Comparative accuracy and resolution assessment of two prototype proton computed tomography scanners

• Published in: Medical physics (Lancaster) Vol. 49; no. 7; pp. 4671 - 4681
• Main Authors: Dedes, George, Drosten, Hubertus, Götz, Stefanie, Dickmann, Jannis, Sarosiek, Christina, Pankuch, Mark, Krah, Nils, Rit, Simon, Bashkirov, Vladimir, Schulte, Reinhard W., Johnson, Robert P., Parodi, Katia, DeJongh, Ethan, Landry, Guillaume
• Format: Journal Article
• Language: English
• Published: United States American Association of Physicists in Medicine 01-07-2022
• Subjects: Computer Science; design choices; image quality; Medical Imaging; proton CT; RSP; spatial resolution; proton CT; image quality; spatial resolution; RSP; design choices
• ISSN: 0094-2405; 2473-4209
• DOI: 10.1002/mp.15657

BackgroundImproving the accuracy of relative stopping power (RSP) in proton therapy may allow reducing range margins. Proton computed tomography (pCT) has been shown to provide state‐of‐the‐art RSP accuracy estimation, and various scanner prototypes have recently been built. The different approaches used in scanner design are expected to impact spatial resolution and RSP accuracy.
PurposeThe goal of this study was to perform the first direct comparison, in terms of spatial resolution and RSP accuracy, of two pCT prototype scanners installed at the same facility and by using the same image reconstruction algorithm.
MethodsA phantom containing cylindrical inserts of known RSP was scanned at the phase‐II pCT prototype of the U.S. pCT collaboration and at the commercially oriented ProtonVDA scanner. Following distance‐driven binning filtered backprojection reconstruction, the radial edge spread function of high‐density inserts was used to estimate the spatial resolution. RSP accuracy was evaluated by the mean absolute percent error (MAPE) over the inserts. No direct imaging dose estimation was possible, which prevented a comparison of the two scanners in terms of RSP noise.
ResultsIn terms of RSP accuracy, both scanners achieved the same MAPE of 0.72% when excluding the porous sinus insert from the evaluation. The ProtonVDA scanner reached a better overall MAPE when all inserts and the body of the phantom were accounted for (0.81%), compared to the phase‐II scanner (1.14%). The spatial resolution with the phase‐II scanner was found to be 0.61 lp/mm, while for the ProtonVDA scanner somewhat lower at 0.46 lp/mm.
ConclusionsThe comparison between two prototype pCT scanners operated in the same clinical facility showed that they both fulfill the requirement of an RSP accuracy of about 1%. Their spatial resolution performance reflects the different design choices of either a scanner with full tracking capabilities (phase‐II) or of a more compact tracker system, which only provides the positions of protons but not their directions (ProtonVDA).