Vision-guided robotic micromanipulation for cell rotation control

• Main Author: Cui, Shuai
• Other Authors: Ang Wei Tech
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Robotic micromanipulation; Single cell manipulation; Image processing; Visual servo control
• Online Access: https://hdl.handle.net/10356/173852
• Institution: Nanyang Technological University

Rotation control of single biological cells is an essential part of several cell surgery applications. It is the starting point for the subsequent analysis. Under a microscope with limited vision and space, biological cells sometimes need to be rotated correctly to make certain cellular structures observable or ready for the following operations. Manual manipulation usually lacks accuracy, stability, and efficiency, even aided with devices like the motorized micromanipulator. Several methods and techniques for robotic cell rotation control are developed and elaborated in this research to address the aforementioned limitations and problems. In biology laboratories and fertility centers, glass capillary micropipette is the most commonly used end-effector for manipulating biological cells. Its tip is commonly required to be aligned with the target object and the platform in relevant applications. The first robotic micromanipulation method that addresses the misalignment caused by the improper incline angle at which the micropipette is held is presented in this thesis. This misalignment error can be accurately identified and eliminated effectively and precisely by the designed vision-guided robotic micromanipulation system. The alignment of the micropipette lays the foundation for further micromanipulation, like rotation, injection, and transfer of biological cells. It guarantees the maneuverability of the proposed cell micromanipulation methods and can reduce the risk of causing unwanted trauma to the cell throughout the operation. Experiment results confirmed that the robotic controller has better efficiency and precision than the traditional manual operation. A friction force-based micromanipulation cell rotation control method is explored and analyzed. Taking mouse embryos as an example, corresponding force models of the embryo while being rotated are established to identify the key parameters for generating the effective rotation in a deterministic way. The operation process is stable and the designed controller is easy to implement with good rotation accuracy. It averagely achieved a rotational precision of one degree in the experiment. Developed and deployed on the standard clinical setups, this method can directly contribute to automating the operation procedures of relevant applications with improved repeatability and stability. Considering the requirement for the rotation precision is not strict in practical applications like embryo biopsy, a hydrodynamic force-based cell rotation method is proposed and developed. By tuning the flow rate of the fluid from the micropipette orifice, the generated torque can always guarantee an effective cell rotation. Although uncertainty seems everywhere at the micro scale, the validated method developed on the basis of the simultaneous perturbation stochastic approximation (SPSA) method for cell in-plane rotation control provides insight into designing controllers for micromanipulation tasks without full knowledge of the system model. Accepting the rotation error no larger than five degrees, results of experiments conducted on mouse oocytes show that the cell can be oriented to the target position with better robustness and a higher success rate (91.7%). Embryo biopsy is taken as an example to rethink and discuss how to apply the proposed cell rotation methods and other related techniques in real applications. An optimized embryo rotation strategy, where the embryo out-of-plane rotation and in-plane rotation are primarily completed by the friction and hydrodynamic force-based method, respectively. Experiment results well suggested the feasibility of the optimized strategy with an overall success rate of 92%. Meanwhile, an image processing system built on the basis of U-Net is developed for selecting the optimal dissection position in the zona pellucida (ZP). The system is proven reliable as it effectively indicated the optimal dissection position in the ZP with a success rate of 95.6%. Based on all experiment results and associated analysis, the methods and techniques introduced in this research exhibit promising potential for future applications.