CellSegm - a MATLAB toolbox for high-throughput 3D cell segmentation - PubMed (original) (raw)

CellSegm - a MATLAB toolbox for high-throughput 3D cell segmentation

Erlend Hodneland et al. Source Code Biol Med. 2013.

Abstract

: The application of fluorescence microscopy in cell biology often generates a huge amount of imaging data. Automated whole cell segmentation of such data enables the detection and analysis of individual cells, where a manual delineation is often time consuming, or practically not feasible. Furthermore, compared to manual analysis, automation normally has a higher degree of reproducibility. CellSegm, the software presented in this work, is a Matlab based command line software toolbox providing an automated whole cell segmentation of images showing surface stained cells, acquired by fluorescence microscopy. It has options for both fully automated and semi-automated cell segmentation. Major algorithmic steps are: (i) smoothing, (ii) Hessian-based ridge enhancement, (iii) marker-controlled watershed segmentation, and (iv) feature-based classfication of cell candidates. Using a wide selection of image recordings and code snippets, we demonstrate that CellSegm has the ability to detect various types of surface stained cells in 3D. After detection and outlining of individual cells, the cell candidates can be subject to software based analysis, specified and programmed by the end-user, or they can be analyzed by other software tools. A segmentation of tissue samples with appropriate characteristics is also shown to be resolvable in CellSegm. The command-line interface of CellSegm facilitates scripting of the separate tools, all implemented in Matlab, offering a high degree of flexibility and tailored workflows for the end-user. The modularity and scripting capabilities of CellSegm enable automated workflows and quantitative analysis of microscopic data, suited for high-throughput image based screening.

PubMed Disclaimer

Figures

Figure 1

Figure 1

Software design of

CELLSEGM

. The solid box surrounds the processing steps occuring in

CELLSEGM

, from image conversion until the post-analysis of the segmentation data. The batch processing

cellsegmentation

is the tool for cell segmentation of high-throughput data. The quality of the resulting segmentation can be assessed in

v

iewsegm, and the processing chain can be restarted on demand (dashed line) with other parameter settings. The separate functions (m-files) can also be executed independently.

Figure 2

Figure 2

Files and folders (inside rectangles) connected to

CELLSEGM

. The BDA license file (

readme.txt

) defines the legal rights and the startup file (

startupcellsegm.m

) is used for setting the path in

MATLAB

to enable

CELLSEGM

. There are four folders at the highest level, containing example files used in this work for demonstration, example data loaded by the example files, the mfiles contained in the class @CellSegm, and one folder containing additional utility files necessary for

CELLSEGM

.

Figure 3

Figure 3

Segmentation of cells in 2D using automatically detected markers in Example 5. A) Raw surface stain, B) automatically detected markers, C) ridge enhanced surface stain, D) watershed image, E) detected cells.

Figure 4

Figure 4

A selection of available subroutines (magenta) and parameter settings (italic) in

CELLSEGM

related to the struct prm . In brackets is the supported data type. For example, to set the threshold

th

in the nucleus method of

segmsurf

, with thresholding method

thrs

, follow the stream of methods

segmsurf

→

getminima

→

nucleus

→

segmct

→

thrs

, and assign it by

prm.segmsurf.getminima.nucleus.thrs.th

in the parameter file. On the other hand, if

segmct

is called upon from the command prompt, the same parameter setting is defined by

prm.thrs.th

.

Figure 5

Figure 5

Smoothing of PC12 cells in 2D by Example 3. A) Raw surface stain, B) smoothing by coherence enhancing diffusion (

method = ’ced’

), C) directional coherence enhancing diffusion (

method = ’dirced’

), and D) Gaussian smoothing (

method = ’gaussian’

). The sharpness is better preserved by the anisotropic filters (B and C), which makes them more suitable for the enhancement of surface stained cells.

Figure 6

Figure 6

Smoothing of stained nuclei of Hela-Kyoto cells in 2D by Example 4. A) Raw nuclei stain, B) smoothing of A by edge enhancing diffusion (

method = ’eed’

). After edge enhancing diffusion the image becomes more piecewise constant and better suited for segmentation.

Figure 7

Figure 7

Segmentation of cells in 3D using automatically detected markers in Example 6. A) Raw surface stain, B) automatically detected markers, C) smoothed and ridge enhanced surface stain, D) watershed image, E) detected cells.

Figure 8

Figure 8

3D view of the segmentation in Example 6. The obtained segmentation is truly a 3D segmentation field. For visualization purposes the cells have been cut at plane 20.

