Magnetic nanoparticles (MNP) have promising applications in biomedicine for therapeutic and diagnostic purposes. To optimally alleviate their possibilities in the human body, non-invasive imaging is required to obtain knowledge on their spatial distribution. In this paper a model is developed that allows to optimize a setup for tomographic imaging of MNPs in order to improve their spatial reconstructions, without the need of extensive calibration measurements. The advantage of this model is that setup design parameters can be taken into account and their impact on the stability and accuracy of the reconstruction can be estimated a priori. Additionally, sources of imaging errors, such as misplaced sensors, can be assessed. Finally, an existing particle imager is optimized so as to improve its imaging capabilities using numerical constrained optimization tools. At realistic noise conditions, an increase in reconstruction quality from 20% to 90% is observed after optimization of the standard setup.