Online Neural Cell Tracking Using Blob-Seed Segmentation and Optical Flow

Existing neural cell tracking methods generally use the morphology cell features for data association. However, these features are limited to the quality of cell segmentation and are prone to errors for mitosis determination. To overcome these issues, in this work we propose an online multi-object tracking method that leverages both cell appearance and motion features for data association. In particular, we propose a supervised blob-seed network (BSNet) to predict the cell appearance features and an unsupervised optical flow network (UnFlowNet) for capturing the cell motions. The data association is then solved using the Hungarian algorithm. Experimental evaluation shows that our approach achieves better performance than existing neural cell tracking methods.

1. Introduction

One primary aim of stem cell biology is to understand the factors influencing cell fate when multipotent cells become specified as terminal functional cell types. In the central nervous system, neural stem cells become specified as neurons, astrocytes, and oligodendrocytes. During this process, cells are constantly sampling their environment, making transient and long-term contacts with neighboring cells via filopodia and lamellipodia. Such behavior is typically recorded as time-lapse videos where the vision techniques could be applied for automatic analysis. In particular, cell tracking serves as an essential tool for the study of cell-cell interactions and thereby plays an important role in discovering the factors that influence the specified fate, and more importantly, the specific changes that correlate with disease progression [2].