napari-storm

A plugin for interactive visualization of Single Molecule Localization Microscopy (SMLM) datasets with Napari. This package uses the (currently experimental) Napari Particles layer developed by Martin Weigert (https://github.com/maweigert).

Installation

napari-storm needs Python 3.10–3.12. It's recommended to install it into its own environment, e.g. with conda:

conda create --name napari-storm python=3.11 pip

conda activate napari-storm

Then install from PyPI. The [pyqt6] extra brings in napari's Qt backend; without an extra, no Qt binding is installed and napari cannot open a window (use [pyside6] if you prefer PySide):

pip install "napari-storm[pyqt6]"

To work on napari-storm itself, install from a clone instead:

git clone https://github.com/napari-storm/napari-storm

cd napari-storm

pip install -e ".[dev,pyqt6]"

Usage

Starting napari-storm

With the environment active -- an Anaconda Prompt after conda activate napari-storm, for example -- run:

napari-storm

This opens napari with the napari-storm dock already docked on the right. python -m napari_storm does the same, and from a clone so does python napari_start.py.

Or start napari itself and open the dock from the Plugins menu:

napari

Importing data into napari-storm

Drag & drop onto the dock widget supported file types (Picasso, ThunderSTORM, MINFLUX, etc.) directly into napari or use the import file dialog.

MINFLUX data is read in both Abberior layouts: the original one, and the flat layout Imspector writes from 24.10 onwards. The newer one is accepted as .npy, .json, .mat, .zarr, and as pyMINFLUX's own .pmx. Which layout a file uses is determined from the file, so there is nothing to choose. A .zarr dataset is a folder rather than a file; since a file dialog cannot select one, pick any file inside it and the whole dataset opens.

If your file is not covered:

• one can either write a custom import function by following the instructions in src/napari_storm/localization_dataset_types/Custom_Import.py, then run it with the Custom button
• try the (experimental) Auto-detect Format button, which will try to extract the headers of your file and lets you assign your data. This should work for any .hdf5, .csv or .npy file.

Basic usage

When a dataset is imported you should be able to see five tabs in the widget: Data Controls, File Infos, Decorators, Data Filter and Data adjustment. In the data controls tab you can change the render range, load a new file, merge the currently open dataset with another file and change your view. There is also the option to change the colormap, add a scalebar or activate rainbow colour coding (for 3D datasets), and adjust the contrast with the slider beneath the colormap. The contrast acts on the reconstruction itself -- the Gaussians summed where they overlap -- so its lower handle hides sparse regions while keeping dense ones, and its upper handle sets how many overlapping localizations it takes to reach full brightness.

The File Infos tab simply displays information on the currently opened datasets.

In the decorators tab you can switch the rendering style from the scientific Gaussian to an alternative visualisation -- points, spheres, uncertainty ellipses and more -- connect each MINFLUX trace's localizations in the order they were measured, activate a grid plane and customize a lot of things for the grid as well as the render range box.

Last but not least is the data filter tab, which gives you the option to filter your displayed datasets by all properties available in the dataset. There you will find two sliders, where the top one lets you change the x-range of the displayed property and the other one controls the cut-off/cut-on of your filter. To apply the filter settings to the dataset simply press one of the apply buttons.

Tips

• Double click or drag the tabs anywhere to detach them from the window. This way you have an overview over all of them at the same time
• For STORM/PALM ... datasets, it is possible to change the rendering options in the data controls tab to Variable-size gaussian, to include the uncertainty values or photon counts for the rendering
• hold shift and drag the mouse for panning

Documentation

Full documentation is at https://napari-storm.readthedocs.io/ — a step-by-step tutorial, how the renderer works, and the embedding API for driving napari-storm from another application.

There is also a custom Q&A GPT for this repo specifically, available at https://chatgpt.com/g/g-68aebb6371a88191877094b48513d690-napari-storm-q-a

To build the docs locally, install the pinned toolchain (once):

pip install -r docs/requirements.txt

Then serve them with live reload:

mkdocs serve

Acknowledgements

napari-storm builds on work by others, with thanks:

• napari-particles by Martin Weigert (BSD-3-Clause), the experimental Particles layer this plugin renders through. The modules derived from it live in src/napari_storm/napari_particles/ and carry their own NOTICE.
• pyMINFLUX by the Single Cell Facility of the D-BSSE, ETH Zurich (source, Apache-2.0). Its MinFluxReaderV2 is the reference for the MINFLUX layout Imspector writes from 24.10 onwards: the version markers that tell the two layouts apart, the per-container quirks, and the .pmx structure were all read from it. napari-storm's reader is an independent implementation and contains no pyMINFLUX code, but it would not exist without theirs having documented the format first.

Issues

If you encounter any problems, please file an issue along with a detailed description.