Deep learning-enabled Inference of 3D molecular absorption distribution of biological cells from IR spectra

Abstract

Infrared spectroscopy-based diffraction tomography has a high potential to be used for 3D reconstruction of intact samples, however, the inverse problem is highly non-linear and remains challenging. Here, the authors solve full-wave inverse scattering problems using deep convolutional neural networks and perform 3D spectral reconstruction by diffraction tomography from scatter-distorted IR spectra.