Potential of dual-energy subtraction for converting CT numbers to electron density based on a single linear...

The conversion of the computed tomography (CT) number to electron density is one of the main processes that determine the accuracy of patient dose calculations in radiotherapy treatment planning. However, the CT number and electron density of tissues cannot be generally interrelated via a simple one-to-one correspondence because the CT number depends on the effective atomic number as well as the electron density. The purpose of this study is to present a simple conversion from the energy-subtracted CT number (ΔHU) by means of dual-energy CT (DECT) to the relative electron density (ρ(e)) via a single linear relationship.

The ΔHU-ρ(e) conversion method was demonstrated by performing analytical DECT image simulations that were intended to imitate a second-generation dual-source CT (DSCT) scanner with an additional tin filtration for the high-kV tube. The ΔHU-ρ(e) calibration line was obtained from the image simulation with a 33 cm-diameter electron density calibration phantom equipped with 16 inserts including polytetrafluoroethylene, polyvinyl chloride, and aluminum; the elemental compositions of these three inserts were quite different to those of body tissues. The ΔHU-ρ(e) conversion method was also applied to previously published experimental CT data, which were measured using two different CT scanners, to validate the clinical feasibility of the present approach. In addition, the effect of object size on ρ(e)-calibrated images was investigated by image simulations using a 25 cm-diameter virtual phantom for two different filtrations: with and without the tin filter for the high-kV tube.

The simulated ΔHU-ρ(e) plot exhibited a predictable linear relationship over a wide range of ρ(e) from 0.00 (air) to 2.35 (aluminum). Resultant values of the coefficient of determination, slope, and intercept of the linear function fitted to the data were close to those of the ideal case. The maximum difference between the ideal and simulated ρ(e) values was -0.7%. The satisfactory linearity of ΔHU-ρ(e) was also confirmed from analyses of the experimental CT data. In the experimental cases, the maximum difference between the nominal and simulated ρ(e) values was found to be 2.5% after two outliers were excluded. When compared with the case without the tin filter, the ΔHU-ρ(e) conversion performed with the tin filter yielded a lower dose and more reliable ρ(e) values that were less affected by the object-size variation.

The ΔHU-ρ(e) calibration line with a simple one-to-one correspondence would facilitate the construction of a well-calibrated ρ(e) image from acquired dual-kV images, and currently, second generation DSCT may be a feasible modality for the clinical use of the ΔHU-ρ(e) conversion method.