Suite3D is a volumetric cell detection algorithm, generally applicable to any type of multi-plane functional 2p imaging where you see cells on multiple planes. For an overview of the algorithms, see our recent preprint.
suite3d.github.io has runnable demos, 3D visualizations, and step-by-step tutorials.
If you run into any kinks, please open an issue and we'll be happy to help you get up and running.
Suite3D needs Python 3.11 or 3.12.
pip install git+https://github.com/alihaydaroglu/suite3d.git
pip install 'cupy-cuda12x>=13.0,<14.0' # GPU (registration); the CPU fallback is much slower
pip install 'suite3d[viz]' # optional: napari 3D viewerRegistration runs on the GPU and needs a system CUDA 12.x install; keep cupy on the 13.x line. If you are not sure which CUDA you have, conda install -c conda-forge cupy will sort it out for you.
To work on the source, clone and install it editable instead:
git clone https://github.com/alihaydaroglu/suite3d.git
cd suite3d
pip install -e ".[all]"Four worked demos live in demos/, each on a real dataset:
| demo | recording | |
|---|---|---|
| 01 | 01-v1-tc030/ |
V1, standard 2P, 7 planes |
| 02 | 02-lbm-ss004/ |
LBM, 22 planes |
| 03 | 03-hippocampus/ |
CA1, standard 2P, 4 planes |
| 04 | 04-sweep/ |
parameter sweep, reuses demo 03 |
Each demo runs two ways: a run_pipeline.py script that goes start to finish, or a walkthrough.ipynb notebook that steps through the pipeline one stage at a time. See demos/README.md for the flags and data layout, or suite3d.github.io to watch them run.
cd demos/01-v1-tc030
python run_pipeline.py --data-root /path/to/data --out-dir ./resultsThe volumetric 2-photon datasets used to test Suite3D are on figshare:
Download and unpack it, then point each demo's --data-root at the folder.