Authors :
Presenting Author: Martina Kolajova, MSC – Mayo Clinic
Behrang Fazli Besheli, PhD – Mayo Clinic
Valentina Hrtonova, MSC – Mayo Clinic
Israt Tasnim, MSN – University of Houston
Enes arslan, MSC – University of Minnesota
Chandra Prakash Swamy, MSC – Mayo Clinic
Kai Miller, MD, PhD – Mayo Clinic
Vaclav Kremen, PhD – Mayo Clinic
Gregory Worrell, MD,PhD – Mayo Clinics
Nuri Ince, PhD – Mayo Clinic
Rationale:
High-frequency oscillations (HFOs) in intracranial EEG (iEEG) are investigated as biomarkers of epileptogenic tissue. However, distinguishing these pathological events from healthy physiological brain activity remains challenging, since physiological HFOs occur during normal brain functions and often overlap with pathological events in frequency, morphology, and amplitude. The objective of this study was to assess whether the previously published Clinical Neural Engineering Lab (CNEL) amplitude-threshold HFO detector (Liu et al., 2016) selectively captures HFOs without leakage of task-related physiological high-frequency oscillatory activities during cognitive task experiments.
Methods:
Intracranial EEG data were obtained from 5 patients with drug resistant epilepsy at Mayo Clinic as part of presurgical evaluation. Recordings were sampled at 2,048 Hz. Each dataset included an initial awake baseline interval followed by approximately 50 minutes of speech/language task recordings (auditory naming, picture naming, reading), allowing comparison between baseline and task conditions. Clinical annotation identified channels within the seizure onset zone (SOZ) and early propagation regions. Functional channels were defined from averaged baseline-normalized time-frequency maps, highlighting task-specific activations. HFOs were detected using a CNEL detector, applied to the ripple (80–300 Hz) and fast ripple (250–600 Hz) bands. With optimized parameters, the detector employs an adaptive amplitude threshold (multipliers of 3 and 4) and excludes events that do not meet burst duration or symmetry criteria. A postprocessing included HFO rate comparisons across conditions and visual inspection for the data quality and artifactual segment rejection.
Results:
Differences in functional areas between baseline and the task phases in HFO rates were not different across multiple thresholds of HFO detector settings (Fig. 1). The morphology of the HFO waveforms originating from eloquent areas were more arbitrary compared those that were detected in the SOZ (Fig. 2). Visual inspection of trigger-aligned data in task phases further confirmed that task-related HFOs were not captured.Conclusions:
Although language task-related physiological activity was visible in averaged baseline normalized time–frequency maps, HFO rates did not increase during tasks at either threshold. These findings suggest that the CNEL amplitude-threshold HFO detector is robust against task-related physiological oscillations and may serve as a reliable tool with minimal leakage of physiological activity into pathological detections. Future work will include larger cohorts and additional tasks to further validate the detector’s robustness. Funding:
This study was supported by the National Institutes of Health’s BRAIN Initiative under award number UH3NS117944 and grant R01NS112497 from the National Institute of Neurological Disorders and Stroke.