Monte Carlo modelling of an extended DXA technique.

The precision achieved in measuring bone mineral density (BMD) by commercial dual-energy x-ray absorptiometry (DXA) machines is typically better than 1%, but accuracy is considerably worse. Errors, due to inhomogeneous distributions of fat, of up to 10% have been reported. These errors arise because the DXA technique assumes a two-component model for the human body, i.e. bone mineral and soft tissue. This paper describes an extended DXA technique that uses a three-component model of human tissue and significantly reduces errors due to inhomogeneous fat distribution. In addition to two x-ray transmission measurements, a measurement of the path length of the x-ray beam within the patient is required. This provides a third equation, i.e. T = ts + tb + tf where T, ts, tb and tf are the total, lean soft tissue, bone mineral and fatty tissue thicknesses respectively. Monte Carlo modelling was undertaken to make a comparison of the standard and extended DXA techniques in the presence of inhomogeneous fat distribution. Two geometries of varying complexity were simulated. In each case the extended DXA technique produced BMD measurements that were independent of soft tissue composition whereas the standard technique produced BMD measurements that were strongly dependent on soft tissue composition. For example, in one case, the gradients of the plots of BMD versus fractional fat content were for standard DXA (-0.183+/-0.037) g cm(-2) and for extended DXA (0.027+/-0.044) g cm(-2). In all cases the extended DXA method produced more accurate but less precise results than the standard DXA technique.