Effects on Image Quality
We have now developed a sound understanding of the consequences of losing shift-invariance, and so it is time to turn to inspecting actual images. We will do so using the important example of spectrally resolved CT scans of a phantom containing solutions of iodine and gadolinium. Exhibiting К-edges at 33 and 50 keV, these contrast agents represent a well-accepted standard to study the spectral imaging properties of a photon counting detector.
12.3.1 Energy Resolved CT Images
A CT scan of a phantom containing iodine and gadolinium is shown in Figure 12.9. It was acquired using a Medipix3RX equipped with a 2-mm CdTe sensor and operated in both SPM (top row) as well as CSM (bottom row). The tube voltage was set to 120 kV to provide a broad spectrum of x-ray energies.
FIGURE 12.9 Left: Phantom made from PMMA, containing iodine and gadolinium contrast agents at two concentrations. Right: Slices extracted from two spectral CT scans of this phantom . using energy thresholds ranging from 28 keV to 88 keV. No further binning into energy windows was performed, that is, there was no upper bound on photon energies in each of the four channels. SPM: single pixel mode: CSM: charge summing mode.
The probably most striking difference between the two modes of operation is represented by the noise visible for the energy threshold of 88 keV. Here, the reduced DQE in SPM, as observed in Figure I2.8a, leads to a substantial noise increase, compared to the CSM scan.
The next feature that differentiates the two scans is the visibility of the gadolinium K-edge. It manifests itself in the CSM measurement (bottom capillary) when raising the threshold from 28 keV to 48 keV, but not in the SPM scan. Last but not least, the contrast of the capillaries with regard to the PMMA background is strongly increased when activating CSM, owing to the better spectroscopic resolution that is a defining characteristic of this mode.
To summarize, both noise and contrast are improved when shift-invariance is restored by employing charge summing. The first effect is due to an improved DQE, the second one because of a much better spectral resolution. In combination, both lead to a twofold improvement in contrast-to-noise ratio  (CNR). The CNR is known to be proportional to the square of the radiation dose. Consequently, charge summing allows us to lower x-ray exposure by a factor of four to arrive at the same image quality as we obtained in SPM. This reduction of radiation dose is substantial.
Ultimately, we are now ready to find out how this translates to the accuracy of material decomposition, the probably most exciting application of multi-energy CT.