Extracting rich information from images

Methods Mol Biol. 2009:486:193-211. doi: 10.1007/978-1-60327-545-3_14.

Abstract

Now that automated image-acquisition instruments (high-throughput microscopes) are commercially available and becoming more widespread, hundreds of thousands of cellular images are routinely generated in a matter of days. Each cellular image generated in a high-throughput screening experiment contains a tremendous amount of information; in fact, the name high-content screening (HCS) refers to the high information content inherently present in cell images (J Biomol Screen 2:249-259, 1997). Historically, most of this information is ignored and the visual information present in images for a particular sample is often reduced to a single numerical output per well, usually by calculating the mean per-cell measurement for a particular feature. Here, we provide a detailed protocol for the use of open-source cell image analysis software, CellProfiler, to measure hundreds of features of each individual cell, including the size and shape of each compartment or organelle, and the intensity and texture of each type of staining in each subcompartment. We use as an example publicly available images from a cytoplasm-to-nucleus translocation assay.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Bone Neoplasms / metabolism
  • Bone Neoplasms / pathology
  • Cell Compartmentation
  • Cells / ultrastructure*
  • Diagnostic Imaging
  • Fluorescence
  • Forkhead Box Protein O1
  • Forkhead Transcription Factors / metabolism
  • Green Fluorescent Proteins / metabolism
  • Humans
  • Image Interpretation, Computer-Assisted / methods*
  • Osteosarcoma / metabolism
  • Osteosarcoma / pathology
  • Pattern Recognition, Automated / methods*
  • Software
  • Tumor Cells, Cultured

Substances

  • FOXO1 protein, human
  • Forkhead Box Protein O1
  • Forkhead Transcription Factors
  • enhanced green fluorescent protein
  • Green Fluorescent Proteins