Automatic detection of subcellular particles in fluorescence microscopy via feature clustering and bayesian analysis

Recent advancement in live cell fluorescence microscopy has enabled image acquisition at single particle resolution, through which biologists can investigate the underlying mechanisms of cellular processes. In this paper, we present a method to automatically detect the features of sub-cellular particles in 2D fluorescence images, including x-y positions, fluorescence intensities, and relative sizes. The method consists of two parts. One is an initial detection method, which finds particle candidates in the images using image filters and clustering algorithms. The other is a MAP-Bayesian based estimation method, which provides the optimal estimations of particle features. The method is evaluated on synthetic data and results show that it has high accuracy. The results on real data confirmed by human expert cell biologists are also presented.