Detection and Tracking Protein Molecules in Fluorescence Microscopic Video

This paper provides a bioimage informatics system of detecting and tracking protein molecules, called APP-GFPs, in a live-cell video captured by a fluorescent microscope. Since both processes encounter many difficulties such as many targets, less appearance information, and heavy background noise, we will try to design the system as robust as possible. Specifically, for the detection, a machine learning-based method is employed. For tracking, a method based on a global optimization strategy is newly developed. Experimental results showed that the speed and direction distributions of molecular motion by the proposed system were very similar to that by manual inspection.