ZUNA: Flexible EEG Superresolution with Position-Aware Diffusion Autoencoders

Announcements
,
1

Zyphra is excited to announce ZUNA, our first foundation model trained on brain data. ZUNA is a 380M-parameter diffusion autoencoder trained to denoise, reconstruct, and upsample scalp-EEG signals. Given a subset of EEG channels, ZUNA can:

  • Denoise existing EEG channels

  • Reconstruct missing EEG channels

  • Predict novel channel signals, given physical coordinates on the scalp

ZUNA outperforms spherical spline interpolation which is ubiquitously used by EEG researchers and practitioners and is included as the default in the widely-used MNE package. ZUNA is easy to use as a pip installable package. Visit Zyphra for more information and support!

3 Likes
4

Thanks a lot for an amazing large-scale, open-weights EEG foundation model.
I was wondering if you plan to compare it to other EEG-FMs?
In my opinion, the EEG-FM-Bench paper could be good start for this.

Love to see ZUNA crush the benchmarks! :slight_smile:

5

Thanks for the pointer. Will look into this.