ESTRO 38 Abstract book

S1027 ESTRO 38

Conclusion Our results suggest that this novel way of converting and correcting CBCT images to RSP for the use in proton therapy treatment planning is feasible. The use of readily available artifact correction techniques are expected to further improve RSP accuracy. The technique could be easily employed in practice using readily available detector hardware. EP-1891 A new hotspot correction algorithm in Modulated Electron Radiation Therapy utilizing 3D printed boli B. Basaric 1 , J. Robar 2 , R. Orbovic 1 , A. Dennis 1 , C. Majcher 1 1 Adaptiiv Medical Technologies Inc., Research and Development, Halifax, Canada ; 2 Nova Scotia Health Authority, Department of Medical Physics, Halifax, Canada Purpose or Objective Modulated Electron Radiation Therapy (MERT) modulates the electron beam by changing the thickness of the bolus resulting in a 3D printed non-uniform thickness bolus that conforms the prescribed dose to the distal part of the PTV. However, there may be a certain trade-off between the resulting conformity and dose homogeneity to the PTV depending on shapes of the PTV and modulated bolus and on the energy of the beam. The objective of this work is to introduce a new hotspot algorithm that changes the shape of the modified electron bolus in such a way it reduces hotspots and keeps dose conformity to the PTV similar to the initial modified electron bolus. Material and Methods Initial modulation of the 3D printed bolus that insures the conformity of the prescribed dose to the distal part of the PTV was designed in the “3D Bolus” software (Adaptiiv Medical Technologies Inc.). Overall, 6 MERT plans were calculated in Eclipse TPS (Varian Medical Systems, Inc.). 5 plans were calculated on a phantom where different shaped PTVs were drawn which RT plans ultimately resulted in different dose homogeneity and an overall good conformity in all cases. Hotspots ranged from 114% (1 small peak case) to 142.2% (1 large peak case). A new hotspot correction algorithm uses PTV, contoured hotspot and modulated bolus RT Structures to estimate the location of peaks in a 3D modulated bolus winged edge mesh in order to reduce the height of these peaks in respect to the corresponding valleys resulting in less scattered radiation toward the hotspot area. The amount of peak reduction on the modulated bolus was scaled from 100% (intact modulated bolus) to 0% (completely reduced peaks resulting in flattened modulated bolus). For each of the 6 cases, hotspot corrected modulated bolus was produced using peak reduction scaling factors of 80%, 60%, 40%, 20% and 0%. Results In all cases, a significant hotspot reduction was noticed which ranged from 23-40% in the 1 large case, 24-33% in the 2 large peaks case, 6-15% in the 1 small peak case, 8- 22% in the 2 small peaks case,10-20% in the no-peaks case and 6-9% in the patient case. In all cases, a progressive reduction of maximum dose to the plan was noticed as the peak reduction scaling factor was closer to 0% while at the same time a progressive loss of conformity of the prescribed dose to the PTV was also noticed. All cases produced at least 1 plan (with one particular peak reduction scaling factor) that satisfied both clinically acceptable level of maximum dose to the plan and conformity of the prescribed dose to target volume that was comparable to the initial modulated bolus electron plan. Conclusion The new hotspot correction algorithm showed in all cases that by scaling the height reduction of the bolus peaks, a clinically acceptable plan can be achieved that satisfies

Purpose or Objective Cone beam CT (CBCT) is becoming a commonly available imaging technology in radiotherapy. However, the presence of artifacts and poor accuracy of Hounsfield Units (HU) preclude its use for treatment planning purposes in proton therapy where range accuracy is crucial. We investigate whether a sufficiently accurate proton relative stopping power (RSP) map can be obtained from a CBCT image when using a patient specific calibration of the HU-RSP conversion curve based on proton radiographic images. Material and Methods We employ a calibration procedure which on one hand uses a proton radiographic image potentially acquired with readily available detector hardware in pencil beam scanning delivery and on the other hand a proton digitally reconstructed radiography based on a CBCT image. The difference between the two is minimized using a recently developed regularized optimization method. The result is a HU-RSP curve which can be used as input to the treatment planning system in combination with the CBCT image instead of the conventional X-ray planning CT. We applied the calibration to head and neck patient cases who were scheduled for proton therapy and received regular CBCT scans prior to treatment sessions. The CBCT images were rigidly registered to the planning CT. Tomographic reconstruction of the CBCT images was performed with the vendor software including only basic artifact correction for now. Proton radiographies were simulated with the GPU accelerated Monte Carlo code “Fred” using an idealized range telescope set-up and the same data processing chain which we apply to experimental data. The X-ray planning CT converted to RSP through a stoichiometric HU-RSP conversion curve was used as simulation input. Results Figure 1 shows an example of an optimized HU-RSP conversion curve to be applied to a CBCT image. It differs noticeably from the conventional one especially in the range of lower HU values. The HU intervals of the curve, which are usually chosen in accordance with human tissue properties, need to be adjusted in this special application and a regular spacing appeared to yield most reasonable results. Figure 2 shows the difference of RSP values between the CBCT and the planning CT. Agreement is overall much better when using the optimized curve, although spatially dependent discrepancies are still visible. We are currently investigating to which extent more advanced artifact correction during the CBCT reconstruction can improve this. Preliminary results of treatment plan recalculation on CBCT images already showed great improvement when using the optimized HU- RSP.