Deep-Learning-Based Real-Time Visual Detection Method for Robotic Cell Microinjection System

Strategic Insights -

Robotic cell microinjection systems have revolutionized the field of life sciences by enabling the precise injection of biological cells. However, detecting cells and microinjector tips in real-time has proven to be a significant challenge due to their transparency and varying sizes. Researchers from the School of Automation have proposed a novel deep-learning based real-time visual detection method to overcome this limitation. This method employs a improved YOLOv8n algorithm, which utilizes a deformable convolution and bidirectional feature pyramid network to achieve high accuracy and precision. The proposed method has been experimentally verified using zebrafish embryo injection, demonstrating an average intersection over union accuracy of 99.5% for cells and 97.8% for microinjector tips.

Key Takeaways:

  • The proposed deep-learning based real-time visual detection method employs an improved YOLOv8n algorithm for cell and microinjector tip detection.
  • The method utilizes a deformable convolution and bidirectional feature pyramid network to achieve high accuracy and precision.
  • Experimental results show an average intersection over union accuracy of 99.5% for cells and 97.8% for microinjector tips.
  • The method also demonstrates 100% and 94.7% success rates for cell and microinjector tip detection, respectively.
  • The detection speed of the proposed method is 118.8 frame rate per second (FPS).
  • The method enables real-time detection of positions of cell and microinjector tip and provides visual feedback.
  • The method is promising in the application of robotic batch cell microinjection with high efficiency.
  • Guozhi Liu, School of Automation, was involved in the research and can be contacted for more information.
  • The research received financial support from National Natural Science Foundation of China and Natural Science Foundation of Jiangsu Province.
  • The proposed visual detection method is expected to have significant implications for robotic cell microinjection systems.

Statistics:

  • Average intersection over union accuracy: 99.5% for cells and 97.8% for microinjector tips.
  • Success rates: 100% and 94.7% for cell and microinjector tip detection, respectively.
  • Detection speed: 118.8 frame rate per second (FPS).
  • Computational resources: 1.96 million parameters and 7.2 giga floating point operations per second (GFLOPs) of computation.

Sources:

  • Journal of Robotics, 2025,2025. (Journal of Robotics - https://www.hindawi.com/journals/jr/). The publisher for Journal of Robotics is Wiley.
  • A free version of this journal article is available at https://doi-org.sdpl.idm.oclc.org/10.1155/joro/8556780.