Preparations for photon external beam radiotherapy treatment planning of small animals

Abstract: Introduction:The XenX is a small animal irradiation system that acquires high-resolution cone-beam tomography (CBCT) images of small animals and treats small animals with higher precision than medical linear accelerators. MuriSlice is a software for CBCT image reconstruction. This thesis aims to optimise the CBCT image reconstruction protocol in MuriSlice to enable the segmentation of CBCT images in the μRayStation 8B treatment planning system. In addition, the calculation of absorbed dose performed by the Monte Carlo dose calculation algorithm in μRayStation 8B based on the reconstructed CBCT images is verified by comparison with radiochromic film measurements. Materials and methods: In the first part of this thesis, the XenX's integrated CBCT was used to acquire a set of projections of a cylindrical water phantom, a 3D printed mouse phantom, and a flat field. The acquired projections were used to optimise selected parameters of MuriSlice. First, the voxel size was selected based on the observed signal-to-noise ratio (SNR) in axial images of the cylindrical water phantom and recommendations of voxel size for CT imaging of small animals. Second, three corrections were performed on each projection, including 1) a correction for a distortion introduced by the lens in the digital camera of the CBCT imaging system, 2) a pixel-by-pixel correction to reduce the noise in the individual pixel of the flat field projections, and 3) a ring artefacts reduction using a 2D Butterworth filter. Third, the corrected projections were filtered with different pre-filters for further noise reduction. A pre-filter was selected based on the SNR and the contrast between soft tissue and bone in a transverse slice of the reconstructed images of the 3D printed mouse phantom. In the second part of this thesis, absorbed dose comparisons were carried out in four different slab phantoms containing various combinations of polystyrene, aluminium, and cork. Each phantom was irradiated by a 220 kV circular beam of X-ray photons (Ø = 10 mm). A set of CBCT projections for each slab phantom was acquired, reconstructed with the optimised protocol, and imported into μRayStation 8B. Inside μRayStation 8B, the reconstructed images of the slab phantoms were segmented into regions with the material composition assigned either based on the CBCT’s Hounsfield unit to mass density calibration or by manual identification. The absorbed dose to the assigned material in the CBCT image of each slab phantom was estimated using the Monte Carlo dose calculation algorithm in μRayStation 8B and then compared with radiochromic film measurements. Results: The reconstruction protocol in Murislice included a voxel size of (100μ〖m)〗^3and a edge-preserving denoising pre-filter. The optimisation of the reconstruction parameters resulted in a significant reduction of noise and near elimination of ring artefacts in the reconstructed image of the 3D printed mouse phantom. Furthermore, the contrast between the structures in the reconstructed CBCT image of the slab phantoms was sufficient to allow for the segmentation of relevant structures in the slab phantoms. The remaining noise resulted in a large fluctuation in estimated absorbed doses when the reconstructed images of the slab phantoms were segmented into regions with the assignment of the material composition based on the CBCT Hounsfield unit to mass density calibration. With the manual assignment of the material composition, the uncertainty in the estimated absorbed dose was within 1 %. The absorbed dose that was estimated with μRayStation 8B based on manual material assignment to the CBCT images of the slab phantoms was in reasonably good agreement with the film measurements, with a mean (max) of absolute difference in the percentage depth dose of 0.4%