The impact of uncertainties in the CT conversion algorithm when predicting proton beam ranges in patients from dose and PET-activity distributions

The advantages of a finite range of proton beams can only be partly exploited in radiation therapy unless the range can be predicted in patient anatomy with <2 mm accuracy (for non-moving targets). Monte Carlo dose calculation aims at 1-2 mm accuracy in dose prediction, and proton-induced PET imaging aims at similar to 2 mm accuracy in range verification. The latter is done using Monte Carlo predicted PET images. Monte Carlo methods are based on CT images to describe patient anatomy. The dose calculation algorithm and the CT resolution/artifacts might affect dose calculation accuracy. Additionally, when using Monte Carlo for PET range verification, the biological decay model and the cross sections for positron emitter production affect predicted PET images. The goal of this work is to study the effect of uncertainties in the CT conversion on the proton beam range predicted by Monte Carlo dose calculations and proton-induced PET signals. Conversion schemes to assign density and elemental composition based on a CT image of the patient define a unique Hounsfield unit (HU) to tissue parameters relationship. Uncertainties are introduced because there is no unique relationship between HU and tissue parameters. In this work, different conversion schemes based on a stoichiometric calibration method as well as different numbers of tissue bins were considered in three head and neck patients. For Monte Carlo dose calculation, the results show close to zero (<0.5 mm) differences in range using different conversion schemes. Further, a reduction of the number of bins used to define individual tissues down to 13 did not affect the accuracy. In the case of simulated PET images we found a more pronounced sensitivity on the CT conversion scheme with a mean fall-off position variation of about 1 mm. We conclude that proton dose distributions based on Monte Carlo calculation are only slightly affected by the uncertainty on density and elemental composition introduced by unique assignment to each HU if a stoichiometric calibration is used. Calculated PET images used for range verification are more sensitive to conversion uncertainties causing an intrinsic limitation due to CT conversion alone of at least 1 mm.