Premium
Quadratic relation for mass density calibration in human body using dual‐energy CT data
Author(s) -
Saito Masatoshi
Publication year - 2021
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1002/mp.14899
Subject(s) - imaging phantom , effective atomic number , calibration , range (aeronautics) , atomic number , reference data , attenuation , digital enhanced cordless telecommunications , quadratic equation , physics , computational physics , nuclear medicine , mathematics , atomic physics , optics , statistics , geometry , materials science , computer science , medicine , telecommunications , database , wireless , composite material
Purpose To derive the mass density ( ρ ) from dual‐energy computed tomography (DECT) data by calibrating electron density ( ρ e ) and effective atomic numbers ( Z eff ) of human tissues. Methods We propose the DEEDZ‐MD method, in which a single polynomial parameterization covers the entire human‐tissue range to establish an empirical quadratic relation between the atomic number‐to‐mass ratio and Z eff . Then, we numerically evaluate the DEEDZ‐MD method in reference human tissues listed in the ICRP Publication 110 and ICRU Report 46. The tissues are considered to have unknown ρ values. The attenuation coefficients of these tissues are calculated using the XCOM Photon Cross Sections Database. The DEEDZ‐MD method is also applied to experimental DECT data acquired from a tissue characterization phantom and an anthropomorphic phantom at 90 kV and 150 kV/Sn. Results The numerical analysis of the DEEDZ‐MD method reveals a single quadratic relation between the atomic number‐to‐mass ratio and Z eff in a wide range of human tissues. The simulated ρ values are in excellent agreement with the reference values over ρ values from 0.260 (lung) to 3.225 (hydroxyapatite). The relative deviations from the reference ρ remain within ±0.6% for all the reference human tissues, except for the eye lens (approximate deviation of −1.0%). The overall root‐mean‐square error is 0.24%. The application of the DEEDZ‐MD method to experimental dual‐energy CT data confirms this agreement within experimental accuracy, indicating the practical feasibility of the method. The DEEDZ‐MD method enables the generation of ρ images with less image noise than the existing DECT‐based conversion of ρ from ρ e and with fewer beam‐hardening artifacts than conventional single‐energy CT images. Conclusions The DEEDZ‐MD method can facilitate the generation of ρ images from dual‐energy CT data without relying on the nontrivial segmentation of different tissues.