Written and maintained by CASRAI Editorial Board
Last updated
Short answer
CellProfiler is image analysis software from the Broad Institute that lets biologists measure cells in microscope images without writing code. You build a pipeline of image-processing modules, adjust their settings to measure a phenotype, run it over a set of images, and export the measurements to a spreadsheet or database. Its companion, CellProfiler Analyst, is for exploring those measurements and classifying phenotypes.
Definition and scope
The original 2006 paper in Genome Biology (Carpenter et al.) presented CellProfiler as the first free, open-source system for flexible, high-throughput cell image analysis. The motivation it gave was automation: biologists could by then prepare and image thousands of samples per day, which made chemical screens possible but left the measurement step as the bottleneck. The software quantifies standard assays such as cell count, cell size and per-cell protein levels, and more complex morphological assays such as cell or organelle shape and subcellular staining patterns.
The project site describes the software as designed for biologists and able to process images automatically, even millions of them. Scope matters here: CellProfiler measures images you have already acquired. It is an analysis tool, not instrument control software for a microscope.
How it is used in practice
The project site describes the workflow in three moves: load an example pipeline made of image-processing modules, adjust the settings so the modules measure the phenotype you care about, and export the results. Results go to a spreadsheet or a database, so the measurements can be analysed in whatever statistics tool the lab already uses.
Two practical points are worth knowing:
- Scale. Per the 3.0 paper (McQuin et al., PLOS Biology, 2018), Distributed-CellProfiler is a script-based interface for running parallel batches on AWS. The paper says it requires basic knowledge of AWS and interaction with the command line, so a cloud batch run is a different skill level from the point-and-click pipeline builder.
- Image formats. Images from many microscopes arrive in proprietary file formats; see opening CZI, ND2, LIF and VSI microscope files for how those formats are read.
Types and related concepts
CellProfiler is a small family of related pieces, summarised below from the project site and the two papers listed under Sources.
| Piece | What it does |
|---|---|
| CellProfiler (2006 paper) | High-throughput 2D cell image analysis; standard assays (count, size, per-cell protein levels) and morphological assays (shape, subcellular staining patterns). |
| CellProfiler 3.0 (2018 paper) | Whole-volume and plane-wise analysis of 3D image stacks; code now 100% Python; ClassifyPixels-Unet plugin runs a U-Net nuclei segmentation model on TensorFlow (can run on CPUs); MeasureImageFocus module predicts image focus using TensorFlow. |
| Distributed-CellProfiler | Script-based interface for parallel batches on AWS; needs some command-line use. |
| CellProfiler Analyst | Explores the measurements and classifies phenotypes with machine learning. Tools named on its page: Image Gallery, Classifier, Plate Viewer, Scatter Plot, Histogram, Density Plot, Table Viewer, Normalization Tool. |
Whole-volume or plane-wise for 3D data
The 3.0 paper describes two ways to analyse a 3D stack. Whole-volume analysis uses more memory but can be more accurate; plane-wise analysis works one plane at a time. The trade-off is memory against accuracy, and it is worth settling before a large run.
Where it matters for labs
CellProfiler matters where imaging produces more images than a person can measure by hand, most clearly in screens that image thousands of samples. This checklist helps decide how to set up a project:
- Assay type. Is the readout a standard count or size measurement, or a morphological one such as shape or staining pattern? Both are covered by the original paper.
- Dimensionality. Are the images 2D, or 3D stacks? For stacks, choose whole-volume or plane-wise analysis, knowing the memory cost.
- Scale. Is this a one-off set, or a large batch? Very large sets are what Distributed-CellProfiler on AWS is for, at the cost of some command-line work.
- Segmentation approach. Classical module settings, or a deep-learning model such as the U-Net nuclei plugin?
- Output. Spreadsheet or database, and whether you will explore and classify the data in CellProfiler Analyst afterwards.
For related bench imaging decisions on this site, see Fiji vs ImageJ for a general image-analysis alternative, measuring leaf area in ImageJ for a worked single-image measurement, and automated cell counters for instrument-based counting.
Common misunderstandings
It does not run the microscope
CellProfiler analyses images after acquisition. It does not control cameras, stages or lasers.
It is not only for programmers
The project positions it for biologists and builds analyses from modules with adjustable settings. Programming knowledge matters mainly for the cloud batch route.
CellProfiler and CellProfiler Analyst are different tools
CellProfiler produces measurements from images. CellProfiler Analyst works on those measurements: it describes itself as providing tools for classifying biological images, and the 3.0 paper calls it a user-friendly tool for classical machine learning based on measured features.
It is not limited to 2D
The 2006 paper was about 2D cell images, but the 3.0 release added 3D stack analysis.
FAQ
What is CellProfiler used for?
Measuring cell phenotypes in microscope images, from simple counts and sizes to shape and subcellular staining patterns, and exporting the results to a spreadsheet or database.
Who develops CellProfiler?
The project site is run by the Broad Institute, which holds its copyright notice. The original 2006 paper (Carpenter et al.) and the 3.0 paper (McQuin et al.) are the two main published descriptions of the software.
Can CellProfiler analyse 3D images?
Yes, from version 3.0, using either whole-volume or plane-wise analysis of 3D image stacks.
Does CellProfiler use deep learning?
The 3.0 paper describes a ClassifyPixels-Unet plugin that runs a U-Net nuclei segmentation model on TensorFlow, which can run on CPUs, and a MeasureImageFocus module that uses TensorFlow to predict image focus. Caffe models were supported earlier, through a CellProfiler-Caffe bridge built into version 2.2.0.
Can it process millions of images?
The project site says it can process images automatically, even millions. For large parallel batches, the 3.0 paper describes Distributed-CellProfiler on AWS.
Sources
- CellProfiler project site, Broad Institute: cellprofiler.org (checked 2026-10-29).
- CellProfiler Analyst page: cellprofiler.org/introduction (checked 2026-10-29).
- Carpenter AE, et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 2006;7(10):R100. PubMed 17076895 (checked 2026-10-29).
- McQuin C, et al. CellProfiler 3.0: Next-generation image processing for biology. PLOS Biology 16(7): e2005970, 2018. journals.plos.org (checked 2026-10-29).
