Purpose: Edge-on-irradiated silicon detectors are being developed for clinical photon-counting CT (PCCT). While silicon detectors have desirable characteristics such as a narrow photopeak, low charge-sharing and no relevant K-fluorescence, many photons undergo Compton scatter (CS) due to the low atomic number of silicon. We investigate the influence of CS on density and spectral imaging tasks.
Methods: We simulate CT imaging using a previously-published edge-on-irradiated silicon detector model that accurately captures the effects of CS and charge sharing. Both tasks are performed assuming a 120 kVp source with 30 cm water filtration. We vary the lower threshold of the lowest energy bin. Dose efficiency is computed using the Cramér-Rao lower bound in the projection domain and the ensemble variance in the image domain.
Results: The density and spectral task dose efficiency improves in both the projection and image domains as the threshold is lowered to include more CS. For a detector with 1-keV-wide energy bins, projection domain dose efficiency for density and spectral imaging improves by 69% and 84%, respectively by including all Compton events, compared to a baseline threshold of 30 keV (where most Compton events are excluded). Corresponding improvements in the image domain for a detector with 8 energy bins and a lowest threshold of 5 keV are 58% and 89%, respectively. Dose efficiency in image and projection domains agree to within 4%.
Conclusion: Including Compton events in edge-on-irradiated silicon PCCT improves density and spectral dose efficiency in both projection and image domains, indicating that these events carry useful information for both density and spectral imaging tasks. This study underlines the importance of including CS in any assessment of task performance using low-Z detector materials such as silicon. Comparison of such detectors with high-Z detectors at equal lowest energy thresholds will severely underestimate the performance of edge-on-irradiated silicon detectors.
Funding Support, Disclosures, and Conflict of Interest: Mats U. Persson discloses past financial interests in Prismatic Sensors AB and research collaboration with GE Healthcare. This study has received funding from MedTechLabs.