Session Item

To fully exploit the benefits of proton therapy, an accurate stopping power ratio (SPR) prediction is necessary. In this study, we evaluated the dose differences between robustly optimized proton plans based on single-energy CT (SECT) and dual-energy CT (DECT) for neuro-oncological patients.

57 patients treated with proton therapy received a planning single energy CT (pSECT), followed by weekly repeat CTs acquired in both SECT (reSECT) and DECT (reDECT) mode (SOMATOM Drive or Confidence, Siemens Healthineers, Forchheim, Germany). On average, each patient had 4.6 sets of paired reSECTs and reDECTs (range: 2-6). Clinical plans were created in RayStation 10A (RaySearch Laboratories, Stockholm, Sweden). All plans were robustly optimized with a range uncertainty of ±3% and an isotropic setup uncertainty of 1 mm.

Commercial DirectSPR software (Siemens Healthineers) was used to create SPR maps of the reDECTs. The contours on the reSECT was copied to the corresponding reDECT. The clinical proton plan was robustly re-evaluated on all weekly reSECTs and reDECTs. Dose-volume histogram (DVH) parameters were extracted from the nominal dose distribution (nom) as well as the voxel-wise minimum and maximum (VWmin/VWmax) dose distributions. As the largest dose difference was expected distally to the target, only organs at risk (OARs) overlapping the clinical target volume (CTV) expanded with 2 cm on the pSECT were included in the evaluations. Moreover, ring structures of 1-5 mm were created around the CTV and included in the analysis. The DVH-parameters for the CTVs, OARs and ring structures were extracted from both the reSECTs and reDECTs, and the DVH-differences were computed. For each patient, the averages of the DVH-differences over the repeat CTs were computed.

For all patients, differences in CTV V94% for both the VWmin and nom dose distributions were within -1.5% and 0.2%, both with medians of 0.0% (and Inter Quartile Ranges (IQR) of 0.2% and 0.0%, respectively, Figure 1). The difference in maximum dose (quantified by D0.03cc) to the CTV for both the VWmax as nom were all within -0.7 and 0.6 Gy, with medians of 0.1 Gy (IQR=0.2 Gy) and 0.0 Gy (IQR=0.2 Gy), respectively. For the OARs, larger differences were observed, of up to 10.9 Gy and 5.8 Gy, with a median of 0.6 Gy and 0.4 Gy (IQR=2.3 Gy and 1.2 Gy) for D0.03cc in the VWmax and nom dose distributions, respectively, for the brainstem surface (Figure 1 and 2). The proton ranges on the DECTs were generally larger than on the SECTs causing dose differences (Figure 2). The same was seen for the V95% for the ring structures in the nom dose distributions which showed an increase in median difference up to 0.9% (IQR= 1.3%) with increasing ring size.

Dosimetric differences in OARs were found between DECT and SECT-based dose distributions due to different implementations of SPR calculation. These implementations have negligible influence for CTV coverage but could influence OAR dose calculations.