IBL Electrophysiology Feature Computation and Region Inference

This repository contains tools for computing electrophysiology features and performing region inference from neural recordings.

Documentation

📚 View Full Documentation

Start here
Tutorial	Install, download a published model, predict brain regions
Compute features on your own data	SpikeGLX, NWB/DANDI or SpikeInterface-Zarr recordings
Run a published model	Region decoder and spatial encoder, from the Hugging Face Hub
Plot features along the probe	Visualise how features vary with depth
API reference	Generated from the docstrings

The site is a Quarto website built from docs/; see docs/README.txt to build it locally.

Published models

Both are public Hugging Face repositories — no account or token needed — and load through the single load_pretrained entry point:

• int-brain-lab/ea-decoder-channel-xgboost — brain region per channel, from features
• int-brain-lab/ea-encoder-channel — features per channel, from position

from ephysatlas import load_pretrained

model = load_pretrained("int-brain-lab/ea-decoder-channel-xgboost", revision="2026_W42")
model.selftest()
out = model.predict(df_features)

Quick Installation

ibleatools requires Python 3.12 or newer.

git clone https://github.com/int-brain-lab/ibleatools.git
cd ibleatools
pip install -e .

Lite vs full install. The command above installs the lite version, which is the default. Lite was developed for ibl-alignment-gui (which depends on it as ibleatools[lite]) and covers S3 table/model downloads, region-classifier inference, the spatial encoder, and LF/CSD/AP feature computation — but not spike/waveform feature computation.

For the full install (spike/waveform features via dartsort, plus the SDSC pipeline), use the [full] extra:

pip install -e ".[full]"

Contributing

Please see our Contributing Guide for details on how to contribute to this project.