Automated Three-Dimensional Image Based Localization of Blastomeres for Single Cell Surgical Applications

Single biological cell surgery requires accurate localization of cell structures and organelles within the cell to permit automated processing. This study proposes an image based system to localize blastomeres in 3-dimensions (3D), in preparation of embryo orientation and ablation. Optical section images along the z-axis (z-stacks) are used to localize the blastomeres using computer vision, and further reconstruction and relocalization are achieved by clustering the 3D data along the z-axis. Two 3D reconstruction algorithms were investigated. The first proposed system processes each z-stack image using local phase coherence maps (LPC) and circle Hough transforms to accurately locate the centroid of the blastomeres while the second system generates and localizes the 3D data, in real time, using the optical flow of the image feed as the z-stacks are acquired. Experimental results demonstrate accurate localization of the blastomeres.