Purpose: Radiologists don't like to review extra images. Leveraging the abilities of photon-counting CT (PCCT) by using energy weighting to form a single set of quantitative greyscale images leverages the unique advantages of PCCT while still providing familiar content and workflow for radiologists, i.e. not requiring them to read additional spectral or material-basis images. We evaluated methods to generate such quantitative greyscale CT images.
Methods: We used CatSim-based simulations of a 120 kVp polychromatic x-ray tube with varying dose coupled with a model of an 8 energy bin PCCT detector within a standard CT gantry. An elliptical water phantom contained inserts (5 cm in diameter) of liver, intestine, and adipose tissue, along with a blood/iodine mixture. The long axis of the phantom was varied from 30 to 50 cm in 2 mm steps. CT acquisitions of the 100 phantoms with independent noise realizations were simulated, and images reconstructed with FBP after applying energy-bin-specific beam-hardening corrections. The 8 energy-bin images were combined using several approaches: uniform, spectral, inverse-spectral, and RMS-optimized weightings. For each case, bias, noise, and RMS error (RMSE) were measured and compared to equivalent results for a standard energy-integrating CT (EICT).
Results: For the EICT images the Hounsfield Unit (HU) values for the 4 inserts varied as phantom size increased – liver: 52 to 38 HU, intestine: 22 to 12 HU, adipose: -244 to -240 HU, and blood/iodine: 260 to 200 HU. However, for the PCCT greyscale images, HU values for all inserts were essentially constant as a function of phantom size for all energy-weighting schemes. PCCT images had generally higher noise but lower RSME compared to EICT.
Conclusion: PCCT can be used to form quantitative greyscale CT images that are invariant to object size and with lower RMSE relative to EICT.
Funding Support, Disclosures, and Conflict of Interest: Supported by a research grant from GE Healthcare
CT, Quantitative Imaging, Energy Spectrum