Quantitative phase imaging (QPI) has emerged as a valuable method in biomedical research by providing label-free, high-resolution phase distribution of transparent cells and tissues. While QPI is limited to transparent samples, quantitative oblique back-illumination microscopy (qOBM) is a novel imaging technology that enables epi-mode 3D quantitative phase imaging and refractive index (RI) tomography of thick scattering samples. This technology employs four oblique back illumination images taken at the same focal planes, along with a rapid 2D deconvolution reconstruction algorithm, to generate 2D phase cross-sections of thick samples. Alternatively, a through-focus z-stack of oblique back illumination images can be utilized to produce 3D RI tomograms, offering enhanced RI quantitative accuracy. However, 3D RI generation requires a more computationally intensive reconstruction process, preventing its potential of a real-time 3D RI tomography. In this paper, we propose a neural network-involved reconstruction technique that significantly reduces the processing time to a third while maintaining high fidelity compared to the deconvolution-based results.
|