CellProfiler Changelog

What's new in CellProfiler 4.2.6

Aug 15, 2023
  • There are no changes to the functional CellProfiler codebase in this release compared to 4.2.5; the only changes are changes intended to make it easier to pip install CellProfiler by updating some pins that did not have wheels on some modern systems, and pinning at the top of their range packages whose updates break existing functionality. If you are using CellProfiler downloaded from our website, there is no need to update; these changes should only affect those installing via a local Python environment.
  • PRs:
  • Tighten some pins, loosen others by @bethac07 in #4801
  • Recreate #4638 in 4.2.x by @bethac07 in #4803

New in CellProfiler 4.2.5 (Dec 15, 2022)

  • Fix subsample values >1 overwriting input image.
  • Patch headless creation of properties files.
  • Update resize.py.
  • Catch "no objects of any type".

New in CellProfiler 4.2.4 (Aug 12, 2022)

  • "Ignore errors" mode
  • 3D CellProfiler Analyst support and better CellProfiler Analyst compatibility with fuzzy column matching
  • Better deep learning compatibility with the SaveCroppedObjects module
  • Lots of bug fixes, and some additional features

New in CellProfiler 2.1.1 Rev 6c2d896 (Jul 26, 2014)

  • Many bug fixes
  • Improvements to the code behind the input modules to address issues with color image stacks
  • Adding a caching layer to image file handling which should improve performance for image stacks
  • Improving the compositing of overlays on images
  • We have fixed a bug in the “fill holes” algorithm which will likely result in slightly different (and we feel, better) segmentations for IdentifyPrimaryObjects and IdentifySecondaryObjects modules with the “fill holes” option enabled. This will not affect most assays, but you may want to confirm this and continue to use 2.1.0 if reproducibility is a concern

New in CellProfiler 2.1.0 Rev 0c7fb94 (Jul 26, 2014)

  • New input modules and project storage files. CellProfiler has four new modules—Images, Metadata, NamesAndTypes and Groups—that are designed to assemble CellProfiler’s image sets in a more flexible and intuitive form. These four modules operate on a list of paths to image files, extracting metadata, filtering and grouping them into channels. The file list and pipeline are now stored in a project file. CellProfiler 2.1.0 assembles large image sets much more quickly than its predecessor and caches image sets in many circumstances to allow quick start-up times for pipelines. The legacy input modules, LoadImages and LoadSingleImages can still be used as before and the LoadData module can both be used as before and can be used with an image set list exported from a project.
  • Multi-core processing. CellProfiler now utilizes multiple cores in analysis mode, which allows for faster processing without any user intervention necessary. It starts a number of worker processes and partitions its work among them. In addition to the increased utilization of CPU resources, this multiprocessing mode allows the workers to run in a 64-bit address space on OS/X even though the UI process is constrained to a 32-bit address space. CellProfiler is still optimized to run on a single core when headless.
  • Measurements saved to disk as HDF5 files during processing. Measurements are now preferentially stored as HDF5 files which are written to disk during the course of the analysis. Previously, measurements were stored in memory and written to a Matlab format file at the end of the run. The change lowers CellProfiler’s memory footprint, especially for long runs, provides incremental measurement output and has paved the way for alternative uses of the HDF5 file.
  • Improved support for Linux. We provide Linux release and trunk builds for CentOS and Ubuntu.
  • Improved help. The help and welcome screen have been extensively revised to make them easier to use and more complete. The help is now searchable.
  • GIT and Github. In the interim between the CellProfiler 2.0 (11710) and the 2.1.0 release, we’ve moved our version control from SVN to GIT (https://github.com/CellProfiler/CellProfiler) and we now use the Github repository issue tracking system to file and resolve issues. We welcome Github pull requests.
  • Other added functionality. We have added new image operations, thresholding methods, and measurements. CellProfiler can load images from OMERO. Users can take advantage of the extended functionality of ImageJ 2.0 in addition to ImageJ 1.0 support.
  • Bug fixes. Many outstanding issues have been addressed, described in the complete list of fixes.
  • Measurements that have changed substantially:
  • The calculations for the following measurements have changed since CellProfiler 2.0 (11710) and yield substantially different results:
  • Neighbors_PercentTouching - The percent touching calculation in CellProfiler 2.0 mistakenly divided the number of touching pixels by the area of the object and multiplied by 100 to get the percent touching. CellProfiler 2.1.0 divides the number of touching pixels by the perimeter of the object (the maximum possible number of touching pixels) and multiplies by 100.
  • The following feature names have changed:
  • Haralick texture features - The Haralick texture features can now be measured using horizontal, vertical, diagonal and anti-diagonal displacements. The feature names have been given an additional “scale” suffix to indicate these four directions. The features measure along the same displacement as CellProfiler 2.0 by default.