One of the fundamental problems in the analysis of single particle tracking data is the detection of individual particle positions from microscopy images. Distinguishing true particles from noise with a minimum of false positives and false negatives is an important step that will have substantial impact on all further analysis of the data. A common approach is to obtain a plausible set of particles from a larger set of candidate particles by filtering using manually selected threshold values for intensity, size, shape, and other parameters describing a particle. This introduces subjectivity into the analysis and hinders reproducibility. In this paper, we introduce a method for automatic selection of these threshold values based on maximizing temporal correlations in particle count time series. We use Markov Chain Monte Carlo to find the threshold values corresponding to the maximum correlation, and we study several experimental data sets to assess the performance of the method in practice by comparing manually selected threshold values from several independent experts with automatically selected threshold values. We conclude that the method produces useful results, reducing subjectivity and the need for manual intervention, a great benefit being its easy integratability into many already existing particle detection algorithms.
Keywords: fluorescence microscopy; image analysis; optical microscopy; unsupervised learning.
Copyright © 2013 Wiley Periodicals, Inc.