Biological Cell Tracking via Multi-Agent Identification and Filtering

Understanding cellular dynamics is a fundamental topic in different biomedical applications. Nowadays, optical microscopy is one of the most used techniques to visualize cell movements. In this paper, we consider a novel cell-tracking algorithm to track multiple cells in optical microscopy videos. The proposed methodology combines two steps. First, we model cell movements and their neighboring interactions according to tailored nonlinear multi-agent systems. Then, we identify model parameters from real cellular trajectories and predict cell movements across different frames of a video. In particular, we use an Extended Kalman Filter that exploits the distributed nature of cell dynamics. Numerical experiments on videos from the Cell Tracking Challenge dataset are performed to validate the proposed method and performance metrics are shown.