Visualize

This module provides interactive visualisation of Catena datasets and predictions using:

• Napari - fast local 2D/3D desktop viewer.
• Neuroglancer - web-based 3D viewer for large volumes and annotations.

It is mainly intended for:

• inspecting raw EM, affinities, and segmentations,
• overlaying synapse annotations and partners,
• sanity-checking training data and model outputs.

Installation

• Install Anaconda/Miniconda
• Make a conda env for NAPARI by running (use the environment_napari.yml file): ###### this will install the particular napari version as mentioned in the yml

  conda env create -f environment_napari.yml -n napari-env 
  

• Make a conda env for NEUROGLANCER by running (use the environment_neuroglancer.yml file): ###### this will install the particular neuroglander version as mentioned in the yml. You may update neuroglancer, but code execution is not guaranteed.

  conda env create -f environment_neuroglancer.yml -n nglancer 
  

[!WARNING]
If you face issues with napari, install from here. Make sure to pip install the other dependencies in environment_napari.yml.

Verify Napari GUI loads

conda activate napari-env
napari

Verify Neuroglancer loads via import

conda activate nglancer
python -c "import neuroglancer"

Usage: NAPARI

[!IMPORTANT] Note that LSDs output segmentations that are smaller than the original EM shape. This is due to application of valid convolutions in the CNN model. So you must pad the output to match the raw shape. Follow example here.

Once Napari is installed successfully, you can run visualise_napari.py:

OPTION 1: - Open the script in an editor. - Edit the shebang to point to your napari-env in Line 1: #!/home/samia/anaconda3/envs/napari-env/bin/python. - Run small datasets (NB: ONLY Do this if your dataset is small, script loads all datasets into napari GUI & RAM):

  catena/visualize/visualise_napari.py -f /path/to/zarr 

- Run large datasets via slicing (Preselect select a ROI in the data to visualise): - 3D

  catena/visualize/visualise_napari.py -f /path/to/zarr -s z1:z2,y1:y2,x1:x2 #3D

- 2D

  catena/visualize/visualise_napari.py -f /path/to/zarr -sf 60 -st 150 #2D slices 60 to 150 across all datasets

OPTION 2: - Run small datasets (NB: ONLY Do this if your dataset is small, script loads all datasets into napari GUI & RAM):

  conda activate napari-env # call the script with the python which has napari
  python catena/visualize/visualise_napari.py -f /path/to/zarr 

- Run large datasets via slicing (Preselect select a ROI in the data to visualise): - 3D

  conda activate napari-env
  cd catena/visualize/visualise_napari.py -f /path/to/zarr -s z1:z2,y1:y2,x1:x2 #3D

- 2D

 conda activate napari-env
 cd catena/visualize/visualise_napari.py -f /path/to/zarr -sf 60 -st 150 #2D slices 60 to 150 across all datasets

Usage: Neuroglancer

[!WARNING] All data is expected to be 3D (or 4D for predicted affinities) per the example script.

• You must edit nglancer_pyconnectomics_example.py to suit your datasets that you want to visualise. This is currently an example script. Specially,

    raw_file: pass your own file with contains raw EM
    scales: change the scales which are set as 8nm in zyx
    names: change the zyx to xyz if your data is in that format
  

• The script expects an affinity dataset and a segmentation. Our raws generally contain raw, pred_affs and segmentation in the same zarr file. You must either comment these lines or provide the files with these datasets to plot them in neuroglancer.

Script features and functionalities

• nglancer_pyconnectomics_example.py

Functionality:

This is an example script that demonstrates how to use neuroglancer to visualize 3D (or 4D for predicted affinities) datasets. It expects an affinity dataset and a segmentation. The script sets up a neuroglancer viewer, defines coordinate spaces, and loads raw data, segmentation, and affinities from a zarr file.

Usage:

[!WARNING] You must edit nglancer_pyconnectomics_example.py to suit your datasets that you want to visualise. This is currently an example script.

• Specially, you need to edit the following:

    raw_file: pass your own file with contains raw EM
    scales: change the scales which are set as 8nm in zyx
    names: change the zyx to xyz if your data is in that format
  

• To run the script:

  conda activate nglancer
  python nglancer_pyconnectomics_example.py
  

  The script will print a URL to the console that you can open in your browser to see the visualization.

Troubleshooting Tips

Check Issue #27 if you are facing problems with neuroglancer installation.