Room: AAPM ePoster Library
In proton therapy, an accurate map of stopping power ratios (SPRs) must be inferred from the patient’s CT scan in order to fully take advantage of proton properties. However, conversion from CT to SPR is a major source of range uncertainty. Hence, calibration of CT scanners must be carefully validated, ideally using materials mimicking human tissue composition. We propose here a validation for several CT calibration methods using liquid organic materials.
A CT scanner was calibrated for a Monte Carlo dose engine using the RayStation, stoichiometric, and continuous stoichiometric methods. The last method allowed for direct interpolation of the materials without using predefined classes. Mixtures were then designed in order to mimic human tissues from the well-established ICRU database regarding three properties: mass density, elemental composition, and water weight fraction. Nine soft tissues and three bones were manufactured using basic materials such as water, amino acids, fatty acids, glucose, calcium compounds, or alcohols. They were placed and scanned inside the Gammex phantom, replacing the original plastic inserts. Finally, SPRs were computed using all three calibrations and compared to theoretical SPRs obtained from the known mixture properties. Predicted elemental fractions were evaluated for soft tissues, that were very close to ICRU tissues.
The SPR relative errors for each method were respectively 1.53±1.40%, 1.37±1.40%, and 1.42±1.29%. The RaySearch method seemed to perform slightly better in bones and fatty tissues, but worse in soft tissues. The continuous stoichiometric method significantly improved the prediction of elemental fractions in comparison with the normal stoichiometric and RayStation methods.
We validated CT calibrations using a set of materials with known properties. These materials facilitate tracking of possible inaccuracies in the calibration curve and could even, in the future, replace plastic materials in the calibration phase itself.
Protons, CT, Quality Assurance
Not Applicable / None Entered.