FogBank: A Single Cell Segmentation across Multiple Cell Lines and Image Modalities

Summary

The separation of touching cells in microscopy images is critical for the counting, identification and measurement of individual cells. Segmentation methods based on morphological watersheds are the current state-of-the-art for cell separation. However, over-segmentation of morphological watersheds is a major problem because of the high level of noise in microscopy cell images. We present a new segmentation method called FogBank that accurately separates cells when they are confluent and touching each other. Figure 1 illustrates segmentation outcomes for a set of algorithmic parameters.

Description

This technique has been successfully applied to phase contrast, bright field, and fluorescence microscopy images, as well as to binary images. FogBank method is based on the morphological watershed principles with two new features to improve the accuracy of related segmentation methods. First, to eliminate the pixel intensity noise that causes over-segmentation, our new method uses histogram binning to quantize the pixel intensities or pixel intensity gradients. We grow watersheds in increments of multiple pixel intensities rather than single intensities. Second, our method uses a geodesic distance mask derived from raw images to incorporate the shapes of individual cells, in contrast to the more linear cell edges that other watershed-like algorithms produce. The segmentation technique is fully automated and does not require any manual region seeding.

Major Accomplishments

An open-source tool is made available for free download. This technique was developed in Matlab, for which the source code is available for download along with a compiled executable.

The paper: J.Chalfoun et al., "Fog Bank: A Single Cell Segmentation across Multiple Cell Lines and Image Modalities", BMC BioInformatics, 2014 (download pdf) (view article)