Particle Tracking In Low SNR Fluorescence Microscopy Images

Particle tracking in fluorescence microscopy images plays an important role in the field of qualitative and quantitative analysis of intracellular dynamics.Usually,there are dozens or even hundreds of micro particles in the camera sight.It is impossible to simultaneously track all the particles manually,so automatic or slightly interactive tracking algorithms are essential for this task.However,fluorescence microscopy images usually suffer from low signal-to-noise ratio(SNR)and low contrast ratio due to photo-bleaching and photo-toxicity.This is one of the main challenges for tracking fluorescence tagged particles in live cells.Most of the popular tracking methods achieve good performance in images with high quality,but none of them get satisfying results when SNR and contrast is low.Conventional methods fail because they highly depend on the results of detection and prior knowledge of particles,dynamic models.To reliably track particles in the low SNR images,we follow the most widely used'Detection&Linking' strategy.We did pre-processing for denoising and enhancement.Particle objects were detected through segment operation.In the linking step,we first implemented a conventional method using motion filter and tried to optimize the algorithm with some constraints and post-processing.Then we proposed a novel method based on minimal path theory to extract complete trajectories.For the first time,we introduced a new definition of topology and the corresponding neighborhood,and solved the difficulty of propagation and back propagation in the spatial-temporal volume constructed from the image sequence of particles 'movement.We also optimized Fast Marching Method(FMM)to conform to the characteristics of particles'appearance and dynamics.The two tracking methods were evaluated and compared on several simulated image sequences and samples of experiment data.Our minimal path based method showed its accurac y and robustness in the task of particle tracking and achieved better performance in low SNR scenario.