In-silico clearing approach for deep refractive index tomography by partial reconstruction and wave-backpropagation

Abstract

Abstract Refractive index (RI) is considered to be a fundamental physical and biophysical parameter in biological imaging, as it governs light-matter interactions and light propagation while reflecting cellular properties. RI tomography enables volumetric visualization of RI distribution, allowing biologically relevant analysis of a sample. However, multiple scattering (MS) and sample-induced aberration (SIA) caused by the inhomogeneity in RI distribution of a thick sample make its visualization challenging. This paper proposes a deep RI tomographic approach to overcome MS and SIA and allow the enhanced reconstruction of thick samples compared to that enabled by conventional linear-model-based RI tomography. The proposed approach consists of partial RI reconstruction using multiple holograms acquired with angular diversity and their backpropagation using the reconstructed partial RI map, which unambiguously reconstructs the next partial volume. Repeating this operation efficiently reconstructs the entire RI tomogram while suppressing MS and SIA. We visualized a multicellular spheroid of diameter 140 µm within minutes of reconstruction, thereby demonstrating the enhanced deep visualization capability and computational efficiency of the proposed method compared to those of conventional RI tomography. Furthermore, we quantified the high-RI structures and morphological changes inside multicellular spheroids, indicating that the proposed method can retrieve biologically relevant information from the RI distribution. Benefitting from the excellent biological interpretability of RI distributions, the label-free deep visualization capability of the proposed method facilitates a noninvasive understanding of the architecture and time-course morphological changes of thick multicellular specimens.