MaGIC-OT: an AI-guided optical tweezers platform for autonomous single-cell isolation in microfluidic devices

Area of Science:

• Biomedical Engineering
• Cellular Biology
• Artificial Intelligence

Background:

• Automating rare cell isolation, such as circulating tumor cells (CTCs), is crucial for liquid biopsy but remains challenging in microfluidic devices.
• Optical tweezers offer precise, non-contact cell manipulation but typically require expert operation.

Purpose of the Study:

• To develop an automated platform, MaGIC-OT (machine-guided isolation of cells using optical tweezers), for single-cell manipulation within microfluidics.
• To integrate classical path planning and deep reinforcement learning (DRL) for autonomous optical tweezer control.

Main Methods:

• Developed a high-fidelity simulation environment to train and evaluate DRL control policies for optical tweezers.
• Employed cooperative, human-in-the-loop training to enhance DRL agent performance.
• Integrated MaGIC-OT with microfluidic chip for on-chip cell isolation.

Main Results:

• Trained DRL agents demonstrated superior speed and isolation success rates compared to expert users in silico.
• Successfully performed proof-of-concept isolation of a cancer cell from a spiked blood sample using the MaGIC-OT platform.
• Showcased improved DRL performance through cooperative, human-in-the-loop training.

Conclusions:

• MaGIC-OT provides a flexible, intelligent framework for automated optical manipulation in microfluidics.
• The platform aligns autonomous control with microfluidic device design for advanced cell isolation.
• Offers a promising pathway toward high-purity, label-free single-cell isolation workflows for liquid biopsies.