Material Quantification Accuracy of Calcium and Iodine Using Deep Silicon Photon Counting CT
Abstract
Purpose
Deep silicon photon counting CT can produce quantitative CT images with every scan, leveraging 8 energy bins to quantify material densities. This study evaluates the material decomposition (MD) and virtual monoenergetic image (VMI) accuracy of calcium and iodine.
Methods
We scanned the Gammex Multi-Energy CT body phantom with HE Calcium 0, 5, 10, 20, 50, 100, 200, 300 mg/ml and Iodine 0.2, 0.5, 1, 2, 5, 10 mg/ml inserts on deep silicon PCCT technology (GE HealthCare). Axial scans were performed at 120 kVp, Large bowtie, and 821 mAs, as well as a repeat scan and at approximately half dose (426 mAs). Deep learning image reconstruction (DLIR) was performed for MD (Calcium/Water and Iodine/Water) and VMI at 70 keV with a Standard kernel and 2.5 mm slices. ROIs were placed on the inserts to measure mean Calcium and Iodine densities, as well as pixel-level noise, and were compared with theoretical values based on elemental composition provided by Gammex.
Results
We found a strong linear relationship between measured and theoretical values for Calcium densities (slope=0.937, offset=2.07, R2=0.9999) and VMI (slope=0.975, offset=6.08, R2=0.9999). Absolute and relative root-mean-squared-errors for Calcium density were 9.00 mg/ml and 4.47% (for HE Calcium ≥5), with pixel-level noise of 4.13 mg/ml. Consistency between repeat scans and at half dose were 0.42 and 0.88 mg/ml, respectively, in absolute RMSE and 2.70% and 2.73% in relative RMSE. The agreement between measured and theoretical values was similarly close for Iodine densities (slope=0.980, offset=0.16, R2=0.9998) and VMI (slope=0.976, offset=4.18, R2=0.9997), with absolute and relative errors for Iodine density of 0.12 mg/ml and 0.44% (for Iodine ≥5), and noise of 0.30 mg/ml.
Conclusion
Deep silicon PCCT yields strong agreement between measured and theoretical densities of Calcium and Iodine, with demonstrated repeatability and consistency across dose, supporting reliable quantitative spectral imaging.