to BioTechniques free email alert service to receive content updates.
CellProfiler™: free, versatile software for automated biological image analysis
Michael R. Lamprecht1, David M. Sabatini1,2, and Anne E. Carpenter1
Full Text (PDF)

Grid Analysis

Many experiments are designed in a grid format, such as cell microarrays, agar plates of yeast patches, and multiwell plates. In experiments with large or numerous grids, it is difficult to identify which reagent corresponds to a spot on this grid. CellProfiler can manually or automatically place a grid on an image and associate each grid location with an annotation such as a sample number or gene name. Each grid location can also be identified as an object, and all available measurements can be made on those objects.

For example, on a cell microarray, more than 5000 individual clusters of cells are grown on a single microscope slide (4,6,20,21,22). Each cluster has been treated with a perturbant, which could be a small molecule, an overexpression plasmid, or an RNA interference reagent. The goal is to determine which perturbants alter cells. When attempting to quantify even a simple phenotype such as the extent of cell death in a spot, it is difficult to determine which spot correlates to which reagent. In this example, CellProfiler places a grid on top of an image (Figure 2B) in a position defined by the user, who specifies the location of known control spots on the array and the number and size of rows and columns. The LoadText module allows the user to load a text file with corresponding sample information for each of the spots on the grid, and the DisplayGridInfo module allows this imported data to be assigned to each of the grid locations (Figure 2C). This quickly allows the user to associate a location in the grid with its reagent. For example, knockdown of the cytokinesis-related gene CG10522 (sticky) shows an unusual large, bright-nuclei phenotype that is visible at low resolution (Figure 2C).

Plates of yeast patches also make use of grids (Figure 2E). Although large screens of yeast patches have been analyzed by eye (23), the number of these screens in progress is rapidly increasing, and visual analysis cannot keep up with the rapid pace at which samples are being generated. Furthermore, quantitative analysis is much preferred, because subtle changes in growth can be identified and the screen can be analyzed statistically. Some software for this application has been developed, but none to our knowledge is fully automated, open-source, and flexible to new assays/unusual measurements.

Because thousands of plates are typically analyzed, the entire process of finding the grid and making measurements is performed automatically by CellProfiler. In the pipeline for this analysis, the images are cropped to remove the plate edges, and any yeast patches that are present are identified. Unlike the cell microarray example, a grid is then defined automatically based on the yeast patches that are identified. This allows for nonuniformity in the precise placement of the grid on the plate, to allow for experimental variation. For this function to work correctly, none of the outside rows or columns can be completely blank. This condition can be satisfied if most patches tend to grow well in the experiment or if control patches exist in two or more opposite corners. These yeast patches can be analyzed in their naturally identified shapes if the patch size or shape is of interest (Figure 2F). Alternately, a circle can be forced into the location of the identified objects to measure, for example, the intensity of each patch, which is a measure of growth (Figure 2G). Once the grid and objects are identified, all available measurements can be calculated for each patch, including measures of growth (Figure 2, F–G, bottom).

Tumor Counting

When tumorigenic cells are labeled with GFP and injected into mice, the resulting tumors in the lungs are readily visible by fluorescence microscopy of the dissected lungs. Accurate, objective quantification of the number and size of the resulting tumors is necessary to understand the process of tumor metastasis (Kimberly Hartwell, unpublished data). The GFP signal from each tumor can be identified by CellProfiler (Figure 3, A–C). CellProfiler counts of identified tumors were comparable to counting by eye (Figure 3D). If the GFP brightness or the shape or texture/smoothness of the tumors is of interest, these measurements can also be recorded.

Wound Healing

The wound healing assay is a standard technique to determine the migration of different cell types in different conditions. In this assay, a confluent monolayer of cells is wounded by scratching it with a pipet tip (24). The monolayer is then imaged at time points to record the size of the wound. In this example, the area of the images covered by cells is calculated by CellProfiler (Figure 3, E–G). While this is not a particularly challenging application, the structure of CellProfiler makes it simple to carry out this quantitative analysis for hundreds of thousands of images, enabling high-throughput screens. In addition, the shape characteristics of the wound border can be measured; for example, to distinguish between samples where all cells have steadily grown toward the middle versus samples where a few individual cells extend into the wound space.

