Source identification in bioluminescence tomography by consensus-based optimization.

Abstract

A consensus-based optimization (CBO) algorithm, which enables derivative and mesh-free global optimization, is presented to localize a bioluminescent source. The light propagation is modeled by the radiative transfer equation. Since CBO can be easily combined with different forward models, it is investigated for a hierarchy of simplified diffusion models; the state-of-the-art diffusive approximation, <i>SP</i>₃, and an adaptive approach combining the model hierarchy with the CBO dynamics. The latter approach allows rendering higher-order models at lower computational cost than with a fixed model. All models ensure an accurate localization of the bioluminescent sources in simulated environments and tissue-mimicking phantoms. While the diffusive approximation gives reliable results for heavily scattering media, higher-order models achieve slightly better localizations in strongly absorbing, but only moderately scattering media.