Home
Introduction to Gut Analysis ToolBox
Last updated
Introduction to Gut Analysis ToolBox
Last updated
Gut Analysis Toolbox is published in Journal of Cell Science as open access. Click here to read.
29/11/2024: There is an issue with the deepImageJ ganglia model prediction, using the most recent version of deepImageJ. We recommend not updating deepImageJ till this has been fixed if you already have GAT.
If you already have GAT and have trouble with the ganglia prediction, please refer to instructions here.
If you are new and installing GAT from scratch, follow installation instructions below and, then please refer to instructions here
An improved ganglia model with better accuracy is in the works! Watch this space!
Gut Analysis Toolbox enables semi-automated analysis of the neuronal and glial distribution within the gut wall (enteric nervous system). It uses a combination of FIJI macros, StarDist and deepImageJ under the hood to do the heavy lifting with automated segmentation.
For large 2D tilescans of gut wholemounts, it is recommended to use QuPath for ENS analysis. The StarDist model for neuron segmentation has been converted for use in QuPath.
What you can do with GAT:
Semi-automated analysis of number of enteric neurons: Uses pan-neuronal marker Hu or anything with similar labelling
Normalise counts to the number of ganglia.
Count number of neuronal subtypes, such as ChAT, nNOS etc..
Spatial analysis using number of neighboring cells.
Calcium imaging analysis: Alignment of images and extraction of normalised traces
Multiplex image alignment and analysis
The deep learning models supplied with GAT enable automated segmentation of:
enteric neuron soma (StarDist)
enteric neuronal subtypes (StarDist)
ganglia (deepImageJ)
Click on the sidebar to navigate through the WIKI.
Disclaimer: It is highly recommended to verify the output from the model and not to use it blindly. We have made every effort to ensure the models works across a range of 2D images, but it is likely you may have to tune parameters by testing on your images. In some cases, you may need to train a new model or contribute data for training.
You can watch GAT instructions as video tutorials on Youtube by clicking on the image below.
Over time, we will be introducing features such as analysis of other cell types/image modalities related to gut research. Feel free to contact us or post a suggestion as an Issue. The documentation is also being updated constantly as we are adding new features. Let us know if something doesn't look right by posting an Issue.
Disclaimer (again!): You should always verify the results of the segmentation as well as the outputs. GAT is by no means perfect and if not configured properly can give you incorrect segmentation results. Test it on couple of your images (smaller crops) from different repeats to make sure it works consistently.
It is recommended to use a fresh install of Fiji for working with GAT.
Download Fiji from this link. You do not need to run special installation software, just unpack it and its ready to go. Watch this video to see the Fiji installation process.
To configure GAT, you will need to install a few plugins first
You will get a pop up window that will check for updates:
Once its done, the ImageJ Updater box will appear. Click Manage update sites on the bottom left and a window pops up:
To install a plugin, tick on the corresponding boxes.
For installing GAT, tick on the following boxes:
BIG-EPFL
CSBDeep
clij
clij2
DeepImageJ
Gut Analysis Toolbox
IJBP-Plugins (MorphoLibJ)
StarDist
PTBIOP
For calcium imaging analysis, you have to enter one update site manually. Follow the instructions here and when it comes to entering the update site, use this: https://sites.imagej.net/Template_Matching/
Once all the boxes have been ticked, click "Apply and Close"
All the dependencies and scripts from the update sites selected above will appear in the ImageJ Updater.
Click “Apply changes” and once its finished, restart FIJI.
After all the installation and configuration, when you restart FIJI, you should see the GAT menu appear on FIJI.
If you have any trouble with installing model files required for segmentation, check the troubleshooting page.
Important NOTE: You need to use an older version of deepImageJ (v3.0.3) for the ganglia prediction to work accurately, please refer to instructions here on how to fix this.
GAT will be updated with features every now and then especially if any bugs have been reported. Usually FIJI may let you know if there are any updates. If not, you can manually update this by going to Help -> Update
If there are any updates, it will appear in the Updater window:
Click Apply changes. The updates will be downloaded. Once its done, make sure you restart Fiji. The newest version of GAT should be available.
GAT is aimed at analysing 2D images. It expects maximum projection or 2D images of wholemounts, however if it detects a 3D image or Z-stack, it will prompt the user to perform a maximum projection.
GAT has been tested on czi, lif, jpg, PNG and tif files. Essentially, any file supported by Fiji or Bioformats.
If the Z planes are out of focus and the final maximum projection is blurry, you can use the extended depth of field projection option.
The models have been trained using the easy-to-use notebooks from ZeroCostDL4Mic. The training data including the notebooks are available here: .
