EGT: an Automated Segmentation Method across Image Modalities and Cell Lines

Summary

We developed a novel and empirically derived image gradient threshold selection method for separating foreground and background pixels in an image. As illustrated in Figure 1, the method works across multiple image modalities and cell lines.

Description

New microscope technologies are enabling image acquisition of terabyte-sized (TB) datasets with hundreds of thousands of images. In order to retrieve and analyze the biological information in these large data sets, segmentation is needed to detect the regions containing cells or cell colonies. Our work with hundreds of large images (each 21 000 x 21 000 pixels) requires a segmentation method that: (1) yields high segmentation accuracy, (2) is applicable to multiple cell lines with various density of cells and cell colonies, and several imaging modalities, (3) can segment TB of image data per day, (4) requires low machine memory consumption, and (5) has a small number of user-settable parameters, with each parameter robust across an entire time-sequence of images, i.e., does not require any changes of parameters for individual images during the segmentation process of hundreds of large images. Therefore, we developed the Empirical Gradient Threshold (EGT) method, a novel and empirically derived image gradient threshold selection method for separating foreground and background pixels in an image that operates on the gradient image histogram and meets all five requirements.

Our method was validated on a reference dataset comprised of 501 validation images with manually determined segmentations and image sizes ranging from 0.36 Megapixels to 850 Megapixels. It includes four different cell lines and two image modalities: phase contrast and fluorescent. The EGT model is derived from this reference dataset with a 10-fold cross-validation method. EGT segments cells or colonies with resulting Dice index measurements above 0.92 for all cross-validation datasets. It has also been visually verified on a much larger dataset (in addition to the reference dataset) that includes: bright field and Differential interference contrast (DIC) images, 16 cell lines, and 61 time-sequence datasets, for a total of 17 479 images.

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. This tool was converted into a Fiji plugin available for download as a Jar file, or from a Fiji update site.

The paper: J.Chalfoun et al., "Empirical Gradient Threshold Technique for Automated Segmentation across Image Modalities and Cell Lines", Journal of Microscopy, 2014 (download pdf)