KEY POINTS
- This multicenter experimental study compared relative stopping power (RSP) estimation across six European clinical and research centers, using an ex vivo biological phantom and a proton CT (pCT) prototype as a common reference. Investigators evaluated nine X-ray CT systems, including five single-energy CT (SECT) and four dual-energy CT (DECT) scanners, with each participating institution applying its own clinical or research calibration method.
- The phantom contained formalin-stabilized bovine adipose tissue, muscle, trabecular bone and compact bone, embedded in a rigid agar-based matrix. Additional external rings created three configurations approximating extremity, head and pelvic anatomical sizes, allowing assessment of scanner calibration, object size and beam-hardening effects. X-ray CT-derived RSP maps were spatially registered and compared voxel by voxel with the pCT-derived reference.
- Intercenter RSP variability was approximately 1% in adipose and muscle, but increased substantially in bone. At the reported 1.5σ variability level, compact bone reached 3.9 ± 0.2% for the extremity and head configurations and 4.6 ± 0.3% for the pelvic configuration. Corresponding variability for trabecular bone increased from approximately 1.2–1.3% to 1.7% with larger phantom size.
- Relative to proton CT, mean absolute RSP deviations were approximately 0.9–1.7% for soft tissues, 2.0–2.3% for trabecular bone, and 4.1–4.6% for compact bone, depending on anatomical configuration. In the pelvic configuration, mean absolute differences were 1.7% for adipose, 1.2% for muscle, 2.3% for trabecular bone and 4.1% for compact bone, demonstrating that calibration discrepancies are strongly dependent on tissue composition.
- More advanced CT technology did not guarantee better stopping-power estimates. For pelvic trabecular bone, median absolute deviations were approximately 1.89% with SECT versus 3.03% with DECT, while corresponding compact bone deviations were 3.15% versus 6.06%. These comparisons involved limited scanner numbers and different institutional calibration protocols, so they do not establish general superiority of SECT over DECT.
- Beam-hardening correction was associated with closer agreement with pCT, particularly in adipose and muscle, whereas using one, two or size-specific Hounsfield unit lookup tables did not consistently outperform alternative strategies. Centers incorporating biological tissues into calibration validation also tended to show smaller soft-tissue deviations. Increasing phantom size generally increased discrepancies, emphasizing the importance of anatomical size and scanner-specific calibration protocols.
- Translating observed RSP differences into representative water-equivalent thickness (WET) estimates produced 2σ intercenter variability of 1.09% for brain and 1.39% for prostate scenarios. These estimates describe the CT calibration-related component, not total clinical proton range uncertainty. Interpretation is further limited by reliance on one imperfect pCT reference, ex vivo tissue, simplified clinical beam-path calculations, and absence of direct patient-level dosimetric or outcome validation.
CLINICAL TAKEAWAY
CT-based stopping-power calibration remains a meaningful source of variation between proton therapy centers, particularly for beam paths involving bone and larger anatomical regions. These findings support standardized calibration procedures, appropriate beam-hardening correction and independent validation of RSP estimates, while demonstrating that scanner technology alone does not guarantee accurate proton range prediction.