Acia-Workflows: Automated Single-cell Imaging Analysis for Scalable and Deep learning-based Live-cell Imaging Analysis Workflows

Accio 🪄 - and your single-cell insights appear - Not quite but - the acia-workflows platform - and your single-cell insights appear to become much easier within a fully automated workflow 🏗️🚀😉

The acia-workflows platform combines three components:

1. the acia framework provides modular image analyis pipelines with state-of-the-art segmentation (Cellpose, CellposeSAM, Contour Proposal Network, Omnipose) and tracking (trackastra, ultrack, PyUAT, laptrack) methods
2. the implementation of workflows in form of a Jupyter Notebook incorporating Python code, extensive documentation, software dependencies and visualization into a single document
3. a collection of workflow examples that are reproducible in the webbrowser.

The collection of application workflows contains the full image analysis pipeline including additional post analysis to investigate living cells. In acia-workflows, we present various worklow applications for single-cell analysis:

• individual analysis steps:
  • cell segmentation
  • cell tracking
• full application workflows:
  • growth rate quantification
  • co-culture characterization
  • alternating oxygen effects on the single-cell level

Moreover, we showcase a simple but efficient way to scale the development workflows to 10s or 100s of time-lapse sequences to generate quantitative and robust insights.

The workflows are implemented in juypter notebooks and contain the software dependencies, data, visualization and extensive text descriptions. Therefore, they are fully reproducible and you can execute them in your local jupyterlab or directly on colab without any installation.

Feel the magic of automation 🪄 and have fun with your live-cell imaging analysis 🦠🔬🚀

Segmentation 🦠

We provide support for multiple segmentation approaches:

Cellpose	CellposeSAM	Contour Proposal Network	Omnipose
Segmentation_Cellpose.mp4	Segmentation_CellposeSAM.mp4	Segmentation_CPN.mp4	Segmentation_Omnipose.mp4
Open notebook	Open notebook	Open notebook	Open notebook

Tracking

Trackastra	Ultrack	PyUAT	Laptrack
Tracking_trackastra.mp4	Tracking_ultrack.mp4	Tracking_pyuat.mp4	Tracking_laptrack.mp4
Open notebook	Open notebook	Open notebook	Open notebook

Complete workflow examples 📔

We present three application workflows that perform the entire analysis of microbial live-cell imaging experiments:

Growth Rate Quantification	Co-culture Analysis	Single-cell reaction to oxygen alternation
Open notebook	Open notebook	Open notebook

Scaling to quantitative insights 🚀

In order to obtain quantitative insights into living cells we implemented feature to scale jupyter notebook workflows across datasets. This yields insights across 10s to 100s of time-lapse sequences and bundles them in a comprehensive overview. We demonstrate the for our example workflows:

Growth Rate Quantification	Co-culture Analysis	Single-cell reaction to oxygen alternation
Notebook	Notebook	Notebook

More Application Workflows

CTC datasets (Hela)	...
tracked.mp4	...
Open notebook	...

Getting started

Instructions to install the software and get the examples running.

Immediately online using google colab

Please have a look at the colab links for segmentation, tracking, and complete case studies.

Local installation (docker)

To use the analsis workflows on your local computer, we provide a docker container with all dependencies

You can install it via the command line:

bash build.sh
docker run -v "./notebooks:/home/notebooks" -u $(id -u):$(id -g) -p 8888:8888 aw-gpu

Note: Local installation is designed for Linux. May differ for other operating systems.

Local installation (conda)

To reproduce and use these analsis workflows on your local computer, you need to install jupyterhub and some dependencies. At best use a linux-based operating system. Please create a new environment and install the dependencies:

git clone https://github.com/JuBiotech/acia-workflows.git
cd acia-workflows

conda create -n acia-workflows python=3.10 jupyterlab -c conda-forge
conda activate acia-workflows
jupyter lab

A browser windows should open that shows the jupyter lab view

Reproducibility

In order to reproduce the results presented in the paper and the workflows, we provide a Docker environment and a reproduction script. In order to reproduce please execute the following:

bash build.sh
bash reproduce.sh --all --gpu

The executed notebooks will be placed in the notebook_runs folder.

Note: If you do not have a GPU avaialable please use the flag --cpu instead. Be aware that this drastically increases the runtime of the reproduction. The full reproduction will take several hours! Note: Please use a linux system for full reproduction.

Contribute

Our work is designed to create sharable and reproducible workflows that also scale to high-throughput data. While we present three common analyses, this is of course does not cover the immense variety and complexity that we seen in live-cell imaging analyses. Therefore, we need you and your opinions 🤩 There are two major ways you can contribute:

1. You have data and an analysis in mind but you cannot extract the insights with your analyses scripts. Open a discussion on your analysis case here and get the opinion of our community.

2. You have developed your own analysis workflow and you are so proud of it that you would like to share it with others. Then also open a discussion here and let's make others benefit from your work by incorporation your script into our collection.

Together we are strong and advancing 💪