Automatic three-dimensional imaging for blastomere identification in early-stage embryos based on brightfield microscopy

摘要

As the first step in manual or automated preimplantation genetic diagnosis (PGD), accurate locating of the blastomere is essential. In order to further improve success rates of the PGD process, it is necessary to obtain three-dimensional information of the blastomeres. In this study, a practical and effective pipeline that allows conventional brightfield microscopy systems to provide 3D imaging feedback for individual blastomeres inside embryos is proposed. The technique consists of two main phases, namely the blastomere identification and the 3D modeling of individual blastomeres. In the first phase, after a series of optical sections are obtained along the z-axis (z-stack), the embryo is located in the z-stack and the blastomere outline is then identified inside the embryo using a coarse-to-fine strategy; In the second phase, a probabilistic model combined with the machine learning approach is designed for the boundary estimation between individual blastomeres, and the 3D coordinate and model of each blastomere are obtained through further relocalization and reconstruction in the z-stack. Experimental results demonstrate that the proposed technique is valid and accurate compared with human perception. Specifically, the proposed technique achieves 90.32% precision rate, and the bias of blastomere position between the proposed method and human segmentation is 1.028 mu m.

关键词