AI-assisted real-time parallel cell sorting with holographic optical trapping

Live cell sorting enables the acquisition of highly purified and functionally preserved populations for applications in disease diagnosis, stem cell research, and precision medicine. However, achieving high-efficiency and fully automated sorting remains challenging. Here, we present a real-time parallel AI holographic optical tweezer (PAIHOT) system that integrates YOLOv11n detection with Kalman filtering and class-matching for stable multi-target tracking and prediction in real time. Predicted trajectories guide the optical trap to the cell periphery, thereby reducing photodamage compared to conventional center-focused trapping. Experimental results demonstrated that PAIHOT achieves sorting purities exceeding 91% across multiple cell types, and the viability assays confirm intact morphology and strong growth activity. The PAIHOT enables high-accuracy, parallel, and low-damage cell sorting in dynamic microscopic environments, providing an effective and robust platform for intelligent, high-throughput single-cell research.