Fig. 1. The design of the OCTOPOD (Optical Cone beam TOmograph for Proton Online Dosimetry) detector. (a) 3D view of the octagonally shaped OCTOPOD detector with an outer diameter of . (b) The schematic frontal cut-plane of the OCTOPOD with the diameter reconstruction volume filled with liquid scintillator (in blue), the octagonal polymethylmethacrylate (PMMA) housing (in green) enclosing the reconstruction volume and connecting to the three pinholes and the sensor plane at the distance of defined by a PMMA cuboid. The setup is repeated for each side of the octagonal PMMA housing to obtain 24 cone beam (CB) projections of the reconstruction volume. (c) The schematic side cut-plane with the long reconstruction volume.

Fig. 2. The correction matrix for the reconstruction algorithm showing the inverse relative sampling efficiency for all voxels in the reconstruction volume. From left to right, the side cut-plane, the top cut-plane and the frontal cut-plane of the 3D distribution are shown. For the side and top view, maximum correction is required for the front and rear part of the reconstruction volume , where the distance to the pinholes is maximized and the effective pinhole size is minimized. For the frontal view taken at the center of the reconstruction volume, the detection efficiency is solely determined by the inverse-square law effect because the effective pinhole size does not change. Hence, the required signal correction increases towards the reconstruction volume center.

Fig. 3. The single-pinhole measurements performed at the experimental proton beamline at University Proton Therapy Dresden (UPTD) with one sensor installed and a rotating detector for three different proton beam setups. The reconstructed 3D signals are shown as the side cut-plane –, the top cut-plane (a)–(c) and the frontal cut-plane (a)–(c). The red lines in each view mark the position of the other cut-planes with respect to the presented view. (a) Experimental setup and reconstruction of the Bragg peak (BP). (b) Experimental setup and reconstruction of the BP with a polymethylmethacrylate (PMMA) block in front of the detector. (c) Experimental setup and reconstruction of the BP with a PMMA block half-blocking the proton beam in front of the detector. Note that here the maximum of the colorbar of the reconstruction was adjusted to the maximum signal in the BP since scattered protons directly detected by the upper sensor positions lead to an artefact in the reconstruction visible in the upper right corner of the side cut-plane.

Fig. 4. The iteration steps of the reconstruction algorithm shown as the laterally integrated 3D reconstruction signal plotted over the penetration depth for the Bragg peak (BP) reconstruction.

Fig. 5. The deconvolution of the measured cone beam (CB) projections compared for single-pinhole and multi-pinhole geometries for an proton Bragg peak (BP, 2 cm polymethylmethacrylate (PMMA) in front of the detector). Increasing the number of pinholes leads to a higher sensitivity but also changes the shape of the obtained projections. The original shape of the CB projections is reconstructed in a deconvolution with the line spread functions (LSFs) of the pinhole geometry using an iterative Richardson–Lucy deconvolution algorithm^[35].

Fig. 6. Comparison of the reconstruction performance for different single pinholes and multi-pinhole grids. (a) The performance comparison between the 1D depth signal profiles of Bragg peaks (BPs) measured with a single pinhole (solid lines) and a 3×3 pinhole grid (dashed lines). The data were simultaneously measured with two different sensors (one sensor with single-pinhole geometry and the other with a 3×3 pinhole grid) and a rotating detector. (b) The 1D depth signal profiles for a 91-pinhole grid, applied for all eight sensors simultaneously. The measured profiles (solid lines, corrected for LET quenching) are compared to simulated (dashed lines) depth dose profiles obtained in a lateral diameter region centered in the BP. Note that the BPs are labeled with an offset of , which needs to be introduced in the Monte Carlo (MC) simulation to benchmark the measured depth dose profiles against RCF measurements.

Fig. 7. Comparison of Monte Carlo (MC) simulation results (dashed lines) with the 3D dose distribution measured with a calibrated stack of radiochromic films (RCFs, type EBT3 from Gafchromic, solid lines). The depth dose distributions shown are derived from the volumetric dose measurement by averaging over a lateral diameter region centered on the proton beam profile. The dose- and fluence-weighted depth-dependent distributions of the linear energy transfer (LET) are derived from the MC simulations and are used to correct the RCF response according to Ref. [37]. The dose in the plateau region was additionally measured with an Advanced Markus ionization chamber (AMC, type 34045, PTW). Since the calibration of the RCFs was performed in the middle of a spread-out Bragg peak, the calibration LET is closer to the LET at the BP maximum than to the lower LET at the entrance plateau. Therefore, the maximum of the LET-corrected RCF dose distribution is normalized to the maximum of the dose distribution measured with the RCF stack, leading to an agreement with the entrance dose measured with the AMC and the simulated depth dose profile.

Fig. 8. The setup for the OCTOPOD measurements at Draco PW. (a) Open rear side of the air-filled vacuum-compatible housing of the OCTOPOD detector. (b) Front side of the OCTOPOD housing with a thin Kapton window to enable the proton detection. The Kapton window is protected by a thin aluminum foil and the vacuum housing of the OCTOPOD detector is covered with a lead plate below a ceramic plate in laser forward direction (-projection sensor direction). (c) Sketch of the experimental setup at Draco PW.

Fig. 9. Comparison of cut-planes through the 3D water dose distribution measured with the OCTOPOD and a stack of radiochromic films (RCFs) as a reference detector. (a) Side, top and frontal cut-planes at water depth. The red box marks a reconstruction artifact, which results from directly detected bremsstrahlung/electrons by the -projection sensor. (b) Frontal cut-planes for varying water depths, as denoted in each subfigure. The red lines in each view mark the position of the other cut-planes with respect to the presented view. The signal was reconstructed with 20 iterations.

Fig. 10. Influence of the laser pulse energy on the proton dose distribution. (a) Reconstructed frontal cut-planes at water depth. Note that the images show the scintillator depth. (b) Reconstructed top cut-planes. (c) Mean depth dose distribution evaluated in a circular region of interest with diameter, as marked by the black dotted circle and lines in the frontal and top cut-planes, respectively. The gray region marks where the dose sensitivity limit is estimated to be. The red lines in each view mark the position of the other cut-planes with respect to the presented view. The signal was reconstructed with 20 iterations.

Fig. 11. Scan of the focal spot position on target showing the top and frontal cut-planes at water depth ( proton energy) of the 3D water dose distribution measured with the OCTOPOD. The blue curved line represents the focus size. The target chamber center (TCC) is at and the positive target position changes are in the direction of the focusing optics. The laser energy on target is , except for the target position (). The red lines in each view mark the position of the other cut-planes with respect to the presented view. The signal was reconstructed with 20 iterations.