BrainMaze - EEG Models#
Ready-to-use trained models for brain electrophysiology (EEG / iEEG), part of the BrainMaze family. Inference only: the models run on ONNX Runtime, on the CPU by default or on an NVIDIA GPU; PyTorch is not needed.
Version 0.1.0. Source: bnelair/brainmaze-eeg-models
model |
module |
output |
|---|---|---|
Seizure probability, CNN + BiLSTM (iEEG; successor of brainmaze-torch) |
probability every 0.5 s; NaN where not evaluated |
|
OpenSpindleNet (scalp EEG / iEEG) |
spindle intervals (s) + confidence, and the time that could not be evaluated |
The package is built for long, multichannel recordings with gaps (NaN). Missing data never turns into silent zeros or a silent “no event”: every result states where the model could not look at the data.