Semi-automatic single particle segmentation on electron micrographs

The selection of single particles from an electron microscopic (EM) micrograph is an essential part of the reconstruction process of biological macromolecules. Despite its importance, this process requires a lot of user intervention by either purely manual or semi-automatic processing (with the aid of graphical interfaces). Since reconstructing a 3D model of these macromolecules at nano-scale resolution requires thousands of particles, the particle selection phase is bound to become a serious bottleneck in the reconstruction process. In this article we propose a semi-automatic procedure for selecting particles from a micrograph that aims to reduce the false positive rate. We achieve this by first using a fuzzy-sets-based segmentation method, that requires very little user intervention, to detect the background, and then calculating a cross-correlation measure only on points with low "affinity" to the micrograph's background.