2Massachusetts Institute of Technology, Cambridge, MA, USA
Careful visual examination of biological samples is quite powerful, but many visual analysis tasks done in the laboratory are repetitive, tedious, and subjective. Here we describe the use of the open-source software, CellProfiler™, to automatically identify and measure a variety of biological objects in images. The applications demonstrated here include yeast colony counting and classifying, cell microarray annotation, yeast patch assays, mouse tumor quantification, wound healing assays, and tissue topology measurement. The software automatically identifies objects in digital images, counts them, and records a full spectrum of measurements for each object, including location within the image, size, shape, color intensity, degree of correlation between colors, texture (smoothness), and number of neighbors. Small numbers of images can be processed automatically on a personal computer, and hundreds of thousands can be analyzed using a computing cluster. This free, easy-to-use software enables biologists to comprehensively and quantitatively address many questions that previously would have required custom programming, thereby facilitating discovery in a variety of biological fields of study.
One of the most powerful methods in biology is the visual analysis of a sample. While nothing can fully replace the expertise of a trained biologist, observing many samples by eye is time-consuming, subjective, and nonquantitative. Certain repetitive tasks in visual analysis are suitable for automation by collecting digital images and processing them with image analysis software. This liberates biologists for more interesting work and has several advantages over visual observations including speed, quantitative and reproducible results, and simultaneous measurement of many features in the image. Efforts to automate visual analysis in biology began several decades ago, but many aspects still need improvement (1).
While numerous commercial and free software packages exist for image analysis, many of these packages are designed for a very specific purpose, such as cell counting (2). Other packages are sold with accompanying hardware for image acquisition (e.g., yeast colony counters), but these are expensive and do not allow measurement of features beyond those that are already built-in. Most commercial software is proprietary, meaning that the underlying methods of analysis are hidden from the researcher. At the other end of the continuum, some software packages are very flexible, especially for interactive analysis of individual images [e.g., Image-Pro Plus, MetaMorph®, and the open-source ImageJ/National Institutes of Health (NIH) Image (3)]. While users can program custom algorithms or record macros, these customized routines are challenging to adapt without knowing a programming language or interacting directly with the macro code.
The CellProfiler™ project was developed to address these software challenges by providing the scientific community with an easy-to-use open-source platform for automated image analysis. The compiled software is freely available for Macintosh®, PC, and Unix platforms at www.cellprofiler.org. It can accommodate adaptation to many biological objects and assays without requiring programming, due to its modular design and graphical user interface. There are many existing software packages available for specific applications in biology, but CellProfiler accomplishes many of the same goals in one open-source program. We recently described CellProfiler's use for cell identification, cell size, intensity and texture of fluorescent stains, cell cycle distributions, and other features of individual cells in images (4). Here we describe its use for a variety of other applications such as yeast colony counting, grid analysis, wound healing, and other visually quantifiable assays.Materials and Methods
All of the image analysis in this paper used the freely available CellProfiler cell image analysis software. The pipelines and images for these examples, as well as others, are available for download (www.cellprofiler.org/examples.htm). The image of yeast colonies (Figure 1) is a plate of Hi90-strain cells plated on synthetic defined medium with 128 µg/mL fluconazole as previously described (5). Images of Drosophila Kc167 cells on cell microarrays (Figure 2, A–C) were prepared as described previously (4,6). Briefly, spots of double-stranded RNA (dsRNA) were printed onto plain slides, and cells were grown on these slides for 3 days before being fixed, stained with Hoechst 33342, and imaged. Images of yeast patches (Figure 2, D–G) were prepared by manually spotting cells (with a 96-well pinning device) onto agar plates containing galactose to induce the expression of a-synuclein and a gene of interest. The cells were grown for 2 days at 30°C prior to imaging (7). Images of green fluorescent protein (GFP)-labeled mouse tumors (Figure 3, A–C) are faces of a mouse lung lobe, dissected out at 8 weeks post-tail vein injection of an established metastatic human cancer cell line overexpressing a gene of interest as described (Kimberly Hartwell, unpublished data). Images of wound healing (Figure 3, E–G) were prepared using MDA-MB-435 cells and imaged at the time points indicated (Lynne Waldman, personal communication). A Drosophila wing imaginal disc from a third larval instar (Figure 3, H–J) was stained with rhodamine-phalloidin to label F-actin, which is concentrated at points of cell-cell contact at the level of the adherens junction lattice (8).