Spectral computed tomography (CT) using novel X-ray photon counting detectors (PCDs) with energy resolving capabilities is capable of providing energy-selective images. This extra energy information may allow materials such as iodine and calcium, or water and fat to be distinguished. PCDs have energy thresholds, enabling the classification of photons into multiple energy bins. The information content of spectral CT images depends on how the photons are grouped together. In this work, a method is presented to optimise energy windows for maximum material discrimination.
Given a combination of thicknesses, the reference number of expected photons in each energy bin is computed using the Beer-Lambert equation. A similar calculation is performed for an exhaustive range of thicknesses and the number of photons in each case is compared to the reference, allowing a statistical map of the uncertainty in thickness parameters to be constructed. The 63%-confidence region in the two-dimensional thickness space is a representation of how optimal the bins are for material separation.
The model is demonstrated with 0.1mm of iodine and 2.2mm of calcium using two adjacent bins encompassing the entire energy range. Bins bordering at the iodine k-edge of 33.2keV are found to be optimal. When compared to two abutted energy bins with equal incident counts as used in the literature (bordering at 54keV), the thickness uncertainties are reduced from approximately 4% to less than 1% (see figure).
This approach has been developed for two materials and is expandable to an arbitrary number of materials and bins.