A new tissue segmentation method to calculate 3D dose in small animal radiation therapy

BACKGROUND: In pre-clinical animal experiments, radiation delivery is usually delivered with kV photon beams, in contrast to the MV beams used in clinical irradiation, because of the small size of the animals. At this medium energy range, however, the contribution of the photoelectric effect to absorbed dose is significant. Accurate dose calculation therefore requires a more detailed tissue definition because both density (ρ) and elemental composition (Z(eff)) affect the dose distribution. Moreover, when applied to cone beam CT (CBCT) acquisitions, the stoichiometric calibration of HU becomes inefficient as it is designed for highly collimated fan beam CT acquisitions. In this study, we propose an automatic tissue segmentation method of CBCT imaging that assigns both density (ρ) and elemental composition (Z(eff)) in small animal dose calculation. METHODS: The method is based on the relationship found between CBCT number and ρ*Z(eff) product computed from known materials. Monte Carlo calculations were performed to evaluate the impact of ρZ(eff) variation on the absorbed dose in tissues. These results led to the creation of a tissue database composed of artificial tissues interpolated from tissue values published by the ICRU. The ρZ(eff) method was validated by measuring transmitted doses through tissue substitute cylinders and a mouse with EBT3 film. Measurements were compared to the results of the Monte Carlo calculations. RESULTS: The study of the impact of ρZ(eff) variation over the range of materials, from ρZ(eff) = 2 g.cm(− 3) (lung) to 27 g.cm(− 3) (cortical bone) led to the creation of 125 artificial tissues. For tissue substitute cylinders, the use of ρZ(eff) method led to maximal and average relative differences between the Monte Carlo results and the EBT3 measurements of 3.6% and 1.6%. Equivalent comparison for the mouse gave maximal and average relative differences of 4.4% and 1.2%, inside the 80% isodose area. Gamma analysis led to a 94.9% success rate in the 10% isodose area with 4% and 0.3 mm criteria in dose and distance. CONCLUSIONS: Our new tissue segmentation method was developed for 40kVp CBCT images. Both density and elemental composition are assigned to each voxel by using a relationship between HU and the product ρZ(eff). The method, validated by comparing measurements and calculations, enables more accurate small animal dose distribution calculated on low energy CBCT images.