ESTRO 38 Abstract book

S295 ESTRO 38

An initial CT CC was determined according to the stochiometric approach. The CC consists of 5 segments: 3 lines that each describe organ-like, fat-like and bone-like tissues respectively and 2 lines as transitions between the primary segments. Unlike in routine practice, 3 types of vacuumed “fresh” animal tissue phantoms were made, consisting of (1) a pig’s head, (2) “thorax”, consisting of ribs, fat, liver and muscle, and (3) femoral bone. These phantoms were scanned on the CT (Somatom, Siemens) and transferred to the treatment planning system (RayStation, RaySearch) to calculate individual pencil beams directed through each phantom. On the CT scans a water slab was added behind the tissue samples to simulate the detector that was used for integral depth dose curve measurements. Tissue phantoms were positioned in the planned position at the proton therapy system (ProteusPlus, IBA) isocenter using the on-board x- ray imaging. A set of shot-through pencil beams of a 210 MeV energy was delivered, and depth dose profiles were measured using a multi-layer ionization chamber (Giraffe, IBA). Measured depth dose curves were compared to TPS calculated ones and the residual range error per spot was defined. Additionally, based on the WEPL of every spot through the tissue, the range error margin according to the published uncertainty recipe of 2.4% + 1 mm was defined. Ratios between measurement based and theoretical range error margins per spot were calculated (fig 1). The CT CC optimization was performed by identifying systematic shifts of the mean range error value per phantom type and minimizing the spread of ratios between residual range errors and range uncertainty margins.

Conclusion The feasibility of using range probing to assess the residual range errors was demonstrated in an institution specific setup. As a result, the published uncertainty margins may be reduced by ∼ 25%, allowing for potentially more conformal proton plans in the future. OC-0565 The optimization of prompt gamma based range estimation in proton therapy using Cramér-Rao theory E. Lens 1 , E. Tolboom 1 , D. Schaart 1 1 Delft University of Technology, Radiation Science and Technology, Delft, The Netherlands Purpose or Objective Various methods for in vivo range estimation during proton therapy based on the measurement of prompt gamma (PG) photons have been proposed. However, optimizing the method of detection by trial-and-error is a tedious endeavor. Here, we demonstrate the use of the Cramér- Rao lower bound (CRLB) to more quickly, and more objectively, arrive at an optimal detector design. Material and Methods The CRLB is based on the Cramér-Rao inequality and can be used to find the lower bound on the variance of any unbiased estimator of a parameter, given a statistical model of the observables. In this study the CRLB was used to derive the smallest possible variance on the proton range obtained from the detected PG photons, making use of the fact that the PG emission process is covered by Poisson statistics. The observables considered are the position, energy, and time of detection of the detected PG photons. We used the TOPAS (Geant4 based) Monte Carlo code to simulate a clinical proton pencil beam with 4·10 9 protons targeting a cylindrical, soft-tissue equivalent phantom. PG photons were scored on a cylindrical surface (i.e. detector; ⌀ =40 cm) coaxially surrounding the phantom. Spatially-, temporally-, and spectrally-resolved PG emission profiles corresponding to different proton ranges were generated by changing the initial proton energy. The detected photons were selected and tallied based on the location, energy, time, and angle of incidence on the simulated ideal detector. From the resulting signals, we calculated the CRLB as a function of several detector setup parameters such as detector size and location, bin size, energy resolution, and photon acceptance angle. Results We obtained relations between the CRLB and different detector setup parameters that allowed us to determine the optimal values for these detector properties. For most detector parameters there is a clear optimal value while the energy resolution and proton bunch width were preferably kept as small as possible. We found comparable CRLB values for proton range estimation based on either spatial, spectral, or temporal information, of 0.26 mm, 0.24 mm and 0.25 mm, respectively, if the detection parameters were optimized

Results When performing the analysis using the initial CT CC, increased range errors were observed for the femoral bone measurement set. Therefore, the slope of the CT CC line segment representing bone-like tissues was adjusted to ensure better agreement between measurements and calculations. Afterwards an independent measurement set for another femoral bone was included in the analysis and all data sets were recalculated using the optimized CT CC. For a full data set (about 1600 spots over all phantoms) the ratio of the actual range error and the uncertainty margin for 1.5σ did not exceed 0.75, indicating that the theoretical uncertainty recipe overestimates the actual range errors (fig 2).