CwA-T: A Channelwise AutoEncoder with Transformer for EEG Abnormality Detection

Youshen Zhao, Keiji Iramina · 2024

Read CwA-T: A Channelwise AutoEncoder with Transformer for EEG Abnormality Detection by Youshen Zhao, Keiji Iramina (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-14522.txt.

Description

Brainwave recordings can flag abnormal activity, but their size and complexity make automated reading hard. The authors propose CwA-T, a two-part system for spotting abnormal EEG. A channelwise convolutional autoencoder first compresses each channel on its own, which cuts computing cost and keeps features tied to single channels, where clinicians can read them. A single-head transformer then classifies the compressed signals, using attention to catch long-range patterns that separate normal from abnormal traces.

On the standard TUH abnormal EEG corpus, the system reaches 85 percent accuracy at the case level, with 91.2 percent specificity. It beats known baselines including EEGNet and Deep4Conv. Just as important for clinics, it is light: about 202 million operations and 2.9 million parameters, far below transformer-heavy rivals. The authors release the code openly and argue the design suits both research and bedside use.

Catalog record

CREATORYoushen Zhao, Keiji Iramina
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-14522

Browse the full library · About the Tower of Babel