Computational and experimental study on accuracy of off-axis reconstructions in optical diffraction tomography

Abstract

We present a study on spatial changes in the accuracy of tomographic reconstructions obtained with two of the most popular tomographic reconstruction algorithms for diffraction tomography—filtered backprojection (FBPJ) and Rytov-based filtered backpropagation (FBPP). We find out that not only FBPJ but also FBPP suffers from a significant loss of accuracy in the off-axis regions of a tomographic reconstruction and this effect is stronger for objects with a high refractive index contrast. Moreover, we propose some modifications to FBPP which allow for significant improvement of the off-axis performance of the algorithm. In the modified algorithm, called the extended depth of focus filtered backpropagation (EDOF-FBPP), scattered waves are backpropagated using a rigorous propagation algorithm, and then the Rytov approximation is applied on extended EDOF images. This modification (1) prevents violation of the Rytov validity condition due to the defocus of scattered waves and (2) suppresses unwrapping errors. The tomographic reconstruction algorithms FBPJ, FBPP, and EDOF-FBPP are extensively tested with numerical simulations supported with rigorous wave scattering methods. The experimental evaluation of the performance of the tomographic algorithms is provided with a tomographic measurement of an optical microtip located 21 μ m from the central axis of the reconstruction.