Using Deep Learning to Identify Cell and Particle in Live-Cell Time-lapse Images

Live-cell time-lapse images generated by biological experiments are useful for observing activities, even for proposing novel hypotheses. In past work, we had proposed a particle-cell relation mining method, abbreviate to PCRM, which involved identifying particles and cells as objects from live-cell time-lapse images at first. Then PCRM is used to track the pathways of particles to calculate the measures as distances between the particles and cells. Finally, the relationship of particles and cells can be quantified by PCRM. The PCRM is useful for biologists to prove their hypotheses. However,it is very time-consuming when identifying the objects among a large number of biological images. Hence, in this paper, we propose a method using deep learning technology, abbreviated to PCOD, to accelerate the particle and cell identification. The PCOD method achieves the accuracies of 90.2% and 99.9% for particles and cells identification, respectively. In this way, the overall particles and cells can be identified in real time.