Deep-learning-based label-free segmentation of cell nuclei in time-lapse refractive index tomograms

Abstract

In order to identify cell nuclei, fluorescent proteins or staining agents has been widely used. However, use of exogenous agents inevitably prevents from long-term imaging of live cells and rapid analysis, and even interferes with intrinsic physiological conditions. In this work, we proposed a method of label-free segmentation of cell nuclei in optical diffraction tomography images by exploiting a deep learning framework. The proposed method was applied for precise cell nucleus segmentation in two, three, and four-dimensional label-free imaging.A novel architecture with optimised training strategies was validated through cross-modality and cross-laboratory experiments. The proposed method would bring out broad and immediate biomedical applications with our framework publicly available.

Refractive Index Changes of Cells and Cellular Compartments Upon Paraformaldehyde Fixation Acquired by Tomographic Phase Microscopy

“…In order to introduce this quantitative parameter to life science applications and medical diagnostics, it is necessary to develop standard sample preparation and data processing procedures for usage in laboratories. Standardized protocols for the preparation of biological samples, for example, drying methods used for scanning electron microscopy (SEM), paraformaldehyde fixation protocols for QPI and fluorescence microscopy or tissue clearing as used for confocal microscopy, optical diffraction tomography (ODT), optical coherence tomography (OCT) or Raman microscopy, are critical for interlaboratory comparison of the results virtual staining (9), for remote diagnostics (7) as well as for novel methods of enhanced 3D RI distribution determination and analysis by use of artificial intelligence (AI) or machine learning (10–12).…”

“…, for remote diagnostics (7) as well as for novel methods of enhanced 3D RI distribution determination and analysis by use of artificial intelligence (AI) or machine learning (10)(11)(12).…”

Microscopy cell nuclei segmentation with enhanced U-Net

“…working as backbone models are not preferred solutions. The possible solution is then to check if modifications of classical U-Net will boost the accuracy of segmentation as [14][15][16] shown. Therefore, in our methods, the encoded branch of conventional U-Net was redesigned to fuse more image features between shallow and deep hierarchy layers (see "Model structure" section).…”

Label-free bone marrow white blood cell classification using refractive index tomograms and deep learning

“…Recent advances in artificial intelligence (AI) have suggested unexplored domains of QPI beyond simply characterizing biological samples. As datasets obtained from QPI do not rely on the variability of staining quality, various machine learning and deep learning approaches can exploit uniform-quality and high-dimensional datasets to perform label-free image classification (Chen et al 2016; Jo et al 2017; Nissim et al 2020; Ozaki et al 2019; Wang et al 2020; Wu et al 2020; Yoon et al 2017; Zhang et al 2020; Zhou et al 2020) and inference (Chang et al 2020; Choi et al 2019; Dardikman-Yoffe et al 2020; Kandel et al 2020; Lee et al 2019; Nguyen et al 2018; Nygate et al 2020; Pitkäaho et al 2019; Rivenson et al 2018). Such synergetic approaches for label-free blood cell identification have also been demonstrated, which are of interest to this work (Go et al 2018; Kim et al 2019; Nassar et al 2019; Ozaki et al 2019; Singh et al 2020; Yoon et al 2017).…”