Room: Exhibit Hall | Forum 9
Purpose: Material decomposition using dual-energy CT has been successfully applied to the quantification of iodine and, more recently, iron concentrations. A potential, and scarcely discussed, bias that can occur in such algorithms is a shift in the CT number caused by the photon starvation effect. Here we investigate the potential impact of this effect.
Methods: 18 vials were prepared containing liver tissue equivalent solution, and dosed with ferric nitrate from 0-20 mg/ml iron concentration. The vials were inserted into a water phantom representing a medium-sized patient. CT scans were performed on a dual-source scanner at 80kVp/140kVp(Sn) at 6 dose levels ranging from 1-32 mGy of volumetric CTDI with 5 repetitions. One 5-mm thick slice was selected for data analysis. A three-material-decomposition algorithm was used to calculate the iron content and fat fraction.
Results: Significant increase in CT number at 80kVp was observed with decreasing doses. Compared to 32mGy acquisition, the increases were 1.5, 3.0, 15.6, 61.6 and 121 HU at 16, 8, 4, 2 and 1 mGy respectively, which led to an overestimation of iron content by 1.8%, 4.4%, 38%, 69% and 649% respectively. An overestimation of fat fraction at lower doses was also observed. Applying iterative reconstruction or smoother kernels did little to alleviate such effects.
Conclusion: Significant bias in CT number can be caused by photon starvation at low kVp, leading to an overestimation of iron content when a three-material-decomposition algorithm is applied. This effect is likely caused by the curvature of logarithm transform used in CT reconstruction. While traditional denoising methods such as iterative reconstruction help reduce the streaking appearance, they do not correct the shift in CT numbers. Therefore, for quantitative study in large patients, it is paramount to avoid photon starvation by using sufficient dose through increased mAs. Alternatively, 100kVp might be advisable over 80kVp.
Funding Support, Disclosures, and Conflict of Interest: This study was in part supported by Siemens Medical Solutions USA, Inc.