Cell State Recognition of Cytopathic Effect with YOLO Detector

Abstract

Cell state recognition and evaluation of its life cycle are the fundamental and critical procedure to identify the pathological mechanism of viruses and analyze pharmacodynamic effectiveness. Modern life science research is based on large-scale biological experiments so that it is inefficiency to estimate and decide the physicochemical results on petri dishes. With the development of target recognition in computer science, especially YOLO (You Only Look Once) approach, we are able to incorporate unified, real-time object detection into cell state recognition in this paper, which realizes automotive process of detection and classification of CPE (cytopathic effect) states. In our work, we build up photomicrograph datasets of Vero cell in CPE state and train our YOLO model with it. Finally, our model learns characteristic pixel structure of cell rounding, syncytium and inclusion bodies, acting as end-to-end detector in cell infected experiments.