CNN-assisted quantitative phase microscopy for biological cell imaging

Abstract

Phase imaging is a solution for the reconstruction of phase information from intensity observations. To make phase imaging possible, sophisticated extra systems are embedded into the existing imaging systems. Contrary, we propose a phase problem solution by DCNN-based framework, which is simple in terms of an optical system. We propose to replace optical lenses with computational algorithms such as CNN phase reconstruction and wavefront propagation. The framework is tested in simulation and real-life experimental phase imaging. To have real experiments with objects close to real-life biological cells, we simulated experimental training datasets on a phase-only spatial light modulator, where phase objects are modeled with corresponding phase distribution to biological cells.