Abstract

We present a multiresolution transform-based method for the extraction of moving filament trajectories from single molecule motility data. Noise-corrupted fluorescence image series are denoised using the multiscale median transform and trajectories are detected in the denoised data set. The presented method reduces noise more efficiently than 2D-anisotropic diffusion and several wavelet based techniques. Fibre trajectories are extracted by segmentation of the denoised image stacks and non-crossing trajectories are unambiguously identified combining the information of 2D (XY) and 3D (XYT) segmentation.

The algorithm is applied and evaluated using experimental data sets – image sequences of fluorescently labeled F-actin molecules and their 2D-trajectories on a myosin coated surface. This so-called ‘motility assay’ is used to analyse kinetics, biochemical regulation and pharmacological modulation of these biologically relevant molecules. The presented method improves signal-to-background discrimination, facilitates filament identification and finally, may contribute to significantly improve the performance of this assay.

© 2007 SPIE