Subspace Based CS-Music for Diffuse Optical Tomography

Abstract

Diffuse optical tomography (DOT) is a low cost imaging modality that reconstructs the optical coefficients of a highly turbid medium. However, the inverse problem is ill-posed, nonlinear, and unstable due to diffusive nature of optical photons through the biological tissue. The conventional DOT imaging methods require the forward problem to be solved repeatedly at each iteration which makes it computationally expensive. Recently, the theory of compressive sensing (CS) has been used in DOT and provided significant reconstruction of sparse objects in many DOT imaging problems. The main objective of this paper is to solve the DOT inverse problem using MMV (multiple measurement vectors) based CS framework and the sparse recovery algorithm like CS-MUSIC (multiple signal classification) is studied. The experimental validation of the CS-MUSIC has been done on a paraffin wax rectangular sample through a DOT experimental set up. We also studied the conventional DOT imaging method like least square method in this paper. The performance metric mean square error (MSE) is used to evaluate the performance of the reconstruction in DOT imaging. Simulation results showed that the CS-MUSIC algorithm outperforms the conventional DOT imaging method in DOT imaging. The advantage of this study is that the forward problem need not be solved repeatedly which are inherent in conventional DOT.