Feasibility study for DEXA using synchrotron CT at 20-35 keV

• Published in: Physics in medicine & biology Vol. 58; no. 4; pp. 1185 - 1205
• Main Author: Midgley, S M
• Format: Journal Article
• Language: English
• Published: England IOP Publishing 21.02.2013
• Subjects: Absorptiometry, Photon - instrumentation; Absorptiometry, Photon - methods; Algorithms; Feasibility Studies; Humans; Models, Statistical; Phantoms, Imaging; Photons; Reproducibility of Results; Scattering, Radiation; Synchrotrons; Tomography, X-Ray Computed - methods; X-Rays
• ISSN: 0031-9155; 1361-6560
• DOI: 10.1088/0031-9155/58/4/1185

A nonlinear model for the x-ray linear attenuation coefficient μ is employed for dual energy x-ray analysis (DEXA). Nonlinear simultaneous equations formed by μ and energy dependent model parameters are solved for the electron density Ne and fourth compositional ratio R4 which has the same 'units' as the atomic number. Computed tomography data was acquired at 20-35 keV using bending magnet synchrotron radiation, a double crystal monochromator, a rotation stage and an area detector. Test objects contained liquid samples as mixtures of ethanol, water and salt solutions with known density and composition. Various noise sources are identified and give μ uncertainties of 1-2%. A fan beam geometry allowed the detection of forward scattered radiation with measured μ being 6% lower than expectations for a narrow beam. Energy dependent model parameters were obtained by solving linear simultaneous equations formed by μ and material parameters based upon Ne and R4. DEXA accuracy was studied as a function of photon energy and sample composition. Propagation of errors analysis identifies the importance of the fractional compositional cross-products whose difference at the two beam energies should exceed 0.1, requiring 10 keV or more separation. For a reasonable approximation for the adjustable model parameters, the mean difference between the DEXA solution and true values (ΔNe, ΔR4) are (1.0%, 0.5%) for soft tissue and (1.5%, 0.8%) for bone like samples.