Comparison of tissue characterization curves for different CT scanners: implication in proton therapy treatment planning

For proton beam therapy, CT imaging is required to calculate dose based on CT pixel values of relative stopping power (RSP). The variation among CT-scanners and the phantom used to derive the relationship CT number-electron density-RSP has not been investigated. Using RMI CT phantoms, 18 CT scanners and a Tomotherapy unit, the Hounsfield unit (HU) variations and associated dosimetric uncertainty were investigated. The variation of HU was within one standard deviation (SD) of the average for 14 out of the 15 scanners tested with the same phantoms. For high density materials (>400 HU) the HU values deviated by more than 4% from the average. The HU-RSP curves of 18 scanners were fitted with a straight line in three HU intervals: -700<HU<0, 0<HU<230 and <230<HU<1,700. The dosimetric impacts of the variation of HU among scanners were <1% in DVH point dose comparisons in a prostate cancer plan. For a head and neck cancer plan the difference was up to 4% due to large inhomogeneities. Our results seem to suggest that the X-ray spectrum of a CT scanner has a smaller effect on the HU-RSP curve than the elemental compositions of the tissue substitutes used in the calibration. Furthermore, a single HU-RSP curve may suffice as a reference curve for proton treatment planning. It is found to agree with the calibration curve obtained for the specific scanner. This study does not address the issue of metallic implants included in the treatment plan.