Versatile Image-Assisted Cell Sorting by Selective Trapping with Spatiotemporal Multiparameter Targeting

Current cell sorting methods often lack versatility, require complex setups, demand large initial cell numbers for reliable sorting, and impose size limitations on target objects. To address these challenges, we introduce two-dimensional sorting with image-guided multiparameter adjustable targeting (2D-SIGMAT), which utilizes dynamic in situ light-activated cell trapping for precise and efficient cell isolation. A key advantage of 2D-SIGMAT is its capability to record and analyze high-resolution images with over ten times more pixels per image without motion blur compared to the other image-assisted sorters, significantly enhancing sorting. We have demonstrated the sorting of objects with a wide range of sizes, from single cells to organoids. Furthermore, we have shown that it is compatible with fluorescent and bright-field imaging, as well as deep neural network models for robust target detection. We have also demonstrated sorting based on high-resolution temporal data, capturing dynamic cellular behavior. Using the deep learning object detection model YOLOv5, 2D-SIGMAT achieves up to 98% recovery efficiency at a throughput of up to 2000 cells per second. Finally, we show that a standard microscope equipped with a ultraviolet (UV) projector can be transformed into a versatile, high-performance cell sorter with a unique scan-select sorting capability with broad application potential.