Program Information
The Effect of Image Noise On Proton Range Estimation Based On Dual-Energy Vs. Photon-Counting-Detector CT
V. Taasti1,2*, G. Michalak2, L. Muren1, J. Petersen1, D. Hansen1, C McCollough2, (1) Aarhus University Hospital, Aarhus, Denmark, (2) Mayo Clinic, Rochester, MN
Presentations
WE-G-605-8 (Wednesday, August 2, 2017) 4:30 PM - 6:00 PM Room: 605
Purpose: Several methods have been proposed for estimating proton stopping power ratios (SPRs) from dual-energy CT (DECT), which has been shown to give more accurate SPR estimates than single-energy CT (SECT). However, unlike SECT, most DECT methods are non-linear, which makes them more sensitive to CT noise. In this study, we investigated the effect of CT noise on proton range errors for different SPR estimation methods. We also developed a SPR method based on photon-counting-detector CT (PCD-CT) that was intended to be robust to noise.
Methods: CT images of the female ICRP phantom at four anatomical locations were calculated theoretically, based on measured energy spectra and detector response for different CT scanner types. Clinically realistic noise values were added to each set of images, sampled from a normal distribution with standard deviation tailored to each scanner type. Different SPR estimation methods were used to calculate SPR images. The range of proton beams of 70 and 115 MeV were estimated based on each set of SPR images to investigate the impact of noise on the range estimation. The noise simulation was redone 50 times and for each simulation the root-mean-square-error was calculated and averaged over all realizations. Results were compared to the noise-free condition.
Results: The presence of noise increased the range errors for some methods by as much as 5 mm. The largest errors were generally found in the lungs. With added noise, the lowest range errors for all slices were found for the two-bin PCD-CT scan type, albeit with different SPR estimation methods.
Conclusion: Generally, CT noise caused uncertainties in the proton range estimation. Further DECT methods will be studied to find the best-suited SPR estimation method in presence of image noise. The proposed PCD-CT method produced good results for the head slice.
Funding Support, Disclosures, and Conflict of Interest: Dr. McCollough received funding from Siemens Healthcare.
Contact Email: