Abstracts

Rationale: Cortical functional mapping is an important step in presurgical evaluation for epilepsy surgery programs. Passive mapping of task-related electrocorticographic (ECoG) activity can be easily implemented in parallel to epilepsy monitoring. Thus, functional mapping which utilizes ECoG signals may be contaminated by pathological high-frequency activities, and removing these activities may improve the mapping accuracy. From a clinical perspective, high-frequency oscillations (HFO) can be divided into pathological HFOs within epileptogenic zones of the brain, and physiological HFOs, which occur together with broadband cortical activity in functional brain areas. In this study, we analyzed ECoG signals during task-related functional mapping and investigated whether removing pathological HFOs improves the mapping quality.

Methods: ECoG recordings were obtained from six patients during cortical mapping tasks. Data were recorded in conjunction with epilepsy monitoring at the Asahikawa Medical University and Megumino Hospital in Japan. The separation between pathological and physiological HFOs was based on selected periods of resting condition and time-locked hand movements. In these periods HFOs were marked manually and served as training data. Thus, epochs in the hand movement tasks were considered as physiological HFOs while epochs in the seizure onset zone of resting-state were considered pathological HFOs. Linear discrimination analysis was trained to classify HFOs based on amplitude, mean frequency, line-length, spectral centroid, and bandpower within 1sec epochs, filtered between 60 and 250Hz. Finally, pathological HFOs were removed from unseen functional mapping data and the task-related activity, defined as a significant difference between 60-170Hz broadband ECoG power during baseline and active periods (ANOVA, p< 0.05), was localized before and after removal.

Results: Pathological and physiological HFOs could be classified with a cross-validated accuracy of 86.64% compared to manual labeling. Test data sets included 2710 processed ECoG locations, including 585 pathological HFOs channels. Before and after removal of classified pathological HFOs in the test data, 896 and 902 ECoG locations showed significant task-related ECoG activity, respectively. Out of the 585 HFO channels, 159 overlapped with functional areas.

Conclusions: Preliminary results indicate that more significant activity can be found during functional mapping with ECoG when pathological HFOs are removed. About 4% (6 out of 159) of the HFO locations in functional areas were not identified due to contamination with epileptogenic waveforms. Hence, epileptogenic signals affect the signal quality for physiological observations. These results have to be confirmed by other functional mapping techniques such as electrical cortical stimulation or functional magnetic resonance imaging to better understand the clinical impact of separating pathological and physiological HFOs.

Funding: Please list any funding that was received in support of this abstract.: This work was supported in part by the European project RapidMaps 2020 and Japanese government Grant-in-Aid No. 16H05434 and No. 17K19708 and JP 17H05900.