Computer vision guided embryo biopsy
Date of Issue2015
School of Mechanical and Aerospace Engineering
Robotics Research Centre
Embryo biopsy, which is a process to remove a part of materials from a cell, is a necessary step of preimplantation genetic diagnosis (PGD). The rapid advancement of the PGD technique has been calling for easy, fast and precise automated embryo biopsy techniques to improve the efficiency, the throughput, and the survival rate of the PGD. In this thesis, a computer vision guided embryo biopsy system is introduced. This research is devoted to demonstrate several techniques that contribute towards the fully automated embryo biopsy system, including the embryo autofocusing, the embryo image segmentation, the embryo orientation, and the zona pellucida (ZP) dissection. A fast and robust autofocusing algorithm consisting of an automatic focus window selection method and an optimum searching algorithm based on the shuffl frog leaping (SFL) algorithm is proposed. The size-adjustable focus window is capable of improving the accuracy of the focus measurement func- tion (FMF) graph of the target object by focusing only on the informative portion of the image, thus leading to an accurate embryo position stimation in Z axis. The proposed improved shuffl frog leaping (ISFL) searching al- gorithm is able to bypass local maxima and reach the global maximum in the non-unimodal FMF graph, which effectively identifies the most suitable focal plane for single blastomere. A technique that is able to select a suitable biopsy position on the zona pellucida automatically is also developed. The proposed technique consists of a method of estimating the elliptical internal and external ZP boundary and two methods of suitable ZP dissection position selection. The validity of this technique is demonstrated by both quantitative and qualitative results. The elliptical ZP boundary estimation can also be applied to measure zona pellucida thickness variance (ZPTV) and other ZP morphology measurements. A 3-point contact automated three-dimensional cell rotational system for single cell manipulation is developed. A high resolution in-house rotational stage and cellular structure recognition algorithms are designed and developed to facilitate the visual servo control. The system is capable of orienting a zebra sh embryo to the ideal viewing cross-section that is aligned with the focal plane of the microscope. Based on the investigations in the axial and lateral oscillations of the piezo-driven micropipette in cell membrane injection and dissection, a piezodriven lateral-oscillation micro-cutter is proposed to perform both single direction and three-dimensional cell dissection. The proposed technique is superior to the classic manual technique in many ways such as higher precision,lower skill requirement and shorter processing time.