Automatic tracking of biological particles
Scientists at the BioQuant centre in Heidelberg have developed an automatic particle tracking method that can be used for time-resolved two- and three-dimensional microscope image data. This powerful computational method achieved the best overall result in an international competition that compared different methods for the quantitative tracking of the position of moving biological particles.
Two years ago, we presented the computer programme “QuantVessel” developed by PD Dr. Karl Rohr and his team at the University of Heidelberg’s BioQuant centre. QuantVessel attracted huge attention because of its efficiency in diagnosing and treating vascular diseases. The software is able to accurately determine the size and shape of blood vessels that move in rhythm with the heartbeat from three-dimensional images (see article of 19th March 2012: Quantification of the morphology of human blood vessels from 3D tomographic image data; link on the right-hand side). The time-resolved quantification of particles in two- and three-dimensional image data is also at the basis of a particle tracking method with which Heidelberg researchers achieved the most accurate results of all the methods presented at an international competition organized as part of the “2012 IEEE International Symposium on Biomedical Imaging” in Barcelona, Spain. The researchers’ powerful analysis method enables the tracking and automatic analysis of the trajectories of the smallest biological particles in microscope images of live cells despite the presence of strong background noise.
“Particle Tracking Challenge“
The results of the Particle Tracking Challenge, in which fourteen bioinformatics teams, including researchers from highly renowned universities such as Yale and Stanford participated, were recently published in the prestigious journal “Nature Methods”.
In the competition, the different image analysis methods were applied to a broad range of two- and three-dimensional image data in order to automatically track the movement of a large number of differing biological particles such as viruses, membrane vesicles and cell receptors. The performance of each method was quantified according to different criteria and the three best methods were determined for each data category. With a total of 150 “Top 3 Rankings”, Dr. Karl Rohr and his colleague Dr. William K. Godinez’s team achieved the best overall result. The second and third placed teams achieved a total of 124 and 103 “Top 3 Rankings”, respectively.
In contrast to conventional “deterministic” methods, the method adopted by the Heidelberg researchers is a probabilistic approach, which means that it is based on a mathematically sound method from probability theory that takes into account uncertainties in the image data, e.g. caused by noise, and exploits knowledge from the application domain. “Compared to deterministic methods, our probabilistic particle tracking approach achieves high levels of accuracy, especially for complicated image data with a large number of objects, high object density and a high level of noise,” explains Rohr. The method is suitable for multi-channel, time-resolved, two-dimensional and three-dimensional microscope images and enables the determination of the trajectories (i.e. movement paths) of objects as well as the quantification of relevant parameters such as speed, path length, motion type or object size. In addition, the probabilistic approach automatically detects important dynamic events such as vesicle-vesicle or virus-cell membrane fusions.
Applications in cell biology and medicine
The tracking of biological particles such as viruses, cell vesicles, granules or internalized receptors in live-cell microscope images is of key importance for the quantitative analysis of intracellular dynamic processes. It is obvious that manually analyzing time-resolved microscope images with hundreds or even thousands of moving objects is not feasible. In recent years, increasing emphasis has therefore been placed on the development of automatic image analysis methods for particle tracking. These methods are computer-based and determine the positions of particles over time. Of all these methods, the probabilistic particle tracking method developed by Drs. Godinez and Rohr stands out for its very high accuracy and reliability in recognizing biological objects and therefore represents huge progress in the tracking of moving objects. The method is being adopted in cooperative projects undertaken in cooperation with research groups in Heidelberg and elsewhere, including research groups from the Department of Infectious Diseases at Heidelberg University Hospital: those led by Prof. Dr. Ralf Bartenschlager and Dr. Alessia Ruggieri that address hepatitis C virus infections, and other groups led by Prof. Dr. Hans-Georg Kräusslich, PD Dr. Barbara Müller and Prof. Dr. Oliver Fackler that are specifically focused on human immunodeficiency virus (HIV) infections. In addition, the renowned microbiologist Prof. David Knipe from Harvard Medical School in Boston and cell biologist David L. Spector from the Cold Spring Harbor Laboratory, New York, are working with the bioinformaticians from Heidelberg.
The “Biomedical Computer Vision” (BMCV) research group led by Dr. Karl Rohr develops computer science methods to automatically analyze microscope images used for basic cell biological research as well as radiological images used for medical applications. The BMCV research group is located at the BioQuant centre at Heidelberg University and is part of the Department of Bioinformatics and Functional Genomics at Heidelberg University’s Institute of Pharmacy and Molecular Biotechnology as well as the DKFZ’s Division of Theoretical Bioinformatics, both of which are headed by Prof. Dr. Roland Eils. Dr. William J. Godinez, who is pursuing postdoctoral work in the BMCV group, was instrumental in the development of the method that achieved best overall results at the Particle Tracking Challenge in Barcelona in 2012.
Chenouard N, Smal I, de Chaumont F, Maška M, Sbalzarini IF, Gong Y, Cardinale J, Carthel C, Coraluppi S, Winter M, Cohen AR, Godinez WJ, Rohr K, Kalaidzidis Y, Liang L, Duncan J, Shen H, Xu Y, Magnusson KEG, Jaldén J, Blau HM, Paul-Gilloteaux P, Roudot P, Kervrann C, Waharte F, Tinevez J-Y, Shorte SL, Willemse J, Celler K, van Wezel GP, Dan H-W, Tsai Y-S, Ortiz de Solórzano C, Olivo-Marin J-C, Meijering E. Objective comparison of particle tracking methods. Nature Methods, March 2014, Volume 11, Issue 3, 281–289; DOI: 10.1038/nmeth.2808