Label-free identification of apoptotic cell death using deep learning for imaging flow cytometry 2564

Abstract Description Apoptosis is a form of cell death critical for embryonic development and tissue homeostasis. Dysregulation of apoptosis is associated with diseases such as oncogenesis and autoimmune disorders; therefore, understanding the mechanism of apoptosis is essential and may lead to development of new therapeutics for treatment of diseases. A common method to identify apoptotic cells is to stain them with live/dead, fluorescent apoptotic markers and measure the fluorescence intensity using flow cytometry. Despite wide usage, this method requires cell staining with fluorescent dyes and antibodies which may disturb the functional status of cells and even prevent users from certain downstream applications. Thus, we designed a machine-learning algorithm based on label-free images from imaging flow cytometry to identify apoptotic cells. Using a cellular apoptotic model, we captured label-free images in a heterogeneous sample using imaging flow cytometry. A deep learning algorithm was applied to differentiate the functional status of cells including live, dead and apoptotic cells. The algorithm achieved an average F1-score of 90% in comparison to the ground truth labeling. In unsupervised clustering, live, dead, and apoptotic cells were clearly distinguishable using deep learning-derived imaging features. The label-free identification of live, dead, and apoptotic cells by deep learning and imaging flow cytometry provided an accurate and efficient way to define cell functional status. Topic Categories Technological Innovations in Immunology (TECH)