Source identification in bioluminescence tomography by consensus-based optimization.

Area of Science:

• Biomedical Optics
• Computational Physics
• Medical Imaging

Background:

• Accurate localization of bioluminescent sources is crucial for in vivo imaging.
• Existing methods often rely on simplified light propagation models.
• Derivative-free global optimization offers a flexible approach for inverse problems.

Purpose of the Study:

• To present and evaluate a consensus-based optimization (CBO) algorithm for bioluminescent source localization.
• To investigate the performance of CBO with a hierarchy of light propagation models, including diffusion approximations.
• To explore an adaptive approach combining model hierarchy with CBO dynamics for improved computational efficiency.

Main Methods:

• Utilized a consensus-based optimization (CBO) algorithm for derivative-free global optimization.
• Modeled light propagation using the radiative transfer equation and simplified diffusion models (e.g., SP3).
• Developed and tested an adaptive CBO approach that dynamically adjusts the complexity of the light propagation model.

Main Results:

• All tested models, including diffusion approximations, achieved accurate bioluminescent source localization in simulations and phantoms.
• The adaptive CBO approach rendered higher-order models more efficiently than fixed-model approaches.
• The diffusive approximation performed well in highly scattering media, while higher-order models showed slight advantages in absorbing media.

Conclusions:

• Consensus-based optimization is a robust and adaptable method for bioluminescent source localization.
• The adaptive model hierarchy approach enhances computational efficiency without sacrificing localization accuracy.
• The choice of light propagation model impacts localization performance based on tissue optical properties (scattering vs. absorption).