The classification of absence seizures using power-to-power cross-frequency coupling analysis with a deep learning...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• High-frequency oscillations are key biomarkers for epileptic tissue.
• Cross-frequency coupling (CFC) reveals complex brain rhythm organization.
• Power-to-power coupling (PPC) is a significant but underutilized CFC measure in seizure classification.

Purpose of the Study:

• To investigate the utility of PPC in classifying absence seizures using scalp EEG.
• To develop and evaluate an automated classification system employing deep learning.

Main Methods:

• Utilized EEG data from 94 absence seizures across 12 patients from Temple University Hospital.
• Calculated pairwise PPC between 2-120 Hz frequencies using EEGLAB.
• Trained a Stacked Sparse Autoencoder (SSAE) on CFC matrices for seizure classification.

Main Results:

• The SSAE model achieved a sensitivity of 93.1%, specificity of 99.5%, and overall accuracy of 96.8%.
• The model successfully distinguished between seizure and background EEG segments not included in training.

Conclusions:

• PPC is a relevant feature for accurate seizure classification.
• Combining PPC with SSAE deep learning offers an effective method for automated absence seizure detection in scalp EEG.