Differentiation between epileptic and functional/dissociative seizures using density spectral array of ictal...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Accurate differentiation between epileptic seizures and functional/dissociative seizures (FDS) is crucial for epilepsy diagnosis.
• Traditional diagnostic methods can be time-consuming and may require specialized expertise.

Purpose of the Study:

• To investigate the efficacy of deep learning applied to density spectral array (DSA) of ictal electroencephalography (EEG) in differentiating mesial temporal lobe epilepsy (mTLE) from FDS.
• To assess the potential of a single EEG channel (Cz) for this differentiation.

Main Methods:

• Retrospective analysis of long-term video-EEG recordings from patients with mTLE and FDS.
• Generation of DSA from EEG data, focusing on the Cz electrode and specific time intervals.
• Training and testing of a ResNet34 convolutional neural network (CNN) model on the DSA datasets.

Main Results:

• The CNN model, utilizing DSA from the Cz electrode, achieved a high area under the curve (0.941) in differentiating between mTLE and FDS.
• Exploratory analyses confirmed the Cz electrode and the Middle 1/3 interval as optimal for differentiation.
• The deep learning approach demonstrated successful discrimination between epileptic and non-epileptic seizures using single-channel EEG.

Conclusions:

• Deep learning analysis of DSA from a single EEG channel (Cz) can reliably differentiate between epileptic and non-epileptic seizures.
• This technique offers a potential practical tool for early screening and triage of epilepsy, particularly beneficial in resource-limited environments.