Purpose: To assess the improvement of diagnostic value and imaging performance of an edge-on silicon photon-counting CT (Si-PCCT) compared to energy-integrating CT (ECT) across various clinical scenarios in a virtual imaging trial.
Methods: We developed a CT simulation platform to model ECT and Si-PCCT. We used a phantom to validate simulations against experimental images acquired with clinical ECT and an investigational Si-PCCT. The phantom had multiple materials of soft-tissue, bone, a mixture of calcium and iodine, and various concentrations of calcium and iodine. The simulation platform was then used to evaluate the imaging performance of Si-PCCT against ECT in terms of task-generic metrics (RMSE, MTF and NPS). Computational human models with lung and liver lesions were imaged virtually under various imaging conditions (radiation dose levels, slice thickness, and reconstruction kernels). We performed radiomic analysis on lung and liver lesions to evaluate the task-specific imaging performance of ECT and Si-PCCT. Simulated pairs of images from ECT and Si-PCCT across all human models were scored (with z-scores) by radiologists based on anatomical conspicuity and image quality features to assess the diagnostic value of Si-PCCT.
Results: Simulation results closely matched to those obtained experimentally from ECT and Si-PCCT. Examination of the task-generic matrices indicated that Si-PCCT outperformed ECT for quantitative imaging. Radiomics analysis showed that Si-PCCT images had lower estimation errors. The observer study showed statistically significant improvement of Si-PCCT over ECT for visualization of lesions in the lung (0.08 ± 0.89 vs. 0.90 ± 0.48) and liver (-0.64 ± 0.37 vs. 0.95 ± 0.55).
Conclusion: Our studies elucidate the enhancement of diagnostic value and quantitative imaging performance of an edge-on Si-PCCT compared to ECT across various clinical scenarios and imaging conditions. This enhancement demonstrates the promise of Si-PCCT for a wide clinical scenario through improved spatial, spectral and temporal resolution, and noise.
Funding Support, Disclosures, and Conflict of Interest: Funding acknowledgement: Research reported in this study was supported by grants from GE Healthcare and National Institutes of Health under award number R01EB001838 and P41EB028744. Disclosure: Zhye Yin is an employee of GE Healthcare.