Your very own automated pipeline for Beta-cell calcium imaging segmentation, registration, tracking, and analysis!
Written by Anne Alsup and Kelli Fowlds through The University of Texas at Arlington.
BetaBuddy is run through the Jupyter Notebook BetaBuddy.ipynb. Follow the steps bellow to run the pipeline and install all necessary packages.
Our pipeline has been tested on ND2 and TIFF Beta-cell image sequences. Other cell types have not been tested on this pipeline. Your image set will need to be in the same directory as BetaBuddy.ipynb.
If you need to convert ND2 images into TIFF, change the filenames after !./bftools/bfconvert-overwrite ./. The first filename will call on your current ND2 image. The second filename is how you want to save your TIFF file. Change the names for your DAPI and calcium dye images.
You will now have two TIFF files in your current working directory.
A directory for your images that will be segmented is created or emptied if one exists. If you have images in CellPoseImg directory, save or move them.
Due to the Linux OS you cannot have any spaces in the name.
A directory for the merged DAPI and calcium dye images is created. Four arguements need to be customized in the terminal command to correctly run this step. This will output a scaled version of your calcium dye images and save all frames separately with the DAPI channel if the user requires one.
Configuration and parameterization of Cellpose. Cells 4 and 6 need to be run to have the widget options appear in the output section. Following the instructions before these cells is very important to run Cellpose correctly. Cell 5 will show you which channels of your image contains the calcium dye and DAPI.
Test Cellpose on only one image to check on your parameters. You can always go back and adjust your parameters if the sample output looks incorrect.
Run Cellpose on all images and save the masks in the Mask directory in CellPoseImg. This step may take around 5-15 minutes depending on the size of your image stack.
The first command will merge your new mask images with your original Beta cell images with an ImageJ macro. The second command will run this merged image stack through the Fiji plugin TrackMate.
An XML titled MaskMerged.xml will be saved in the current directory along with the MaskMerged.tif. You can load this into Fiji through Plugins -> Tracking -> Load TrackMate File and see the calculated tracks. You can also open the Track tab and click on TrackIDs to see their location.
A CSV will be saved with fluorescent intensities from every cell grouped together by their TrackID over time. This CSV will be saved in the current directory as experimentjup.csv.
Note: The TrackIDs in the CSV files will be off by one
Background subtraction is performed by sampling 100 pixels outside of the ROIs created by Cellpose. These pixel intensities are saved as a CSV.
Last step!! The R script Beta_Buddy_Stats.R is run from the terminal to conduct background subtraction, normalization, preliminary statistical analyses, and data visualization. Three plots and the final CSV file has only tracks that span the full image stack and identifies tracks in the top 20% of signal variance over all frames.
This pipeline must be run on a Linux OS with GPU or CPU access and at least 16GB for image stacks.
The directory that holds your BetaBuddy.ipynb must also have the following:
- Calcium image stack and/or DAPI image stack as TIFF or ND2 file(s)
- DAPIMerge.ijm
- MaskMerge.ijm
- LabelLapLinuxTerminalJup.py
- bftools unzipped directory
- Beta_Buddy_Stats.R
- Fiji.app unzipped directory
To use the widgets within the Jupyter Notebook:
pip3 install ipywidgets
jupyter nbextension enable --py widgetsnbextensionInstall Cellpose:
python -m pip install cellposeCellpose dependencies:
pip3 install "opencv-python-headless<4.3"pip3 install scikit-imagepip3 install matplotlibCUDA and Pytorch will be necessary to run Cellpose with GPU. The following command is for CUDA 11.0, but you can find the correct CUDA installation command for you here.
conda install pytorch torchvision torchaudio cudatoolkit=11.0 -c pytorchThe Bio-Formats software tool will allow us to convert our ND2 files into TIFF files. The "bftools" directory will need to be inside the folder with your BetaBuddy.ipynb
wget https://downloads.openmicroscopy.org/bio-formats/6.9.1/artifacts/bftools.zip
unzip bftoolsYou can install Fiji through the terminal with the following commands. Once you unzip the folder, you will need to move the 'Fiji.app' folder out of 'fiji-linux64.zip' and directly into the directory with BetaBuddy.ipynb.
wget https://downloads.imagej.net/fiji/latest/fiji-linux64.zip
unzip fiji-linux64.zip
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent