional to the RiFi period λ. Conclusion: Our findings clearly indicate that the dose inhomogeneity induced by RiFis does not add uncertainties to the dose distribution in the clinical setting. The new RiFi design can be used in treatments to obtain homogeneous PTV dose coverage with fewer energy steps while improving lateral penumbra, thereby reducing the required treatment time.

Fluence inhomogeneities due to a ripple filter induced Moiré effect.

At particle therapy facilities with pencil beam scanning, the implementation of a ripple filter (RiFi) broadens the Bragg peak, so fewer energy steps from the accelerator are required for a homogeneous dose coverage of the planning target volume (PTV). However, sharply focusing the scanned pencil beams at the RiFi plane by ion optical settings can lead to a Moiré effect, causing fluence inhomogeneities at the isocenter. This has been experimentally proven at the Heidelberg Ionenstrahl-Therapiezentrum (HIT), Universitätsklinikum Heidelberg, Germany. 150 MeV u(-1) carbon-12 ions are used for irradiation with a 3 mm thick RiFi. The beam is focused in front of and as close to the RiFi plane as possible. The pencil beam width is estimated to be 0.78 mm at a 93 mm distance from the RiFi. Radiographic films are used to obtain the fluence profile 30 mm in front of the isocenter, 930 mm from the RiFi. The Monte Carlo (MC) code SHIELD-HIT12A is used to determine the RiFi-induced inhomogeneities in the fluence distribution at the isocenter for a similar setup, pencil beam widths at the RiFi plane ranging from σχ(RiFi to 1.2 mm and for scanning step sizes ranging from 1.5 to 3.7 mm. The beam application and monitoring system (BAMS) used at HIT is modelled and simulated. When the width of the pencil beams at the RiFi plane is much smaller than the scanning step size, the resulting inhomogeneous fluence distribution at the RiFi plane interfers with the inhomogeneous RiFi mass distribution and fluence inhomogeneity can be observed at the isocenter as large as an 8% deviation from the mean fluence. The inverse of the fluence ripple period at the isocenter is found to be the difference between the inverse of the RiFi period and the inverse of the scanning step size. We have been able to use MC simulations to reproduce the spacing of the ripple stripes seen in films irradiated at HIT. Our findings clearly indicate that pencil beams sharply focused near the RiFi plane result in fluence inhomogeneity at the isocenter. In the normal clinical application, such a setting should generally be avoided.

Dose build-up effects induced by delta electrons and target fragments in proton Bragg curves—measurements and simulations

Dose build-up effects in the entrance channel of proton Bragg curves were investigated in detail by means of simulations and experiments. There are two relevant dose build-up effects. Firstly, the δ-electron build-up effect which takes place in the first few millimeters of the tissue until an equilibrium state of the forward-scattered δ-electrons is reached. Secondly, the target fragment build-up effect that covers the first centimeters in the entrance channel of the proton Bragg curve. These target fragments are created in inelastic interactions of the beam protons with the target nuclei and partially have low kinetic energies and/or high atomic numbers compared to the incident beam protons. Consequently, the target fragments possess high LET values and thus an increased RBE. However, the production cross sections relevant for target fragmentation in ion beam therapy still have large uncertainties. Therefore, in this work target fragmentation was investigated indirectly by measuring low-noise proton Bragg curves with the focus placed on their build-up regions. The measurements clearly show the magnitude and shape of the two different build-up effects. Additionally, with the application of a magnetic filter, it was possible to separate the measurement of the target fragment build-up effect from the δ-electron build-up effect. Corresponding FLUKA Monte Carlo simulations were carried out for the experimental setup. A comparison of the experimental results with the FLUKA predictions enabled the assessment of the precision of FLUKA models, e.g. the δ-electron production models and the nuclear event generators which are responsible for target fragmentation reactions. It could be shown that the relevant models worked well to reproduce both build-up effects.

Dosimetric comparisons of carbon ion treatment plans for 1D and 2D ripple filters with variable thicknesses.

A ripple filter (RiFi)-also called mini-ridge filter-is a passive energy modulator used in particle beam treatments that broadens the Bragg peak (BP) as a function of its maximum thickness. The number of different energies requested from the accelerator can thus be reduced, which significantly reduces the treatment time. A new second generation RiFi with 2D groove shapes was developed using rapid prototyping, which optimizes the beam-modulating material and enables RiFi thicknesses of up to 6 mm. Carbon ion treatment plans were calculated using the standard 1D 3 mm thick RiFi and the new 4 and 6 mm 2D RiFis for spherical planning target volumes (PTVs) in water, eight stage I non-small cell lung cancer cases, four skull base chordoma cases and three prostate cancer cases. TRiP98 was used for treatment planning with facility-specific base data calculated with the Monte Carlo code SHIELD-HIT12A. Dose-volume-histograms, spatial dose distributions and dosimetric indexes were used for plan evaluation. Plan homogeneity and conformity of thinner RiFis were slightly superior to thicker RiFis but satisfactory results were obtained for all RiFis investigated. For the 6 mm RiFi, fine structures in the dose distribution caused by the larger energy steps were observed at the PTV edges, in particular for superficial and/or very small PTVs but performances for all RiFis increased with penetration depth due to straggling and scattering effects. Plans with the new RiFi design yielded for the studied cases comparable dosimetric results to the standard RiFi while the 4 and 6 mm RiFis lowered the irradiation time by 25-30% and 45-49%, respectively.