Simulation studies of optimum energies for DXA: dependence on tissue type, patient size and dose model

Dual-energy x-ray absorptiometry (DXA) is a well established technique for measuring bone mineral density (BMD). However, in recent years DXA is increasingly being used to measure body composition in terms of fat and fat-free mass. DXA scanners must also determine the soft tissue baseline value from soft-tissue-only regions adjacent to bone. The aim of this work is to determine, using computer simulations, the optimum x-ray energies for: a number of dose models, different tissues, i.e. bone mineral, average soft tissue, lean soft tissue and fat; and a range of anatomical sites and patient sizes. Three models for patient dose were evaluated: total beam energy, entrance exposure and absorbed dose calculated by Monte Carlo modelling. A range of tissue compositions and thicknesses were chosen to cover typical patient variations for the three sites: femoral neck, PA spine and lateral spine. In this work, the optimisation of the energies is based on: (1) the uncertainty that arises from the quantum statistical nature of the number of x-rays recorded by the detector, and (2) the radiation dose received by the patient. This study has deliberately not considered other parameters such as detector response, electronic noise, x-ray tube heat load etc, because these are technology dependent parameters, not ones that are inherent to the measuring technique. Optimisation of the energies is achieved by minimisation of the product of variance of density measurement and dose which is independent of the absolute intensities of the x-ray beams. The results obtained indicate that if solving for bone density, then Elow in the range 34 to 42 keV, Ehigh in the range 100 to 200 keV and incident intensity ratio (low energy/high energy) in the range 3 to 10 is a reasonable compromise for the normal range of patient sizes. The choice of energies is complicated by the fact that the DXA unit must also solve for fat and lean soft tissue in soft-tissue-only regions adjacent to the bone. In this case the ranges of energies suggested above are acceptable only for the average phantom. In extreme cases the variance-dose product can be up to 50 times higher than for the ideal energies.