Purpose: Multi-energy CT has been enabled by detector advancements (e.g., photon-counting, multi-layer) and facilitates material decomposition and density estimation of tissues and contrast agents. We propose a novel source-side innovation called spatial-spectral filters comprised of tiled k-edge metals. This is a potentially inexpensive modification to existing systems without the need for energy-discriminating detectors. In this work we simulate and characterize design parameters associated with spatial-spectral filters to make informed choices for hardware design and implementation.
Methods: The impact of focal spot size and gap width between filter tiles on imaging performance was investigated. Simulation studies used a numerical phantom with cylindrical contrast inserts of iodine, gadolinium, and gold (nanoparticles) with concentrations of 0.1-3.2~mg/mL; and a spatial-spectral filter with lead, gold, lutetium, and erbium tiles (thicknesses 0.10-0.25~mm). Focal spot widths between 0.0-0.5~mm and gap widths between 0-50~microns were simulated. Scenarios where the data generation model was matched and mismatched with the model-based material decomposition (MBMD) were performed to characterize the sensitivity to model accuracy. Root-Mean-Squared Error (RMSE) of the insert concentration estimates was computed as a function of model mismatch.
Results: For model-matched data, a 0.5~mm focal spot width and 50~micron gap width increased RMSE less than 1%. For the model-mismatched case, RMSE rose dramatically for the for both gap width and focal spot width.
Conclusion: Model-mismatch can significantly impact spatial-spectral filter imaging performance, and has the potential to enforce strict fabrication requirements. However, we find that when modeled, nonidealities like an extended focal spot and manufacturing gaps have a small effect on performance. This suggests that with proper calibration, spatial-spectral filters can achieve accurate density estimates in physical systems. This work will guide ongoing hardware design and implementation of multi-energy CT using spatial-spectral filters.
Funding Support, Disclosures, and Conflict of Interest: This work was supported, in part, by NIH grant R21EB026849.
Not Applicable / None Entered.