Abstract
BACKGROUND AND PURPOSE: Dose Painting (DP) in radiotherapy is a strategy to account for the tumor microstructural heterogeneity. This study evaluated a DP approach in carbon ion radiotherapy (CIRT), using cell count estimates derived from diffusion-weighted magnetic resonance imaging (DWI). MATERIALS AND METHODS: Thirty-seven large sacral chordoma (SC) patients were analysed. Voxel-wise cell count was estimated from DWI using a published microstructural model. A Poisson-based tumor control probability (TCP) model, fitted on 27 patients, guided DP optimization in a research version of RayStation 2024A. The approach was tested on 10 patients, comparing uniform-dose plans against two strategies: dose redistribution (DR), which maintained the mean gross tumor volume (GTV) dose, and dose escalation (DE), which allowed a 3 % increase. Plan evaluation on targets and organs at risk (OARs) included dose-volume histogram metrics (D(95%), D(50%), D(1%)) and dose-averaged linear energy transfer (LET(d))-volume histogram metrics (L(98%), L(50%), L(1%)) to assess clinical acceptability. TCP gain quantified the benefit of biologically targeted strategies. TCP uncertainty was evaluated by propagating the microstructural model's errors to generate best- and worst-case cell count maps. RESULTS: DE and DR plans met clinical acceptability criteria. DE increased TCP from 75.5 ± 5.6 % to 83.3 ± 3.9 % (p < 0.001), with -3 to +5 percentage points variation under uncertainty. DR plans showed a TCP gain of 1.8 ± 1.0 percentage points. No significant dose or LET(d) increase was observed in OARs, while DE plans showed a lower L(98%) in GTV. CONCLUSIONS: Dose painting based on microstructural modelling in CIRT showed potential to improve TCP while sparing OARs.