Abstracts

Rationale: The different regions of the brain work as a part of a network and resting state networks have been established to play significant role in physiological functions. These networks are constituted with neuronal populations that participate in synchronous electrophysiological activity. Intracranial EEG recorded using steroEEG can be used to generate the network connectivity across different regions of the brain. This approach can be used to generate resting state network and also to identify epileptogenic changes in the network.


Methods: Coherence is a widely used measure to determine the frequency-resolved functional connectivity between pairs of data. We evaluated the resting state of the brain network with iEEG signals recorded in 12 patients undergoing epilepsy surgery evaluation with sEEG, with coherence measures. The iEEG was sampled at 2000Hz continuously for several hours and the network connectivity was estimated based on periods of recording when no clinical or electrphysiological seizures were noted. Coherence was calculated for low frequency (1-10 hz), medium frequency(10-30 hz) and high frequency components(30-120 hz). The coherence was calculated for each pair of the electrodes based on 1 second of data, generating a 2-dimensional matrix. To establish the significant components of the coherence-based connectivity an autoencoder model was trained with multiple periods of recording. A encoder-decoder based neural network model was used for this purpose.


Results: The trained autoencoder was capable of capturing the most significant components of resting coherence networks. The decoder model also enabled in developing anomaly detection based seizure identification. The amount anomaly noted in the decoder output was as a surrogate marker for identifying epilepsy onset. The coherence based network was unique in each patient with maximal connectivity with the rest of the brain regions noted for insula and hippocampal regions.

Conclusions: Autoencoder can enable refining the coherence based networks and this method can help in accurately identifying disruptions in the connectivity. The disruptions in the connectivity can be used for detecting abnormal neural functioning as seen in epilepsy and other neurological disorders.

Funding: None