Computational image analysis of subcellular dynamics in time-lapse fluorescence microscopy

Abstract

The use of image segmentation and motion tracking algorithms was adapted for analyzing time-lapse data of cells with fluorescently labeled protein. Performance metrics were devised and algorithm parameters were matched to hand-created ground-truth data. The performance of these algorithms in this domain was compared. Finally, the optimal algorithms were selected and used to acquire statistics on existing data, in order to reproduce previous studies on the cell cytoskeleton. New data was acquired to extend previous results and further test the algorithms on a different cell line, under both widefield and confocal microscope conditions.