Treatment Planning for the Carbon Ion Radiotherapy

In the three therapy beam times 2007 about 50 patients were planned and treated with our treatment planning software, TRiP98, and its development version, TRiP98BEAM. In the past, the standard method was to optimize the biological target dose distribution for the contributing partial fields singly. Due to the enormous effort for the computation of biological effects, only a simple, but still sufficiently accurate dose calculation algorithm could be used for numerical optimization [1]. The recently developed fast approximation of biological effects [2], allowed to introduce more sophisticated dose optimization algorithms into clinical practice. Dose calculation has been refined to reflect the full beam profile (including broadening by multiple scattering, if so desired), which leads to a better coverage of target edges. Dose optimization can now be performed for an (theoretically) arbitrary number of fields simultaneously (multi-field optimization, MFO). The sparing of healthy tissue (organs at risk, OAR) is incorporated and is controlled by specifying constraints, i.e. a maximum OAR dose and weight factors to impose tissue-specific ”penalties”. This new method [3] was repeatedly verified by means of biological dosimetry [4, 5]. Extensive tests were performed to ensure the robustness of the generated plans [6], i.e. the carbon ion fluence maps. Both, particle maps and partial field dose distributions show the required benign behaviour, i.e. contiguous scanner patterns and no steep gradients within the partial fields. To be on the safe side, the scanner patterns obtained with the new algorithms are always recalculated with the standard TRiP98 version for plan assessment. The new methods were cautiously put into clinical operation as early as 2006. Over time, an interesting variant of the MFO evolved: single field optimization with constraints (SFO/constraints). It means, that with the same algorithms as for MFO (but much less computing resources), dose distributions can sometimes be improved already for singly optimized partial fields. This is due to some redundancy inherent in the narrowly overlapping beam spots of the GSI raster scan procedure.