Motion analysis of nanoscale intracellular objects, commonly studied using fluorescence microscopy imaging, requires tracking of large and time-varying numbers of spots in noisy image sequences. Conventional approaches to tracking in molecular cell biology typically consist of two separate stages: object detection and temporal data association. Recent comparison studies have shown common techniques based on this principle to fail in practical cases of poor imaging conditions. In this project we develop new techniques for multiple-object tracking based on nonlinear Bayesian approaches. Since these better integrate available temporal information and application-specific prior knowledge, they can be shown to perform superiorly.
Very fast and sensitive algorithms have been developed for single molecule detection in fluorescence microscopy images. In addition we have developed tools for the generation of synthetic but realistic images used in the international particle tracking challenge that we are co-organizing. The latter aims to objectively compare all state-of-the-art algorithms for particle tracking in the field.
Figure: Procedure used in the particle tracking challenge for generating images simulating different imaging scenarios, motion models, and microscopes.
- [Erik Meijering]
- [Ihor Smal]