Photon-Counting Computed Tomography (PCCT) for Radiation Therapy

Photon-counting computed tomography (PCCT), which provides CT data with intrinsic spectral information, is a topic of increasing interest in radiation therapy. Early studies with prototype PCCT scanners have shown similar or slightly better relative stopping power (RSP) accuracy compared to methods based on dual-energy CT (DECT). The first-generation clinical PCCT scanners have been recently introduced. Their performance for evaluation of quantitative information for radiation therapy treatment planning such as relative electron density (RED), effective atomic number (EAN), and RSP has not been investigated yet. We aimed at assessment of RED, EAN, and RSP relying on the virtual monoenergetic images (VMIs) provided by a first-generation clinical PCCT scanner. Two customised cylindrical PMMA phantoms, housing tissue-equivalent inserts, with 15 cm and 13 cm diameter for calibration and evaluation, respectively, were scanned at a clinical PCCT scanner (NAEOTOM Alpha, Siemens Healthineers, Forchheim, Germany) at a tube voltage of 120 kVp with 40 mGy CTDI \({ }_{\mathrm {vol}}\) . VMIs with 1 mm slice thickness and 0.38 \(\times \) 0.38 mm2 pixel size were reconstructed at energies ranging from 40 keV to 190 keV in 10 keV increments. We exploited various combinations of a low-energy and a high-energy VMI to convert measured CT numbers into RED, EAN, and RSP using Saito and Sagara’s approach. The mean estimated values of RED, EAN, and RSP for the inserts were compared to reference values obtained from either vendor information (RED & EAN) or water column measurements (RSP). Root-mean-squared errors (RMSEs) were calculated over phantom inserts. The RMSEs for RED estimation mostly (102 out of 105 VMI pairs) ranged between 1.51% and 1.54% for the calibration phantom and 0.55% and 0.56% for the evaluation phantom. The optimal pair of VMIs, with 60 keV as the low energy and 180 keV as the high energy, showed the overall best accuracy for EAN and RSP. Comparable accuracy of RSP, 1.27% (0.54% when excluding the porous lung inserts) and 0.71% for the calibration and evaluation phantoms, respectively, to that reported in previous DECT studies was demonstrated. In conclusion, we performed tissue-equivalent material measurements at a first-generation clinical PCCT scanner and evaluated the accuracy of RED, EAN, and RSP estimation. Comparable RSP accuracy to DECT was obtained. The results presented were obtained in the first year of operation of the scanner. We are currently analysing data obtained with subsequent measurements to further compare results of not only tissue-equivalent inserts but also biological tissue between PCCT and DECT in terms of RSP accuracy and spatial resolution.