Background: Boron neutron capture therapy (BNCT) is a binary radiotherapy combining biochemical targeting with neutron irradiation. However, monitoring the boron distribution is a fundamental problem in BNCT. Prompt gamma rays emitted by boron capture reaction can be used to address the issue. Materials and Methods: The general-purpose Monte Carlo toolkits Geant4 and MCNP were used for the simulations. A cubic phantom with soft tissue was used to study the prompt gamma emission during BNCT. The Chinese hybrid phantom with arbitrary tumors was constructed and used to acquire the 0.478 MeV prompt gamma rays in BNCT. Tomographic images were reconstructed with the maximum likelihood expectation maximization (MLEM) algorithm. Results: Comparison between MCNP and Geant4 showed a similar gamma rays emission rate in soft tissue. Up to 30 gamma ray peaks were found in the simulation, and 0.478 MeV prompt gamma ray from boron was clearly observed. The single brain tumor with variable diameter from 1 cm to 4 cm in the heterogeneous anthropomorphic phantom was each time found to be recognizable in the reconstructed image. Furthermore, in a patient with four tumors, the variable distance between the source and the tumors leads to a neutron attenuation thus resulting in an inhomogeneous number of prompt gammas. Conclusion: The SPECT system for a heterogeneous phantom in BNCT was simulated with Geant4. The results show that BNCT-SPECT is valid for the reconstruction of the boron capture interaction position for a heterogeneous patient.