Overview

Let’s face it, when dealing with segmentation of microscopy data we often do not have time to check that everything is correct, because it is a tedious and very time consuming process. Cell-ACDC comes to the rescue! We combined the currently best available neural network models (such as Segment Anything Model (SAM), YeaZ, cellpose, StarDist, YeastMate, omnipose, delta, DeepSea, etc.) and we complemented them with a fast and intuitive GUI.

We developed and implemented several smart functionalities such as real-time continuous tracking, automatic propagation of error correction, and several tools to facilitate manual correction, from simple yet useful brush and eraser to more complex flood fill (magic wand) and Random Walker segmentation routines.

See the table below how it compares to other popular tools available (Table 1 of our publication).

Feature	Cell-ACDC	YeaZ	Cellpose	YeastMate	DeepCell	PhyloCell	CellProfiler	ImageJ/Fiji	YeastSpotter	YeastNet	MorphoLibJ
Deep-learning segmentation	✅	✅	✅	✅	✅	❌	✅	✅	✅	✅	❌
Traditional segmentation	✅	❌	❌	❌	❌	✅	✅	✅	❌	❌	✅
Tracking	✅	✅	❌	❌	✅	✅	✅	✅	❌	❌	❌
Manual corrections	✅	✅	✅	✅	✅	✅	✅	✅	❌	❌	✅
Automatic real-time tracking	✅	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
Automatic propagation of corrections	✅	❌	❌	❌	❌	✅	❌	❌	❌	❌	❌
Automatic mother-bud pairing	✅	❌	❌	✅	❌	✅	❌	❌	❌	❌	❌
Pedigree annotations	✅	❌	❌	✅	✅	✅	✅	✅	❌	❌	❌
Cell division annotations	✅	❌	❌	❌	❌	✅	✅	✅	❌	❌	❌
Downstream analysis	✅	❌	❌	❌	✅	✅	✅	✅	❌	❌	❌
3D z-stacks	✅	❌	✅	❌	✅	❌	✅	✅	❌	❌	✅
Multiple model organisms	✅	❌	✅	❌	✅	❌	✅	✅	❌	❌	✅
Bio-formats	✅	❌	❌	❌	❌	❌	✅	✅	❌	❌	✅
User manual	✅	✅	✅	✅	✅	❌	✅	✅	✅	✅	✅
Open source	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
Does not require a licence	✅	✅	✅	✅	✅	❌	✅	✅	✅	✅	✅

Is it only about segmentation?

Of course not! Cell-ACDC automatically computes several single-cell numerical features such as cell area and cell volume, plus the mean, max, median, sum and quantiles of any additional fluorescent channel’s signal. It even performs background correction, to compute the protein amount and concentration.

You can load and analyse single 2D images, 3D data (3D z-stacks or 2D images over time) and even 4D data (3D z-stacks over time).

Finally, we provide Jupyter notebooks to visualize and interactively explore the data produced.

Do not hesitate to contact me here on GitHub (by opening an issue) or directly at my email padovaf@tcd.ie for any problem and/or feedback on how to improve the user experience!

Bidirectional microscopy shift error correction

Is every second line in your files from your bidirectional microscopy shifted? Look here for further information on how to correct your data.