Quantitative Enhancement of Spectral Imaging in Photon-Counting CT via Dynamic Energy Thresholding

Photon-counting computed tomography (PCCT) offers several advantages over conventional spectral systems including improved material separation, reduction of artifacts, and better quantitation. However, imaging performance largely depends on the spectral separation set by energy thresholds, and optimal thresholding can vary by the particular diagnostic task as well as patient anatomy. In most commercial photon-counting detectors (PCDs), the energy thresholds are predefined and fixed, and the capability for dynamic threshold modulation during a scan remains underexplored. In this work, we propose to determine optimal dynamic thresholds that maximize spectral imaging performance given knowledge of patient anatomy from a prior scan or a low-dose 3D scout. Through two-bin PCD simulations under various conditions, we investigated changes in optimal thresholds and assessed potential improvements in dual-energy material decomposition. The simulations include single-pixel decomposition across varying material thicknesses and total photon counts after attenuation, as well as the decomposition of a digital pelvic phantom, which has iodine enhancement in the prostatic arteries. Results demonstrate that the optimal energy threshold changes from 35 to 75 keV in water-calcium decomposition; replacing a fixed threshold with the optimal threshold yields RMSE reductions over 30%, especially in low-count regimes (total photon counts < 100); and pixel-wise threshold modulation in pelvic CT simulations provides 6.9% and 17.2% RMSE reductions for water and iodine, respectively, with reduced streak artifacts. These findings show that dynamic threshold modulation can offer substantial improvements in spectral imaging accuracy for low-dose protocols and challenging anatomical regions when integrated into PCCT systems.