Abstract
Microinjection of foreign materials into individual cells is a critical technique in genetics, transgenics, and drug discovery. However, manual injection of zebrafish embryos is labor-intensive and prone to human error. This paper presents a microrobotic system for fully automated zebrafish embryo injection. By integrating computer vision and high-precision motion control, the system achieves an injection speed of 15 embryos per minute with a 98% survival rate and 99% success rate. This high-throughput approach offers a scalable solution for biological and pharmaceutical research.
Introduction
Microinjection is a cornerstone of modern biological research, enabling the introduction of DNA, RNAi, proteins, and drugs into single cells. Zebrafish embryos are particularly valuable models due to their transparency and rapid development. Despite their utility, the manual injection process is notoriously difficult. Technicians must operate under a microscope for extended periods, leading to fatigue and inconsistent results. While semi-automated systems have been proposed, they often require human intervention for target selection or alignment. This research addresses the need for a "hands-off," fully automated system capable of high-throughput operation without sacrificing precision or embryo viability.
System Design and Methodology
The developed microrobotic system utilizes a multi-layered approach combining hardware precision with intelligent software control.
Hardware and Immobilization
The system features two high-precision, three-degrees-of-freedom (3-DOF) microrobots. Microrobot-A serves as a precision XY stage to hold the embryos, while Microrobot-B controls the injection needle. To solve the challenge of sample immobilization, the researchers moved away from traditional agarose gel trenches. Instead, they developed a vacuum-based holding device featuring a plate with an array of 400 μm holes. Negative pressure is used to rapidly secure embryos into a regular grid, a process that takes only seconds.
Computer Vision and Structure Recognition
The "brain" of the system is a computer vision pipeline that identifies embryo structures in approximately 45 milliseconds. Using a CMOS camera and optical microscope, the system employs active contours (snakes)—energy-minimizing splines—to precisely locate the boundaries of the chorion (outer shell) and the cytoplasm (the target). This allows the system to distinguish between the yolk and the cell portion, enabling specific deposition destinations to optimize molecular diffusion.
Control Strategy
The system follows a "looking-then-moving" strategy. It detects the vertical height of the needle once per batch and then utilizes PID control to coordinate both robots in parallel. To maximize throughput, the system is designed so that while one embryo is being retracted after injection, the next is already being moved into the field of view.
Experimental Results
The system was validated through two primary experimental trials:
- Fluorescent Dye Trials: 350 zebrafish embryos were injected with Rhodamine B. The system achieved a 99% success rate in material deposition and a 98% survival rate, demonstrating that the robotic touch is gentle enough to maintain embryo viability.
- Biological Validation: 210 embryos were injected with a "no tail" (ntl) morpholino. This functional test resulted in a 98.5% phenotypic rate, where the resulting fish lacked tails, confirming the system's efficacy for genetic research.
The system operates at a speed of 15 embryos per minute (with chorion unremoved), significantly outperforming manual methods in both speed and consistency.
Discussion and Conclusion
The integration of parallel robot movement and robust structure recognition allows this system to overcome the "dead time" inherent in manual or semi-automated processes. The use of adaptive thresholding and convex hull algorithms ensures the system remains robust despite variations in embryo size or orientation.
While the current setup focuses on zebrafish, the underlying immobilization and control methods are theoretically applicable to other biological targets, such as mouse oocytes or Drosophila embryos. Future work will focus on further automating the initial sample loading process and adapting the vision algorithms for different cell types. This fully automated system represents a significant step forward in enabling high-throughput biological and pharmaceutical research.
Summary updated