Tracking Multiple Particles in Fluorescence Time-Lapse Microscopy...
W. J. Godinez, K. Rohr
IEEE Trans Med Imaging, vol. 34, issue 2, February 2015
Automatically tracking the movements of biological particles such as virus particles or cell vesicles in microscopy images is important to understand intracellular dynamic processes. We have developed a probabilistic approach for tracking multiple particles in fluorescence microscopy image sequences. Our approach combines a localization scheme that uses multiple bottom-up detections as well as a top-down ellipsoidal sampling scheme based on Gaussian probability distributions determined by a Kalman filter. Compared to random sampling used in particle filters, the ellipsoidal sampling scheme requires fewer samples which reduces the computation time. For position estimation, the multiple measurements from the localization scheme are integrated into the Kalman filter via principles of the probabilistic data association algorithm. To track objects in close proximity, we propose an algorithm which computes the image support for each object relative to its neighbors. In addition, we incorporate multiple motion models in our approach.
The approach has been successfully applied to synthetic 2D and 3D images as well as to real 2D and 3D microscopy images. In addition, the approach was successfully applied to the 2D and 3D image data of the Particle Tracking Challenge at the IEEE International Symposium on Biomedical Imaging (ISBI) 2012 (Nature Methods 2014). Among 14 different approaches, our approach yielded the highest number of top-3 rankings. Currently, the tracking approach is used in biological studies, for example, to provide detailed spatial-temporal information about HIV (human immunodeficiency virus) and HCV (hepatitis C virus) particles to better understand virus infection.