Tissue Topology

In a developing tissue or at other sites of cell-cell contact (e.g., tumors and surrounding stromal cells), it is useful to determine the number of neighbors each cell has to better understand the processes underlying the topology (25). CellProfiler can identify cells in tissues (Figure 3, H–J). In addition to typical measurements, the MeasureObjectNeighbors module can determine the number of cells neighboring each cell and record this measurement. The cells can then be color-coded by how many neighbors it has (Figure 3J), or the data can be exported to further analyze the topology of the tissue.


CellProfiler is a flexible platform that can automate the analysis of images to address a wide variety of biological questions. For many assays, described here and previously (4), it eliminates the tedium of repetitive visual analysis and produces rapid, quantitative, and accurate results. The modular design of the software provides an infrastructure for image analysis that is applicable to diverse assays. Its open-source code allows programmers to design and contribute new algorithms to the project. It is our hope that CellProfiler will become a widely used platform, through which advanced algorithms are made conveniently available for automatic biological image analysis.


We gratefully acknowledge contributions of images from researchers at the Harvard Medical School: Matt Gibson (Figure 3, H–J) and at the Whitehead Institute for Biomedical Research and Massachusetts Institute of Technology: Leah Cowen (Figure 1), Douglas B. Wheeler (Figure 2, A–C), Aaron Gitler (Figure 2, D–G), Kimberly A. Hartwell (Figure 3, A–D), and Lynne K. Waldman (Figure 3, E–G). We also thank the other members of the CellProfiler project software team for software development and lab members for helpful comments: Robert A. Lindquist, Shomit Sengupta, and David A. Guertin. This work was supported by a Merck/CSBi postdoctoral fellowship (A.E.C.), a Novartis fellowship from the Life Sciences Research Foundation (A.E.C.), a Society for Biomolecular Screening Academic grant (A.E.C.), Department of Defense (DOD) Tuberous Sclerosis Complex (TSC) research program grant no. W81XWH-05-1-0318-DS (D.M.S.), NIH grant no. R01 GM072555-01 (D.M.S.), and the Keck Foundation (D.M.S.).

Competing Interests Statement

The authors declare no competing interests.

1.) Murphy, R.F., E. Meijering, and G. Danuser. 2005. Special issue on molecular and cellular bioimaging. IEEE Trans. Image Process 14:1233-1236.

2.) Selinummi, J., J. Seppala, O. Yli-Harja, and J.A. Puhakka. 2005. Software for quantification of labeled bacteria from digital microscope images by automated image analysis. BioTechniques 39:859-863.

3.) Abramoff, M.D., P.J. Magalhaes, and S.J. Ram. 2004. Image processing with ImageJ. Biophotonics Int. 11:36-42.

4.) Carpenter, A.E., T.R. Jones, M.R. Lamprecht, C. Clarke, I.H. Kang, O. Friman, D.A. Guertin, J.H. Chang CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol..

5.) Cowen, L.E., and S. Lindquist. 2005. Hsp90 potentiates the rapid evolution of new traits: drug resistance in diverse fungi. Science 309:2185-2189.

6.) Wheeler, D.B., S.N. Bailey, D.A. Guertin, A.E. Carpenter, C.O. Higgins, and D.M. Sabatini. 2004. RNAi living-cell microarrays for loss-of-function screens in Drosophila melanogaster cells. Nat. Methods 1:127-132.

7.) Cooper, AA., A.D. Gitler, A. Cashikar, C.M. Haynes, K.J. Hill, B. Bhullar, K. Liu, K. Xu. 2006. α-Synuclein blocks Ergolgi traffic and Rab1 rescues neuron loss in Parkinson's models. Science 313:324-328.

