Abstracts

Rationale: The occurrence of epileptic waves and seizures is strongly related to sleep stages. Analyzing the relationship between sleep stages and scalp electroencephalogram (EEG) abnormalities is crucial for understanding the pathophysiology of epilepsy. There are few reports of direct classification of sleep stages from electrocorticography (ECoG). The aim of this study is to classify sleep stages using ECoG from the hippocampus and parahippocampal gyrus.


Methods: Study participants included patients over 17 years of age with intractable epilepsy who underwent subdural electrode implantation for clinical purposes at the The University of Tokyo Hospital between November 2013 and February 2023. Among 79 patients, we selected 18 patients who implanted subdural electrodes placed in the bilateral basal temporal lobes. The sleep periods were estimated based on videos recorded simultaneously with ECoG. Time-frequency analysis using wavelet transform was performed on the sleep ECoG. We performed a principal component analysis using 41 features corresponding to the frequency band power values of 0.5 to 200 Hz. In addition, we attempted to classify the sleep stage from ECoG by k-means clustering.


Results: 64 sleep sessions were obtained. The average sleep duration per session was 9 hours and 8 minutes. We extracted 16 sleep sessions from 11 patients with at least 6 hours of sleep, sufficient for analysis, and with few interruptions due to noise or seizures. The results of the wavelet transform showed that the power values clearly varied in both the low and high frequency bands depending on the time of day during sleep. In particular, the increase in the low frequency band affected the first principal component. The k-means clustering identified 3 or 4 clusters based on the power value of each frequency band.


Conclusions: In the present study, a clear increase or decrease in the power value of the low-frequency band was observed, and we classified the sleep sessions into 3 or 4 clusters using K-means clustering based on the power values of the frequency band. The ECoG of the hippocampus and parahippocampal gyrus has the potential to classify sleep stages. However, scalp EEG, electromyogram, and electrooculogram were not measured simultaneously with ECoG. Further investigation is needed to compare the relationship between the result of this study and conventional classification of sleep stages.


Funding: No funding was received.