Centroid calculation of the blastomere from 3D Z-Stack image data of a 2-cell mouse embryo

Highlights

• 3D Z-stack Mouse embryo image data was captured by a bright field inverted microscope.

• Existing image processing operations were used to segment cell area in each 2D image.

• 3D centroid coordinates were calculated from segmented z-stack images.

• Cell volume was reconstructed in 3D for visualization of the calculated 3D centroid.

• Cell manipulation tasks can be automated with 3D centroid and image-based feedback.

Abstract

During cell surgery, cell manipulation tasks such as placement of cells, extraction of organelles, blastomeres, or other structures from within a cell are essential. To enable these tasks, it is necessary to locate the centroid of such objects within a cell, to be able to position micromanipulators to carry out such tasks. Calculation of the centroid of objects within a cell is necessary to enable vision-based feedback control to micromanipulators for automation of cell tasks. In this paper, we propose a methodology to address the calculation of the centroid of blastomeres within a mouse embryo. This is achieved with z-stack image data from 2D cell images. Z-stack images are captured at uniform intervals over the cell depth, by keeping the embryo fixed and moving the focal plane. Individual (2D) images of the z-stack are then processed. Blastomeres in each image are segmented with existing image processing techniques and converted to mask images. From these masked images, the 2D blastomere image centroids are calculated, allowing the centroid of the blastomere volume in 3D to be computed. The proposed calculation of the blastomere centroid is a critical step in determining the centroid of objects within a cell, to facilitate automation of cell surgery tasks.

Introduction

Automation of cell surgery tasks is made possible through the use of image data, actuators, feedback controllers and micro-manipulators, such as micropipettes [1], microneedles [2], [3] and micro-grippers [4]. Micromanipulation can also be achieved through the use of electric currents, optical tweezers and electromagnets [5], [6].

Cell micromanipulation has been used in gene injection, in-vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI) [7] and other single-cell tasks. During manipulation, biological material from the inside of the cell may be biopsied or a protein, sperm, or drug compound may be injected into the cell. For example, to fertilize an embryo, sperm is inserted into the oocyte with a micro-needle injector. Micromanipulators come into direct contact with biological cells, hence exert contact forces and pressure on the cells which are being manipulated. To facilitate the automation of this interaction, image-based feedback control techniques are used, with images captured with microscopes and processed during cell manipulation. Quantitative measurements such as cell location, the centroid of the cell and centroid of internal cell structures, and the location of the micromanipulator must be calculated from these images to facilitate cell surgery automation. These data provide the feedback signals to control micromanipulators, micropipettes to achieve successful automation.

This paper presents an approach to estimate the centroid of blastomeres within a mouse embryo. To facilitate these calculations, 2D planar microscope images, obtained at a uniform interval in the vertical coordinate direction, over the entire volume of the embryo are processed. With the fixed mouse embryo on the 2-DOF (degree of freedom) microscopic stage, the microscope focal plane (objective) is moved vertically, and a new image is obtained while the focal plane (objective) movement is stopped at uniform displacements in the vertical direction. The capture of these cell cross-sectional images, also known as z-stack images, Guisti et al. [8] is made possible because the embryos are transparent. Fig. 1 represents the image plane arrangements during the collection of a z-stack image dataset. After capturing these z-stack images, each 2D image is processed to identify the blastomeres in the image. The centroid of each blastomere in each 2D z-stack image is calculated. The centroid of the blastomeres can then be obtained from the full set of z-stack images, providing the 3D coordinates of the blastomeres. With the 3D centroid coordinates of each blastomere determined, foreign material could then be injected, or the blastomere could be extracted in a subsequent cell surgery process. Thus, the major contribution of this paper is the estimation of the 3D centroid coordinates of a blastomere which represents an important first step towards the automation of single-cell surgical processes, using 3D coordinate data. The paper utilizes existing image processing software such as MATLAB and well-known image processing methods, as they have reached a high level of sophistication over the years.

In the following, Sec. 2 reviews the literature pertaining to single-cell manipulation and the techniques used which are almost exclusively based on the use of 2D image data. Sec. 3 discusses the experimental set-up for capturing the z-stack dataset. An image segmentation method applied to the blastomere images (2D images of a 2-cell mouse embryo) is proposed in Sec. 4. This is followed by the (3D) centroid calculation of blastomere volume from multiple-segmented z-stack images in Sec. 5. At the end, evaluation of the calculated centroid coordinates and segmentation method with Intersection over Union (IoU), reconstruction of the blastomere volume, computational requirements for the proposed approach, comparison between 2D and 3D image-based feedback-controlled cell manipulation, and evaluation of the proposed segmentation method with the state-of-the-art segmentation methods are discussed in Sec. 6.