Transformer based model for Seizure detection in EEG data

  • Implemented a novel 4-channel selection method, reducing EEG data requirement of a time series transformer to 17%
  • Utilized Data Uncertainty Learning and Data Leakage prevention methods coupled with a Hybrid Vision Transformer
  • Hypothesized and established age and gender correlation in the prediction of seizures on the CHB-MIT EEG database

Download Slides ; Download Report ; Github Repository