Deep Learning Diffuse Optical Tomography

Abstract

In this paper, we propose an interpretable deep learning approach to solve the inverse scattering problem of diffuse optical tomography (DOT). Unlike the conventional approaches, which consider a neural network as a black-box, our proposed network is designed to learn a mapping that inverts the Lippmann-Schwinger integral equation, which describes the non-linear physics of photon migration in turbid media. By using real data from a prototype DOT system, we showed that our deep neural network, trained with only simulation data, can accurately recover the 3D distribution of optical anomalies in various experiments without the use of an exogenous contrast agent.