A rapid and efficient 2D/3D nuclear segmentation method for analysis of early mouse embryo and stem cell image data

Abstract

Segmentation is a fundamental problem that dominates the success of microscopic image analysis. In almost 25 years of cell detection software development, there is still no single piece of commercial software that works well in practice when applied to early mouse embryo or stem cell image data. To address this need, we developed MINS (modular interactive nuclear segmentation) as a MATLAB/C++-based segmentation tool tailored for counting cells and fluorescent intensity measurements of 2D and 3D image data. Our aim was to develop a tool that is accurate and efficient yet straightforward and user friendly. The MINS pipeline comprises three major cascaded modules: detection, segmentation, and cell position classification. An extensive evaluation of MINS on both 2D and 3D images, and comparison to related tools, reveals improvements in segmentation accuracy and usability. Thus, its accuracy and ease of use will allow MINS to be implemented for routine single-cell-level image analyses.

Graphical abstract

Figure 1

Image Analysis of Cells and Mouse Embryos and a Schematic of Preimplantation Embryo Development (A) Schematic showing the experimental setup used for static and live imaging of stem cell and mouse embryo specimens. Notably, samples are maintained in liquid culture, and images are acquired on inverted microscope systems. (B) Examples of imaging acquisition of 3D static immunostaining (left) or 3D live imaging of fluorescent reporter (right). (C) Schematic diagram showing 2D, 3D, and 4D image data acquisition and analysis. (D) Differential interference contrast (DIC) images of CAG:H2B-GFP transgenic fluorescent reporter expressing embryos at two-cell, compact morula, early, and late blastocyst stages merged with 2D and 3D renderings of GFP channel showing nuclei labels and a schematic diagram of lineage specification during preimplantation development (Schrode et al., 2013). Scale bar, 20 μm.

Figure 2

Procedure for Nuclei Detection and ID (A) Users provide an input image from, for example, a mouse embryo, as shown here. (B1 and B2) The first (B1) and second (B2) eigenvalues of the Hessian matrix are computed from the smoothed image at different scales. (C) A binary segmentation is obtained by thresholding the respective eigenvalues in (B1) and (B2). (D) The final detection is obtained by combining the binary segmentations in (C), and each nucleus is assigned with a unique number using connected component analysis.

Figure 3

Flow Diagram of Proposed Algorithm for Nuclear Segmentation (A) Users provide an input image from either cell culture imaging, in columns (A1) and (A2), or live embryo imaging, in column (A3). (B) Detection is performed to locate each nucleus. (C) Graph coloring is used to separate proximate nuclei by assigning different colors to them. (D and E) Iteratively, a color is selected (D), and geodesic segmentation is called to segment the entire body of the nuclei (E). (F) The final segmentation is obtained by combining the segmentations from (E). Scale bar, 20 μm.

Figure 4

Multistep Classification: Multiple Embryo Extraction and Outlier Removal (A) The embryo separation algorithm successfully detects two embryos in (i) and five embryos in (ii). False detections from the background are mistaken for true embryonic cell nuclei (yellow arrows). (B) Outlier removal discards most of false detections (yellow arrows). True cell nuclei can be misclassified as outliers if they are located at the embryo boundary (red arrow). (C) Maximum intensity projections at single time points 3D time-lapse movie over (i) a 540 min period and (ii) a 1500 min period. (D) Performance evaluation of (i) multiple embryo extraction with outlier removal on the data set described in (C) and (ii) nuclear segmentation over an extended period. Scale bar, 20μm.

Figure 5

Multistep Classification: TE versus ICM Classification (A) Schematic of TE versus ICM lineage allocation in preimplantation mouse embryo. (B) Preimplantation embryos over various stages immunostained with TE (CDX2, green) and ICM (NANOG and GATA6, magenta) markers. (C) Schematic diagram of TE versus ICM classification procedure by MINS. (D) Performance evaluation of lineage classification. Scale bar, 20 μm.

Figure 6

Overview of the MINS Platform and Its Comparison with Related Tools (A) The main GUI of MINS. The top boxes contain functions for parameter loading and saving. The middle boxes correspond to the entire processing pipeline. The bottom boxes allow batch processing on a large number of data sets. (B) The processing pipeline and the output of each modules. (C) Detailed outputs ease any downstream analyses, either manually or by integration with other software tools. Overlay of segmentation and raw data allow rapid and straightforward inspection of the results. A segmentation information summary provides easy access to quantitation results. (D) Top: volume rendering of a raw 3D CAG:H2B-GFP mouse embryo data set and the segmentation output generated by FARSIGHT, ilastik, and MINS. For each segmentation, each segmented object is assigned a unique color descriptor. Bottom: visualization of a 2D section of the same data set. Scale bar, 20 μm.

Figure 7

Application of MINS for Quantitative Fluorescent Measurements (A) Nuclear segmentation and quantitative fluorescent measurements of mouse ES cells grown under different conditions that have been stained for the pluripotency-associated factor NANOG. Stem cells display a heterogeneous pattern of NANOG expression when grown under standard serum + LIF conditions (left column) but either downregulate its expression when LIF is absent (middle column) or markedly increase its expression in the presence of the 2i inhibitors (right column). Quantitative fluorescent measurements via MINS are indicative of the culture conditions used (scatterplot at right). (B) Nuclear segmentation and quantitative immunofluorescence of a 100-cell-stage embryo stained for the EPI-specific factor NANOG and the PrE-specific factor GATA6. There are two distinct population of cells within the ICM, either expressing high levels of GATA6 (blue arrowheads) or high levels of NANOG (red arrowheads). A single cell expresses similar levels of NANOG and GATA6 and could thus be categorized as unspecified for either EPI or PrE (green arrowhead). Quantitative fluorescent measurements via MINS are indicative of the two distinct cell populations within the ICM (scatterplot at the right). (C) Comparison of quantitative analysis of 3D time-lapse imaging data performed either manually or with MINS software on embryos carrying the Pdgfrα^H2B-GFP reporter cultured ex utero. Select single time points from a representative 3D time-lapse movie are shown on the top row. GFP intensities of individual cells identified in each embryo at selected time points are shown at the scatterplot to the bottom. Scale bar, 20 μm.