A biologically adapted dose-escalation approach, demonstrated for 18F-FET-PET in brain tumors.

Purpose: To demonstrate the feasibility of a biologically adapted dose-escalation approach to brain tumors.

Material And Methods: Due to the specific accumulation of fluoroethyltyrosine (FET) in brain tumors, (18)F-FET-PET imaging is used to derive a voxel-by-voxel dose distribution. Although the kinetics of (18)F-FET are not completely understood, the authors regard regions with high tracer uptake as vital and aggressive tumor and use a linear dose-escalation function between SUV (standard uptake value) 3 and SUV 5. The resulting dose distribution is then planned using the inverse Monte Carlo treatment- planning system IKO. In a theoretical study, the dose range is clinically adapted from 1.8 Gy to 2.68 Gy per fraction (with a total of 30 fractions). In a second study, the maximum dose of the model is increased step by step from 2.5 Gy to 3.4 Gy to investigate whether a significant dose escalation to tracer-accumulating subvolumes is possible without affecting the shell-shaped organ at risk (OAR). For all dose-escalation levels the dose difference Delta D of each voxel inside the target volume is calculated and the mean dose difference Delta D and their standard deviation sigma Delta D are determined. The dose to the OAR is evaluated by the dose values D OAR 50% and D OAR 5%, which are the dose values not exceeded by 50% and 5% of the volume, respectively.

Results: The inhomogeneous dose prescription is achieved with high accuracy (Delta D < 0.03 +/- 0.3 Gy/fraction). The maximum dose can be increased remarkably, without increasing the dose to the OAR (standard deviation of D OAR 50% < 0.02 Gy/fraction and of D OAR 5% < 0.05 Gy/fraction).

Conclusion: Assuming that regions with high tracer uptake can be interpreted as target for radiotherapy, (18)F-FET-PET-based "dose painting by numbers" applied to brain tumors is a feasible approach. The dose, and therefore potentially the chance of tumor control, can be enhanced. The proposed model can easily be transferred to other tracers and tumor entities.