Presenter(s)
Y. Li1, W. Li1, S. Yu2, K. W. Li3, L. S. Geng1, and Y. Zhang2; 1School of Physics, Beihang University, Beijing, China, 2Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education/Beijing), Department of Radiation Oncology, Peking University Cancer Hospital & Institute, Beijing Cancer Hospital & Institute, Beijing, China, 3CAS Ion Medical Technology Co., Ltd., Beijing, China
Purpose/Objective(s): Dual-energy CT (DECT) enhances material differentiation by leveraging energy-dependent attenuation properties particularly for carbon ion therapy. This study systematically verified a recently proposed machine-learning-based DECT (ML-DECT) elemental decomposition method in dose calculation, robustness of noise, dose uncertainty estimation, and dose monitoring of carbon ion therapy.
Materials/Methods: The ICRP110 human phantom was used as ground truth. The calculated DECT numbers of each voxel were used as input of the ML-DECT method to obtain the elemental composition. Up to 5% Gaussian noise was added to the DECT numbers and then used as input of the ML-DECT method to obtain the elemental composition. Relevant sources of uncertainties were analyzed and propagated. The phantom was then irradiated by a set of carbon ion pencil beams via Monte-Carlo simulation. The physical and biological doses without and with noise, dose uncertainty, and signals of annihilation photons were scored. All the results were compared with those obtained from a parameterized DECT (PA-DECT) method and the ground truth.
Results: Compared with the PA-DECT method, the gamma passing rates of physical and biological dose without noise of the ML-DECT method were reduced by up to 4% and 10%, respectively, under criteria of 1mm%, 1%. In the presence of noise, the improvement become to 11% and over 20%. The relative dose uncertainty of the ML-DECT method was reduced by 5% (physical) and 8% (biological). The mean relative error of the signals of annihilation photons of the ML-DECT method was reduced by 4%.
Conclusion: In the application of carbon ion therapy, the ML-DECT method showed improved dose accuracy and robustness to noise, reduced dose uncertainty, and improved dose monitoring signals.