Proton Range Verification with PET Imaging in Brain and Head and Neck Cancers

Objectives: Uncertainty in proton range in tissue is a limiting factor in accurate and optimal planning for proton therapy[1]. Positron emission tomography (PET) has the potential to verify range by imaging the positron emitters generated by nuclear reactions of protons with elements in the tissue[2]-[5]. Images can be taken shortly after any given treatment fraction and compared to treatment plan-based Monte Carlo (MC) simulations of the PET activity to check beam delivery. Accurate MC simulations, including biological effects, and appropriate comparison methods are necessary for appropriate assessment of dose accuracy. In this study we compared two MC simulation schemes to our measured PET images and are reporting our most recent clinical study results. Methods: We scanned 15 patients undergoing proton therapy for brain and head and neck tumors. All patients were imaged with a brain PET/CT scanner within 3.5 minutes of treatment with 2 Gy of protons from one or two beams. Most patients had repeated scans to check to reproducibility, resulting in 40 scans total. MC simulations of the expected PET distribution were generated using TOPAS, a Geant4 based particle simulator[6]. Radioactive decay and biological washout were applied to the MC to match the timing of the PET scans. Adjustments were made to the simulations and decay scheme to account for differences between cerebral spinal fluid (CSF) in the head and the whole-body parameters typically used in the simulations. CSF has a similar composition to water, thus ventricles and resected regions were given material properties similar to water, rather than the default values with a greater carbon contribution[7]. Additionally, less biological washout was applied to CSF regions, given the slower flow than in the highly perfused gray and white matter. Reconstructed images were compared to the both the “original MC” and the CSF “modified MC” for accuracy. Profiles along the beam were generated and compared by shifting one profile relative to the other until the difference is minimized[8]. The shift distance is used in generating a map across the beam to designate areas of agreement or disagreement in range. Results: Overall, the agreement between PET and MC-PET images was within a few millimeters. The average range difference for all patients and beams ranged from -1.85 to 2.77 mm with an average of 0.48 mm for the modified MC, and -1.87 to 3.83 mm with an average of 1.10 mm for the original MC. The RMSE of the differences ranged from 1.24 to 4.67 mm with mean 3.13 mm for the modified MC, and 1.92 to 6.07 mm with mean 3.43 mm for the original MC. While use of the modified MC resulted in smaller range differences, the majority of range differences detected by the automatic algorithm were due to remaining differences in the actual and predicted washout, and thus are not true range errors. A graphical user interface displaying the PET and MC, the 2D shift map, and profiles of selected points on the shift map was used to spot check the results. See Figure 2 for an example range comparison and 2D range map. Conclusions: Despite the often-complicated geometry and the high washout, it is possible to use profiles to compare measured and expected range differences of proton treatments in the head. We are able to provide an assessment of range differences between PET and MC-PET, and thus planned and delivered dose, across the entire beam cross section. The modified MC parameters for CSF regions made the profile shape of the MC more realistic and improved the comparison with PET. Results of the range comparisons can be used on a patient/fraction basis to check individual fractions, or population basis to determine the overall accuracy achievable for a given beam/anatomical location combination. Increasing the PET signal by reducing the delay between treatment and imaging has the potential to greatly improve the results. In-beam PET systems could provide higher imaging quality for even more accurate comparisons.