Abstract ID: 114 The impact of dual-energy CT tissue segmentation for low-dose rate prostate brachytherapy Monte Carlo dose calculations

Abstract

Purpose To evaluate the impact of a novel tissue segmentation method based on dual-energy CT (DECT) for low-dose rate (LDR) brachytherapy dose calculations, by comparison with a reference single-energy CT (SECT) segmentation method. Methods A virtual patient geometry is created using the DICOM-RT of a real patient pelvis SECT scan, where known elemental compositions and varying densities are overwritten in each voxel to define a reference phantom. Simulated CT images are generated using XCOM attenuation coefficients, with a 100 kVp spectrum for SECT, and 80 and 140Sn kVp for DECT. Tissue segmentations for Monte Carlo (MC) dose calculations are performed with both SECT and DECT and compared with the reference geometry. For SECT, the method of Schneider et al. [1] is used and for DECT, the eigentissue decomposition of Lalonde & Bouchard [2] used in combination with a Bayesian estimator. A LDR prostate brachytherapy treatment is planned with 125I sources and calculated using the MC code Brachydose for all three cases. Dose distributions and dose-volume histograms (DVH) are compared to the reference dose distribution to investigate the accuracy of the tissue segmentation methods. Results For noiseless images, DECT-based tissue segmentation outperforms the SECT procedure with a relative dose distribution root mean square error (RMSE) of 3.08% versus 8.02%, and provides DVH closest to the reference for all tissues. For a medium level of noise (12 HU), Bayesian eigentissue decomposition performs better with a dose calculation RMSE of 6.11% and 8.49% for DECT and SECT, respectively. Both methods yield similar DVHs for the prostate while DECT segmentation remains more accurate for organs at risk. Conclusion Our study shows that DECT-based tissue segmentation has the potential to provide LDR brachytherapy dose distributions with higher accuracy that conventional SECT in a clinical context, even in the presence of noise.