Mesh optimization for Monte Carlo based optical tomography

Fluorescence Molecular Tomography aims to reconstruct the 3D distribution of fluorescent markers in bio-tissues based on 2D surface measurements of emitted photons. This technique requires an accurate model of light propagation, the gold standard of which is created by the Monte Carlo (MC) method. One drawback of MC is the computational burden associated with the need to simulate large packet of photons to sample the volume to be imaged with accuracy. Recent developments in MC formulation and massively parallel computing geared towards optical tomogrpahy have allowed to alleviate this issue. Especially, mesh based MC techniques have shown favorable computational costs compared to voxel-based MC. Herein, we investigate the potential of mesh optimization strategies for computationally efficient and accurate Monte Carlo based optical tomography. Using a mouse model created from μCT data and average murine optical properties, we investigate the potential of an iterative mesh refinement strategy. Performances of the method are evaluated in the image space. Our preliminary results indicate that accuracy improves over several iterations when mesh refinement is employed.