- MeV using three different Monte Carlo codes; EGS41, MCNP2, and PHOTON3. These codes differ in their input files and transport calculations, and were used to verify the internal consistency of the set-up and of the input data. The energy range of 0.662-10 MeV was chosen to cover energies of interest in body-composition studies. The superior efficiency of the BGO detectors has to be weighed against their inferior resolution, and their higher price than that of the NaI detectors. Since the price of the BGO detectors strongly depends on the crystal's size, its optimization is an important component in the design of the entire system. METHOD The intrinsic response functions were calculated for a point source at 10 cm from the base of a cylindrical detector, where the source emission was forced into the solid angle subtended by the detector. The energy deposition in the detector, i.e., the photon spectrum, was calculated using 512 bins 20 keV wide. Special attention was paid to the energy cut-off used in the codes. While this value is fixed at 10 keV in the PHOTON code, to improve accuracy we altered it in the other two codes to 1 keV. The main effect of the cut-off value is in the valley area beside the photopeak (Fig 1a.). The calculated photon spectra were convoluted with an energy-resolution function of the type a+bE½+cE that was fitted to the experimental data for each detector reported in the literature4. The convolution consisted of Monte Carlo normal sampling of each energy bin. We found that from about 2 MeV and above, the results from the PHOTON code deviated systematically from the results obtained using the other two codes (Figs. 1a, 1b, and 1c). We attributed this discrepancy to inadequate bremsstrahlung cross-sections used in the PHOTON code (we are verifying this with the authors of the code). The Photon code was not used from further at high energies.