Figure 9

Figure 9

Segmentation of cells using nucleus markers in 2D from Example 7, executed for plane five in the image stack. A) Raw surface stain, B) raw nucleus stain, C) surface stain minus nucleus stain, D) markers (blue) derived from the nucleus stain superimposed onto the surface stain, E) cell markers, F) smoothed segmentation image, from C, G) watershed image, H) detected cell areas.

Figure 10

Figure 10

Segmentation of cells using nucleus markers in 3D from Example 8, visualized for plane two. A) Raw surface stain, B) raw nucleus stain, C) surface stain minus nucleus stain, D) markers (blue) from nucleus stain superimposed on the surface stain, E) cell markers, F) smoothed input image, from C, G) watershed image, H) detected cells. All cells have been detected.

Figure 11

Figure 11

Segmentation of cells using manually drawn markers in 3D from example 9, visualized for plane five. A) Raw surface stain, B) smoothed surface stain used for segmentation, C) cell markers drawn manually, D) background markers (in an imaging plane other than the cell markers), E) watershed image, F) detected cells.

Figure 12

Figure 12

Segmentation of Hoechst stained Hela-Kyoto nuclei in 2D using

segmct

from Example 10. A) Input image showing stained nuclei, B) after edge enhancing diffusion, C) segmentation by adaptive thresholding (

prm.method = ’adth’

) without splitting of cells, and D) with splitting of cells. Note that the connected nuclei are now disconnected. E) Segmentation by iterative thresholding (

prm.method = ’thrs’

) without splitting, and F) after splitting. Both methods are successfull.

Figure 13

Figure 13

Segmentation of Hoechst stained nuclei in 3D using

segmct

, from Example 11. A) Raw nucleus stain, B) segmentation with iterative thresholding (

prm.method = ’itth’

) without splitting of cells, C) after splitting of cells. Note that after splitting several connected nuclei are disconnected into their separate compartments.

Figure 14

Figure 14

A batch processing of two data sets from two experimental conditions, as described in Example 12. The data is visualized for plane seven. A1-A4) Data set one and two in the two conditions. B1-B4) Segmentation using

CELLSEGM

. C1-C4) Segmentation using

CELLPROFILER

. For the strongly stained cells,

CELLPROFILER

provides a larger segmentation than

CELLSEGM

. For the weakly stained cells,

CELLPROFILER

is missing large cell fractions compared to

CELLSEGM

, probably due to uneven illumination. However, a correction of the uneven illumination pattern uneven did not improve the results (data not shown). For visualization, the objects segmented in

CELLPROFILER

where eroded by one voxel to highlight the contours.

Figure 15

Figure 15

Segmentation of a paraffine embedded human skin biopsy. Upper row: Light microscopical image of the whole sample, visible are the layers of the epidermis and dermis, including a part of a hair follicle. Stained are CD44 (red) and p53 (brown). Middle row: One plane of a 3D confocal fluorescence image stack of CD44 (VulcanRed; white), overlaid with the segmentation from

CELLSEGM

(red). For visualization purposes, the contours were dilated with a structural element of one pixel radius, and then closed with a structural element of four pixel radius. Lower row: Segmentation results using

CELLSEGM

(no dilation and no closing here). The segmentation is essentially confined to the cells expressing the marker at the plasma membrane to a sufficient amount.

Similar articles

Cited by

References

    1. Sacan A, Ferhatosmanoglu H, Cosku H. CellTrack: an open-source software for cell tracking and motility analysis. Bioinformatics. 2008;24(14):1647–1649. doi: 10.1093/bioinformatics/btn247. - DOI - PubMed
    1. Tscherepanow M, Zöllner F, Kummert F. Advances in Mass Data Analysis of Images and Signals in Medicine, Biotechnology, Chemistry and Food Industry (MDA 2008) Springer Lecture Notes in Computer Science Volume 5108; 2008. Automatic segmentation of unstained living cells in bright-field microscope images; pp. 158–172.
    1. Wu K, Gauthier D, Levine MD. Live cell image segmentation. IEEE Trans Biomed Eng. 1995;42:1–12. doi: 10.1109/10.362924. - DOI - PubMed
    1. Yu D, Pham TD, Yan H, Whang B, Crane DI. Segmentation of cultured neurons using logical analysis of grey and distance differences. J Neurosci methods. 2007;166:125–137. doi: 10.1016/j.jneumeth.2007.06.028. citeseer.ist.psu.edu/wu95live.html. - DOI - PubMed
    1. Wählby C, Lindblad J, Vondrus M, Bengtsson E, Björkesten L. Algorithms for cytoplasm segmentation of fluorescence labelled cells. Anal Cell Pathol. 2002;24:101–111. - PMC - PubMed

LinkOut - more resources