Spindle Density Analysis of Adult Epilepsy based on Automatic Detection Algorithms in EEG

Abstract

Electrophysiological investigations of sleep provide an important advantage for recording spontaneous neural activity and quantifying brain function. This study combined artificial intelligence technology to quantitatively analyze spindle density in adult patients with epilepsy. All patients received one-night sleep electroencephalogram monitoring. We employed a convolutional neural network-based sleep staging system to predict sleep macrostructure. Then we applied two species of advanced algorithms: Spindler and Latent state spindle detector, to automatically detect sleep spindle during non-rapid eye movement sleep stage 2. And we calculated three kinds of frequency range spindle density involving 11-17 Hz, 9-12 Hz, and 12-15 Hz. Our results suggested that 11-17 Hz and 12-15 Hz spindle density in adult epilepsy predominated in parietal and 9-12 Hz spindle density in prefrontal regions.