Live imaging technologies have revolutionized the study of numerous biological processes. Specifically, advanced imaging modalities have recently been developed enabling the real-time acquisition of millions of 3D cell features, opening the path for a better understanding of tissue development, repair, homeostasis and pathology. However, our capacity to analyze huge datasets generated from live imaging is lagging behind and the full potential of such imaging methods can only be reached when automatic, robust and user-friendly methods are developed to detect specific cellular events from the huge amounts of data generated. Moreover, a classical problem associated with live-imaging is that the functional studies of rare cellular events, such as division or apoptosis, entails their imaging at high temporal and spatial resolution. O-NEAT (On- and offline Networks for Event Aware Topological detection) addresses the classical and general problems associated with live-imaging of cells and tissues by implementing advanced imaging methods based on Deep learning to detect and predict cellular events (such as division, or apoptosis). We have developed the Offline-NEAT software that utilizes a trained Convolutional Neural Network followed by Long Short-Term Memory and dense layers to recognize cell events as an action sequence enabling the categorization of the cellular event. Building on Offline-NEAT, we have implemented an online intelligent imaging modality that predicts the occurrence of cellular events and adjusts the microscopy parameters to image these cellular events with high temporal and spatial resolutions. The cellular event can be predicted in epithelial tissues in 2D. As a whole, O-NEAT offers a novel imaging platform facilitating the detection and the acquisition of cellular events that should improve the exploration of cell dynamics in physiological and pathological contexts.
