Assisted reproductive technologies (ART) are a host of technologies and procedures related to fertility and reproduction. Preimplantation genetic diagnosis (PGD), a procedure within ART, is a form of embryo biopsy in which a single cell, called a blastomere, is non-destructively extracted from the embryo for genetic analysis. PGD procedures are currently performed manually with low success rates, potentially caused by operator performance variability and contamination through handling, and are financially, physically, and psychologically taxing for the patient. The automation of PGD can increase the procedure repeatability and hence improve overall success rates. This thesis presents novel methods for the automation of certain PGD embryo biopsy procedures using robotic micromanipulators, computer vision and automation algorithms. This dissertation is separated into three main chapters, each pertaining to the automation of the three main phases of PGD biopsy:​

The first step in PGD biopsy automation is automated embryo position and orientation control using independently controlled parallel plates. Computer vision algorithms are used for tracking the embryo position and orientation. A novel 3D mapping technique for use with Hoffman modulation contrast microscopy is described and used for ablation zone optimization. Open and closed loop position and orientation control is demonstrated using the parallel plates.

The second step of PGD biopsy automation performs a novel two-stage optimization for laser zona drilling to reduce embryo heating. The first stage uses computer vision algorithms to identify embryonic structures and determines the optimal ablation zone farthest away from critical structures such as blastomeres. The second stage combines a genetic optimization algorithm with a thermal model of LZD to optimize the combination of laser pulse locations and pulse durations.

The third step of PGD biopsy automation involves blastomere extraction using the displacement method. A process flow for blastomere extraction automation is developed, and implemented using algorithms for non-vision-based feedback micropipette position control, and computer vision algorithms for blastomere extraction event detection and blastomere tracking during retrieval.

For each PGD biopsy step, automation experiments were performed on mouse embryos. Successful demonstration of the above methods proves the feasibility of biopsy automation, and presents a step towards fully automated ART.