Accuracy and reproducibility of effective atomic number and electron density measurements from sequential dual energy CT.

Purpose: This study assesses the accuracy of effective atomic number (Z ) and electron density measurements acquired from dual energy CT and characterizes the response to clinically relevant variables representative of challenges in patient imaging, including: phantom size, material position within the phantom, variation over time, off-center positioning, and large cone beam angle.

Methods: The Gammex Multi-Energy CT head and body phantoms were used to measure Z and electron density from 35 rod inserts that mimic tissues and varying concentrations of iodine and calcium. Scans were performed on a Canon Aquilion ONE Genesis CT scanner over a period of 6 months using default dual energy protocols appropriate for each phantom size. Theoretical Z and electron density values were calculated using data provided by the phantom manufacturer and compared to the measurements. Sources of variance were separated and quantified to identify the influences of random photon statistics, ROI placement, and variation over time. A subset of measurements were repeated with the phantom shifted in the vertical and horizontal directions, and over all slices in the volumetric scan.

Results: All measurements showed strong correlation (r > 0.98) with their corresponding theoretical values; however, the system did demonstrate a bias of -0.58 atomic units in the body phantom and 0.28 atomic units in the head phantom for Z measurements. The mean absolute percent error (MAPE) was 6.3% for the body phantom and 3.2% for the head phantom. Electron density measurements of the body and head phantoms gave MAPE values of 4.6% and 1.0%, respectively. Z and electron density measurements significantly varied within the solid water background, showing a positional dependence within the phantom that dominated the total standard deviation in measurements. Z values dropped by 0.2 atomic units when the phantom was off-center; electron density measurements were less affected by phantom position. Along the z-axis, the accuracy drops off markedly at more than 50-60 mm from the central slice.

Conclusion: The Canon dual energy system offers an accurate way of measuring the Z and electron density of clinically relevant materials. Accuracy could be improved further by calibration to remove bias, careful attention to centering within the FOV, and avoiding measurements at the edges of the cone beam.