Abstract

AI Enabled 2D and 3D Image-Guided Cell Analyzers and Sorters

Yu-Hwa Lo, Professor, University of California San Diego

Single cell analysis is playing an important role in biological and medical research.  Flow cytometers and fluorescence-activated cell sorters (FACS) are workhorse for single cell analysis, offering high throughput and the ability of isolating cells of high interest, especially those rare cells or new cell types.  On the other hand, microscopy, especially 3D microscopy, can produce high information contents, revealing important insight in cell phenotypes and high spatial information although cell isolation in a microscope platform with techniques such as laser microdissection is cumbersome and subject to cell damage. We present results of image-guided cell sorters that possess the merits of high throughput and high content imaging in a single system.  The image-guided FACS system uses a microfluidic cartridge with an on-chip piezoelectric actuator to sort cells of interest into tubes or 384-well plates.  The cell sorting criteria is based on the 2D images of cells, defined by either operator specified gating criteria or AI with or without human instructions. Going beyond the systems with 2D images, we further demonstrate flow cytometers that produce 3D fluorescent, scattering, and transmission images of each single cell at a throughput of 1000 cells/s, which is 1000-10,000 times faster than a fluorescent confocal microscope. The cell tomography offers far more information than 2D images and is more adaptive to AI and machine learning. Both the 2D and 3D image-guided cell analyzers and sorters are powered by artificial intelligence with convolutional neural networks.  With AI capabilities based on semi-supervised learning, both systems have shown exceptional capabilities for classification and cell type discovery.  Results from several experiments, including the diagnosis of leukemia and NASH diseases, DNA damage assessment, protein translocations under drug treatments, cell cycle detection, and cell fate prediction (the so-called Day Zero experiment) will be discussed in the presentation.