High-speed optical diffraction tomography (ODT) with deep-learning approach (Conference Presentation)

Abstract

Optical diffraction tomography (ODT) is a powerful label-free three-dimensional (3D) quantitative imaging technique. However, current ODT modalities require around 50 different illumination angles to reconstruct the 3D refraction index (RI) map, which limits its imaging speed and prohibit it from further applications. Here we propose a deep-learning approach to reduce the number of illumination angles and improve the imaging speed of ODT. With 3D Unet architecture and large training data of different species of cells, we can decrease the number of illumination angles from 49 to 5 with similar reconstruction performance, which empowers ODT the capability to reveal high-speed biological dynamics.