Experimental Benchmarking and Validation of a Clinical Photon-Counting CT Simulation
Abstract
Purpose
To develop and validate the first open-source, CdTe-based photon-counting CT (PCCT) simulation framework using CATsim. The goal was to emulate clinical PCCT performance and quantitatively compare simulated versus experimental iodine and calcium Hounsfield units (HU) across varying concentrations to identify sources of systematic bias.
Methods
The framework was configured to replicate a clinical scanner (Siemens NAEOTOM Alpha) using specific parameters: CdTe detector, coverage, focal spot size, tin filtration, and a bowtie filter. Simulations at 120 and 140 kVp utilized spectral modeling, scatter estimation, and a simplified Beer-Lambert detector efficiency model. Experimental data were acquired at 120 and 140kVp using a Gammex 472 phantom with iodine (2–20 mg/mL) and calcium (50–400 mg/mL) inserts. Agreement between experiment and simulation results was assessed via linear regression and Lin’s Concordance Correlation Coefficient (CCC).
Results
Comparison via grouped bar charts and linear regression revealed a linear systematic underestimation of simulations relative to experimental results. High linearity (R2 > 0.99 for both materials at both voltages) confirms this observation. Lin’s CCC was 0.93 for iodine (moderate agreement) and 0.89 for calcium (poor agreement) per the McBride scale.
Conclusion
This study implements an open-source CATsim model for clinical PCCT benchmarking. The linear systematic underestimation suggests that analytical simplifications in the detector efficiency model are primary error drivers. Future refinements will employ Monte Carlo techniques to better capture the complex spectral response of CdTe detectors.