Although great efforts had been made to improve the physical phantoms used for calibrating in vivo measurement systems, for technical reasons they can oniy provide a rough representation of human tissue. Substantial corrections m w t therefore be made to calibration factors obtained with such caiibration phantoms for extrapolation to a given individuai. These corrections are particularly crucial and delicate in low-energy in vivo measurement when absorption in tissue is significant. To improve caiibration for such special conditions, the posîibility has been raised of using voxelised numerical phantoms associated with Monte Carlo computing techniques. In the method described below, a mathematical phantom, consisting of a voxelised representation derived from scanner images is used, with a specially-designed interface making it possible to not only reconstruct widely-differing contamination confgurations and specify associated tissue compositions, but also automatically create an MCNP4b input file. After validation of the different sources and geometries, the complete procedure of reconstruction of the phantom and simulation of "'Am lung measurement was carried out using a tissue equivalent calibration phantom of the type commoniy used for lung calibration for actinides. The purpose of this work was to extend the use of this principle to the reconstruction of numerical phantoms on the basis of physiological data of individuak obtained from maguetic resonance and scanner images. The resulîs obtained and the current limitations of this approach in the context are discussed. Développement de fantômes numériques voxélisés associé au code Monte Carlo MCNP : application à la mesure anthroporadiamétrique.