Model details

Transformer

  • Developers: Alireza Amirshahi
  • Institution: EPFL
  • Contact: alireza.amirshahi@epfl.ch
  • Source:
  • License: CC-BY 4.0 International license.
  • Reference paper: https://www.biorxiv.org/content/10.1101/2023.01.24.525308v2
  • citation:

Model details

Signal Processing pipeline:

Apply Short-time Fourier transform (STFT) on the raw inputs. Then, feed them to a vision transformer.

Model Description

A vision transformer model with four layers of encoder.

Model Features

The window length is 12 seconds.

Model Channels

17

Model Training

Siena Scalp EEG

Model Imbalance

Repeat the seizure windows by the ratio of the number of seizures over the non-seizure windows.

Results