Deep learning and electrocorticography to tailor epilepsy surgery

S. Hoogteijling, E. Schaft, G. Huiskamp, S. Straumann, P. Smits, M. Demuru, P. van Rijen, F. Leijten, T. Gebbink, M. van Putten, M. Zijlmans

Research output: Contribution to journalMeeting AbstractAcademic

Abstract

Purpose:Intra- operative electrocorticography (ioECoG) is used to delineate epileptogenic tissue. This delineation can be difficult due to the paucity of accurate epileptic io-ECoG biomarkers. Computer- aided pattern recognition algorithms can be useful in the delineation process but need to be validated. This requires a large ioECoG data-set. Our aim is to construct a training and test set to train and validate a convolutional neural network (CNN) for binary classification of ioECoG channels as epileptic or non- epileptic.Methods:We retrospectively included patients who had an Engel 1A outcome from the RESPect database – a data-base with intracranial electroencephalography data from patients who underwent epilepsy surgery from 2008 on at University Medical Center Utrecht. Patients undergoing amygdala- hippocampectomy were excluded. The ioECoG channels were measured at 2048Hz and labelled as being inside or outside the resected area based on pre- and pos-tresection photos. All resected channels were assumed epileptic, given that all patients had Engel 1A outcome. We split the patients into an 80% training and 20% test set after stratification by age and pathology. The CNN will be trained and validated to dichotomously classify single ioECoG channels on the training and test set, respectively.Results:In total 113 patients (mean age: 18 [0- 68] years) were included where 57 had frontal lobe epilepsy, 43 tem-poral lobe epilepsy, and 13 another anatomical location; 112 patients showed MRI abnormalities; 49 patients had tumour tissue, 39 cortical development malformation, 20 other pathology types, and 6 no abnormalities confirmed by pathology. Per patient, 1- 6 pre- resection ioECoG re-cordings were made with 6- 36 electrodes where at least 1 electrode covered the resection area.Conclusion:This work is the first to build a large ioECoG dataset to train and validate a CNN. This dataset will be used to test whether a CNN can classify a single ioECoG channel as epileptic or non- epileptic
Original languageEnglish
Pages (from-to)104-104
JournalEpilepsia
Volume63
Issue numberS2
DOIs
Publication statusPublished - Sept 2022

Fingerprint

Dive into the research topics of 'Deep learning and electrocorticography to tailor epilepsy surgery'. Together they form a unique fingerprint.

Cite this