Room: AAPM ePoster Library
Purpose: Previous study has verified the feasibility of a novel design of proton computed tomography (pCT) through Monte Carlo simulation with a cylinder phantom. In this study, we further evaluated the ability of the pCT to image real human anatomy by using a set of virtual human head phantoms.
Methods: The pCT detector was composed of a multiple-layer ionization chamber (MLIC) and two perpendicular strip chambers, recording the location and residual range of exiting protons from the imaging object. Proton stopping power ratio (SPR) was reconstructed through filtered-back projection. The human head phantoms were derived from a cohort of real patient CT scans acquired at our hospital. The Hounsfield Units (HUs) of CT images were converted to the composition of materials based on the Stoichiometric calibration. Uncertainty from Monte Carlo simulation and imaging reconstruction were investigated by overriding the head phantom with one material of water and three materials of adipose, brain, and bone respectively. The reconstructed SPR and water equivalent path length ( WEPL) were compared with ground truth in the overridden datasets and compared with calculated SPR from the Stoichiometric calibration. Signal-to-noise ratio (SNR) were reported in homogeneous region of interest (ROI).
Results: Histogram of the percentage difference between reconstructed SPR from proton CT and the ground truth or calculated SPR showed that average percentage difference was -0.57%, -0.55%, and 0.54%, and the average percentage difference of WEPL were -0.38%, -0.37% and 0.96% in the datasets overridden with water, with three materials and no overridden dataset respectively. SNR in ROI was 28.91, 26.35, and 27.33 accordingly.
Conclusion: SPR reconstructed from the novel pCT design has acceptable image quality and excellent accuracy when imaging real human anatomy, indicating that the pCT design has great potential to be used in clinic for localization and plan adaption in proton therapy.