FMT inversion using XCT image priors without strong anatomy function correlations

Fluorescence molecular tomography (FMT) is a relatively new optical imaging modality that uses fluorescent markers that accumulate in specific regions, to monitor cellular and subcellular functional activity in-vivo in small animals. In FMT, near-infrared (NIR) light sources at excitation wavelength are directed to the animal body at different projection angles. The excitation light travels diffusely into the tissue and part of the photons are absorbed by fluophores, which re-emit part of the light at a longer wavelength. Then, both the excitation and fluorescence projected light intensities are measured using a charge-coupled device (CCD) camera placed opposite the source. These multiple projections are combined to obtain the distribution of fluophore concentration. However, due to the diffusive nature of light propagation in biological tissue the image reconstruction is an ill-posed inverse problem, and in addition, it is also underdetermined. Hence, in order to obtain a stable and meaningful solution one has to use regularisation methods. Prior structural anatomical information from X-ray computer tomography (XCT) images can be incorporated into the reconstruction to improve optical imaging performance. However, some image priors may lead to biased solutions. We aim to develop and analyse the performance of different priors, which improve the accuracy of the FMT problem without biasing the solution.

This work is part of the European project FMT-XCT.