Longitudinal CT number characterization of a novel upright CT for proton therapy planning

The combination of upright CT and patient positioner for particle therapy presents potential for high precision image-guided radiation therapy and online adaptive radiation therapy, where reliable CT performance is essential.

To evaluate the longitudinal stability of upright CT for proton radiation therapy, benchmark against standard of care conventional CT, and establish the feasibility of a simplified phantom configuration for efficient routine quality assurance (QA).

Following consensus guidelines, a calibration phantom with tissue mimicking inserts located at variable phantom positions was scanned on an upright CT (120 kV, 250 mAs) 14 times over 7 months. CT number intersession repeatability was assessed via standard deviation (SD). To assess size dependency due to beam hardening, body and head phantom measurements were evaluated. A Hounsfield look-up table (HLUT) for stopping power ratio (SPR) estimation was derived for proton dose calculations. The phantom was scanned on a conventional CT (120 kV, 250 mAs) for benchmarking. An anthropomorphic phantom (ATOM) was scanned on upright and conventional CT. Proton plans were developed in the phantom for prostate and spine stereotactic body radiation therapy on upright CT using pencil beam scanning techniques, robust optimization (3-5 mm setup, 3.5% range uncertainties), and Monte Carlo dose calculation. Dose was computed on co-registered conventional CT datasets. Dose agreement on upright and conventional CT was assessed. To develop an efficient approach for routine QA, a simplified phantom configuration (1 scan with 4 bone inserts) was scanned on upright CT over 8 months (15 imaging sessions, 5 images/session). Inter- and intra-session repeatability were assessed. The Wilcoxon signed-rank tests were used to test for significant differences between head/body and consensus/simplified phantom configurations.

Using consensus phantom configurations, upright CT demonstrated excellent longitudinal CT number stability with minimal inter-session SD ≤4.9 HU. System upgrades and recalibration introduced marginal offsets (difference in CT numbers ∆HU ≤14.4 HU). Size dependency of beam hardening was identified with statistically significant (p <0.05) differences in upright CT numbers (maximum ∆HU = 136 HU in cortical bone), leading to |∆SPR| up to 0.079 comparing body and head phantom results. Comparing proton dose calculated on upright and conventional CT using corresponding consensus-derived HLUTs, for both spine and prostate plans, local dose differences were found at distal end of beam path due to SPR differences with minimal target coverage differences <0.3% and gamma pass rates >99% at 1 mm/1%. Using a simplified phantom configuration, longitudinal upright CT number stability was also excellent (inter-session SD ≤3.2 HU and intra-session SD≤1.6 HU) although when comparing results of the consensus versus simplified phantom configurations, upright CT numbers were substantially different (p <0.05, maximum ∆HU = -76 HU in cortical bone) yielding |∆SPR| up to 0.044.

Upright CT demonstrated excellent longitudinal CT number stability as required for treatment planning and as a step toward adaptive proton therapy. Differences were noted between body and head phantom results, suggesting value in size-specific calibration protocols. The feasibility of using a simplified phantom configuration was demonstrated to support efficient QA to monitor machine stability. Future work includes extending the investigation to multi-institution validation.