ESTRO 38 Abstract book

S500 ESTRO 38

PO-0931 Application of a thin, energy-layer specific multi-leaf collimator for proton pencil beam scanning. C. Winterhalter 1,2 , G. Meier 1 , D. Oxley 1 , D.C. Weber 1,3,4 , A. Lomax 1,2 , S. Safai 1 1 Paul Scherrer Institute, Center for Proton Therapy, Villigen PSI, Switzerland ; 2 ETH Zurich, Physics Department, Zurich, Switzerland ; 3 University Hospital of Zurich, Radiation Oncology Department, Zurich, Switzerland ; 4 University Hospital of Bern, Radiation Oncology Department, Bern, Switzerland Purpose or Objective For pencil beam scanning (PBS) proton therapy, lateral fall-off has been shown to be inferior when compared to passive scattering proton therapy. For shallow tumours adjacent and close to critical structures, this might lead to unnecessary high doses to organs at risk. As such, collimation could potentially improve PBS proton dose distributions. Material and Methods For four patients with tumours located close to the brainstem, we have investigated the potential advantages of the use of a thin, multi-leaf collimator for energy-level specific collimation. By thin, we mean a collimator with a thickness just sufficient to stop protons with ranges in water of up to 15 cm, beyond which lateral penumbra is anyway determined by scattering in the patient. Additionally, two different pencil beam placement techniques - rectilinear grids (4 mm spacing, clinical approach) or contour scanning - have been analysed with and without collimation using different expansions for the contour scanning scenarios (0mm, 1mm, 2mm, 3mm, see Meier et al (Phys. 2017 Med. Biol. 62; 2398)). For the optimization process, both collimated and un-collimated pencil beams have been included in an analytical dose calculation, but all final dose distributions have been calculated using Monte Carlo (TOPAS). Results Figure 1 shows the improvement in penumbra when moving from grid to collimated contour scanning for an example patient field. For this case, grid scanning combined with collimation reduced the V30% outside the target by 20% in comparison to the un-collimated grid scenario. For contour scanning alone, a maximal V30% reduction outside the target of 26% was achieved (0mm contour expansion) which increased to 33% with collimation (2mm contour expansion). These improvements however come at the cost of reduced target dose homogeneity. As such, the best un- collimated/collimated dose distributions (i.e. scenarios retaining dose homogeneity while reducing dose to the normal tissue) were achieved with a 1mm/3mm contour expansion, without and with collimation respectively (c.f. figure 1). Finally, for four patients, collimated (shallow) and un-collimated (deep) fields have been combined and compared to the best un-collimated approach (Table 1), showing that the use of a collimator could reduce the V30% by 0.8-8.0% and the mean dose to the brainstem by 1.5- 3.3% for such combined plans.

Purpose or Objective Knowledge-based planning (KBP) has the potential to improve plan quality. We have recently implemented KBP (Rapidplan, Eclipse, Varian Medical Systems), in our clinic. Here we report on comparison between non-KBP and KBP treatment plans for advanced head and neck cancer (HCN) patients. Material and Methods The KBP model was validated on twenty HNC patients. Three target volumes (TV) are covered by a SIB plan, with prescribed doses (PD) of 66Gy (PTV1), 60Gy (PTV2) and 50Gy (PTV3). Two planning strategies were applied: one with optimization parameters from a generic HNC template and one with patient specific KBP optimization parameter. All plans were three-arc VMAT plans. The two plans were compared using quantitative metrics for target coverage, homogeneity and conformity, steep dose fall- off at PTV1, Integral dose, low-dose bath and mean doses to the brainstem, salivary glands, oral cavity, lips, thyroid, and swallowing structures. P-values were calculated using Wilcoxon signed rank test with p < 0.05 considered significant. Results Target parameters are shown in fig. 1. No differences were observed for target coverage (data not shown). Target homogeneity was slightly higher for KBP than non- KBP plans. Three out of the four metrics for target conformity show that at least PTV1 and PTV3 are less conformal for KBP than non-KBP. The steep dose-fall of PTV1 is higher for KBP than non-KBP. Comparisons for normal tissues are shown in fig. 2. Using KBP decreases the mean doses to the pharyngeal constrictor muscles (PCM), the glottis larynx and the thyroid gland, whereas the remaining OAR mean doses remain unchanged. The volumes receiving 10 Gy and 15 Gy are decreased by KBP and the integral dose decreases slightly from 127.0J [min: 54.0J; max: 226,3J] to 126.0J [52.9J; 224.1J] by KBP (p- value: 0.018).

Conclusion KBP gives similar target coverage as previously accepted clinical plans, with the advantage of reducing the mean doses to a number of OAR, slightly lowering the volume receiving 10 Gy and 15 Gy as well as the integral dose. Since the conformity of PTV3 is decreased, the OAR benefits are due to the steeper dose gradient.