8.) Gibson, M.C., A.B. Patel, R. Nagpal, and N. Perrimon. 2006. The emergence of geometric order in proliferating metazoan epithelia. Nature Aug 9; [Epub ahead of print].

9.) Jones, T.R., A.E. Carpenter, and P. Golland. 2005. Voronoi-based segmentation of cells on image manifolds. ICCV Workshop on Computer Vision for Biomedical Image Applications:535-543.

10.) Jones, T.R., A.E. Carpenter, P. Golland, and D.M. Sabatini Methods for high-content, high-throughput image-based cell screening. MIAAB 2006 Workshop Proceedings.

11.) Malpica, N., C.O. de Solorzano, J.J. Vaquero, A. Santos, I. Vallcorba, J.M. Garcia-Sagredo, and F. del Pozo. 1997. Applying watershed algorithms to the segmentation of clustered nuclei. Cytometry 28:289-297.

12.) Meyer, F., and S. Beucher. 1990. Morphological segmentation. J. Vis. Commun. Image Rep. 1:21-46.

13.) Ortiz de Solorzano, C., E.G. Rodriguez, A. Jones, D. Pinkel, J.W. Gray, D. Sudar, and S.J. Lockett. 1999. Segmentation of confocal microscope images of cell nuclei in thick tissue sections. J. Microsc (Oxford) 193:212-226.

14.) Wahlby, C. 2003.Algorithms for applied digital image cytometry. Center for Image Analysis. Uppsala University, Uppsala:75.

15.) Wahlby, C., I.M. Sintorn, F. Erlandsson, G. Borgefors, and E. Bengtsson. 2004. Combining intensity, edge and shape information for 2D and 3D segmentation of cell nuclei in tissue sections. J. Microsc. 215:67-76.

16.) Vincent, L., and P. Soille. 1991. Watersheds in digital spaces—an efficient algorithm based on immersion simulations. IEEE Trans. Pattern Anal. Mach. Intell. 13:583-598.

17.) Boland, M.V., M.K. Markey, and R.F. Murphy. 1998. Automated recognition of patterns characteristic of subcellular structures in fluorescence microscopy images. Cytometry 33:366-375.

18.) Gabor, D. 1946. Theory of communication. J. Instit. Electr. Engineer. 93:429-441.

19.) Haralick, R.M., K. Shanmuga, and I. Dinstein. 1973. Textural features for image classification. IEEE Trans. Syst. Man. Cybern. SMC3:610-621.

20.) Bailey, S.N., D.M. Sabatini, and B.R. Stockwell. 2004. Microarrays of small molecules embedded in biodegradable polymers for use in mammalian cell-based screens. Proc. Natl. Acad. Sci. USA 101:16144-16149.

21.) Wheeler, D.B., A.E. Carpenter, and D.M. Sabatini. 2005. Cell microarrays and RNA interference chip away at gene function. Nat. Genet. 37(Suppl):S25-S30.

22.) Bailey, S.N., S.M. Ali, A.E. Carpenter, C.O. Higgins, and D.M. Sabatini. 2006. Microarrays of lentiviruses for gene function screens in immortalized and primary cells. Nat. Methods 3:117-122.

23.) Measday, V., K. Baetz, J. Guzzo, K. Yuen, T. Kwok, B. Sheikh, H. Ding, R. Ueta. 2005. Systematic yeast synthetic lethal and synthetic dosage lethal screens identify genes required for chromosome segregation. Proc. Natl. Acad. Sci. USA 102:13956-13961.

24.) Yu, A.C., Y.L. Lee, and L.F. Eng. 1993. Astrogliosis in culture: I. The model and the effect of antisense oligonucleotides on glial fibrillary acidic protein synthesis. J. Neurosci. Res. 34:295-303.

25.) Classen, A.K., K.I. Anderson, E. Marois, and S. Eaton. 2005. Hexagonal packing of Drosophila wing epithelial cells by the planar cell polarity pathway. Dev. Cell 9:805-817.

  1    2    3