Subcellular particles tracking in time-lapse confocal microscopy images.

Automatically tracking and analyzing the mobility of live subcellular structures will expedite the understanding of signaling pathways, protein-protein interaction, drug delivery, protein synthesis and functionality. Traditional computer vision tracking methods produce yet-to-be-satisfactory results due to the complexity of the particles recorded in spatial-temporal video sequences from confocal images. The difficulties arise from diverse modalities of motion patterns (translational, Brownian, or sessile), changes in behavior during tracking, and cluttered background. In this paper, we present an effective framework to detect and track subcullular particles in different motion modalities. The methodology begins with a Divergence Filter design for motion modality detection. After that, an improved a' trous wavelet is presented for segmenting particles. Represented by Euclidean Distance Map which contains information on object position, size, and intensity, the multiple particle tracking is carried out by solving a linear assignment problem. The proposed framework can also simultaneously evaluate particle population change by automatically counting the number of newly appeared or disappeared particles in time space.