Compressed sensing based diffuse optical tomography reconstruction using linearization approach

Abstract

Diffuse optical tomography (DOT) is non-invasive imaging modality that uses near infrared (NIR) light, wavelength of 700 to 1000nm. The purpose of diffuse optical tomography is to reconstruct optical properties of highly scattering biological tissue when anomalies are inside tissue. However, to solve inverse scattering problems in diffuse optical tomography, there are some difficulties since the problems are non-linear. In this study, we present reconstruction method that converting given problem to compressed sensing(CS) problem using linearization approach, and estimating the position of non-zero support as well as accurate value of absorption coefficient simultaneously using M-SBL algorithm, actually SBL algorithm, after applying $Calder \acute{o} n$ pre-conditioner to reduce computational complexity of Green’s function. Simulation and real data showed that proposed approach give higher accuracy compared to iterative shrinkage thresholding (IST) method, which is compressed sensing based reconstruction.