Skip to main content
U.S. flag

An official website of the United States government

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

3D Segmentation of Stem Cells from Thousands of Confocal Microscopy Images

Published

Author(s)

Peter Bajcsy, Mylene H. Simon, Stephen J. Florczyk, Carl G. Simon Jr., Derek Juba, Mary C. Brady

Abstract

We address the problem of estimating 3D segmentation performance when segmentation is applied to thousands of confocal microscopy images (z-stacks) of cells. With a very large number of z-stacks, manual inputs to validate each segmentation result are clearly labor prohibitive. Thus, we characterize segmentation performance in three steps. First, we design six candidate segmentation methods based on imaging and geometrical assumptions about cells. Next, we evaluate the segmentation methods against manual segmentations of statistically representative z-stacks. In order to minimize the overall manual labor, we introduce sampling and orthogonal projections. Finally, we verify the quality of segmentations visually in 2D and 3D by using comparative and multi-resolution visualization approaches. We demonstrate the methodology by applying it to a data set of 1253 mesenchymal stem cells (134,804 files) that were stained for actin and reside on 10 different types of biomaterial scaffolds. After constructing and evaluating six candidates of 3D segmentation methods, the most accurate 3D segmentation method achieved an average precision of 0.82 and accuracy of 0.84 measured by a Dice similarity index, the probability of segmentation success 0.85 based on visual verification, and the computational efficiency of 42.3 h to process all z-stacks. While the most accurate segmentation was 4.2 times slower than the second most accurate method it consumed on average 9.65 times less memory per z-stack segmentation. We observed consistency of the segmentation accuracy estimates measured by using quantitative evaluation and qualitative visual verification approaches. All segmented cells are available for 3D viewing and for downloading in a web browser at https://isg.nist.gov
Citation
Journal of Microscopy

Keywords

3D segmentation, segmentation evaluation, sampling, visual verification, confocal imaging, stem cells

Citation

Bajcsy, P. , Simon, M. , Florczyk, S. , Simon, C. , Juba, D. and Brady, M. (2016), 3D Segmentation of Stem Cells from Thousands of Confocal Microscopy Images, Journal of Microscopy (Accessed May 19, 2024)

Issues

If you have any questions about this publication or are having problems accessing it, please contact reflib@nist.gov.

Created March 1, 2016, Updated February 19, 2017