Eighteen months in the making, this is the first version of CellProfiler that can identify objects in 3D images volumetrically – the result of a collaboration with the Allen Institute for Cell Science who funded the project together with NIH. If you’ve not yet seen it, the Allen Cell Explorer is a real visual and biological treat! So many researchers require completely automated analysis of 3D images, as more complex cell organoids are entering mainstream research. The new capabilities of CellProfiler aim to address this growing need.
Although there are no massive changes in the remainder of CellProfiler’s interface, a LOT has improved under the hood since the last release. The thousands of researchers using CellProfiler will probably notice the change in speed though: not just startup speed but also roughly 2-fold improvement in the time it takes to process a typical pipeline. It adds up particularly if you are paying for cloud computing resources! Speaking of which, we recently created Distributed-CellProfiler, which allows running jobs on Amazon Web Services, even if you’re not a computational expert.
We’ve made substantial progress simplifying CellProfiler’s installation. In addition to the macOS and Windows releases of CellProfiler we’ve started packaging a CellProfiler release for Linux that will ease installation across Linux distributions. We’ve also started packaging CellProfiler for a variety of formats, for example, a Python wheel is now available from the Python Package Index and a Docker image is now available from Docker Hub. In an effort to see new uses for CellProfiler we’ve made CellProfiler much simpler to compile on a variety of familiar and unusual platforms by requiring fewer dependencies and only using ubiquitous build systems.
Those taking a peek at the code will notice 3.0 is massively trimmed down, in part due to better integration with Python’s scientific community. We’ve contributed most of CellProfiler’s fundamental image analysis, image processing, and image segmentation algorithms to scikit-image making them readily available as a package to those writing their own imaging applications.
And for the machine learning enthusiasts out there, CellProfiler is the first biologist-friendly software we are aware of that can integrate deep learning! You might have noticed convolutional neural networks in the news, as they’ve been massively successful lately, especially for computer vision tasks. There are now preliminary demonstrations of using a TensorFlow and a Caffe model with CellProfiler.
Using the new 3D functionality within CellProfiler 3.0, a monolayer of HeLa cells is segmented, with the nuclei and cytoplasm of each cell uniquely labeled. CellProfiler measures 3D aspects of each object (e.g., volume, surface area), providing the data required to quantify 3D spatial relationships between cellular components. (Credit: Allen Institute for Cell Science and Kyle Karhohs)