Transfer Learning for the Identification of Paediatric EEGs With Interictal Epileptiform Abnormalities

Background

Interictal epileptiform abnormalities are important indicators for epilepsy diagnosis, but identifying these patterns requires highly trained experts and is time-consuming.

Research

This study developed a transfer learning-based method to identify paediatric EEG recordings containing interictal abnormalities without relying on manual feature engineering or expert annotations. Spectrograms from multiple EEG channels were analysed using pretrained deep learning models.

Potential Impact

The approach demonstrates the potential of AI to assist in recognising abnormal paediatric EEGs and support more efficient epilepsy diagnosis.