Purpose: Proton Computed Tomography (pCT), reconstructed from the integral depth dose (IDD) detected by a multiple-layer ionization chamber (MLIC), suffers from inferior image resolution and suboptimal accuracy in proton stopping power determination due to the large spot size. Although a previous study showed significant improvement with beamline collimation, it would require modification of beamline and also cause extra ambient neutron doses. Therefore, this study proposes a novel and universal model-based reconstruction for pencil beam scanning (PBS) machines with various spot sizes.
Methods: The IDD of an uncollimated proton beam detected by MLIC was modeled as a weighted sum of percentage depth doses of elemental infinitesimal beamlets evenly spaced by 1mm. The weights followed the source distribution. Squared L2 norm between calculated and measured IDDs were minimized at each gantry angle by correcting the water equivalent path length (WEPL) of all beamlets. Two iterations of correction with an intermediate smoothing were performed as a strategy to escape local minima. The corrected WEPLs were filtered back-projected to the final stopping power ratio (SPR) map. The reconstruction method was verified on a digital cylindrical water-based phantom with tissue-equivalent inserts and an ICRP adult phantom. SPR accuracy, spatial resolution, and imaging dose were evaluated.
Results: The mean absolute deviation of SPR in ROIs was 0.66%, 0.33%, and 0.48% for the cylindrical phantom, head, and lung region of the adult phantom, respectively. MTF(10%) derived from radial edge spread function of a cortical insert was 5.30cm⁻¹. The mean imaging dose was 1.71cGy and 1.95cGy at the head and lung region of the adult phantom, accordingly.
Conclusion: A novel model-based reconstruction method has been developed for pCT that can be utilized by all commercial PBS machines without complicated detector tracking individual protons. Excellent SPR accuracy, improved spatial resolution, and low imaging dose were achieved without additional collimation.