Adaptive particle manipulation with AI-guided optoelectronic tweezers

The ability to manipulate particles at the nanoscale with optoelectronic tweezers (OETs) and other techniques has driven biomedical breakthroughs by enabling precise analysis of complex biological systems. However, traditional OET-based techniques rely on predefined patterns and cannot readily adapt to crowded or dynamic environments. Wang et al. developed an OET sorting method that uses machine learning to precisely manipulate particles in a complex, evolving system.

The team’s low-resource technique combines artificial intelligence methods like deep learning and object tracking with a path-predicting algorithm — the artificial potential field (APF) — that minimizes particle collisions and automatically recalculates trajectories in real time. Notably, the algorithm enables particles to autonomously bypass static obstacles and other particles so that they maintain stable and precise navigation in complex, microfluidic environments.

“The artificial potential field (APF) algorithm enabled multiple particles to move simultaneously along independent paths without collisions, even in dense and dynamic environments,” co-author Liuyong Shi said. “We also introduced a self-feedback mechanism that enabled the OET to move with the particles it was manipulating, ensuring continuous tracking.”

In the study, real-time images of particles in a microfluidic chip were captured on a microscope, then analyzed using computer vision models for classification, detection, and tracking. The results were then fed into an APF-based algorithm that treated targets as attractive sources and obstacles as repulsive ones. For each frame, the team then computed force vectors that guided OET components to manipulate particle trajectories while avoiding collisions.

Future work will focus on improving image frame rates, model decision-making times, and expanding the study to real biological samples.

Source: “Artificial potential field-enhanced optoelectronic tweezer technology for path planning and intelligent sorting of particles,” by Tianyi Wang, Shizheng Zhou, Jianghao Zeng, Guibiao Qian, Zhihao Wu, Jingming Bai, Lian Sun, Zhihang Yu, Hong Yan, Teng Zhou, Hongming Chen, and Liuyong Shi, Physics of Fluids (2025). The article can be accessed at https://doi.org/10.1063/5.0294345 .