Authors :
Presenting Author: Birgit Frauscher, MD – Duke University Medical Center
Veronique Latreille, PhD – McGill University and Montreal Neurological Institute-Hospital; Justin Corbin-Lapointe, MSc – École de Technologie Supérieure; Laure Peter-Derex, MD, PhD – Lyon University Hospital; Jean-Marc Lina, PhD – École de Technologie Supérieure
Rationale:
Accumulating evidence indicates that sleep rhythms mirror healthy brain networks. In epilepsy, specific sleep rhythms such as spindles and slow waves are altered by epileptic activity. In addition to the oscillatory brain activity, the non-oscillatory (or scale-free) components of the background electroencephalogram (EEG) may provide further information about the complexity of the underlying neuronal network. Prior work highlighted the role of such scale-free metrics in predicting the ictal phase. Studying these sleep features may be a promising avenue to identify the seizure-onset zone (SOZ) in focal drug-resistant epilepsy (DRE), which may ultimately benefit surgery planning. Here we propose an in-depth analysis of intracerebral sleep oscillatory and non-oscillatory interictal brain activity patterns using multiple EEG features to identify the SOZ in DRE.
Methods:
We analyzed intracranial EEG (iEEG) data of 38 adults with focal DRE implanted with depth electrodes (n=2463 bipolar channels) for clinical monitoring at the Montreal Neurological Institute (MNI). We estimated the following oscillatory and non-oscillatory metrics during early N2 and N3 sleep: (i) spindle rates, (ii) slow wave rates, (iii) rhythmic spectral power (delta 0.5-4 Hz, theta 4-8 Hz, alpha 8-13 Hz, sigma 10-16 Hz, beta 13-30 Hz, and gamma 30-80 Hz), and (iv) scale-free metrics (H exponent). Using a within-subject design, we compared each metric across the epileptic network: (i) Seizure-onset zone (SOZ; n=499 channels), (ii) Irritative zone (IZ; n=1314 channels) and (iii) Normal zone (NZ; n=650 channels). To quantify iEEG abnormalities within the epileptic network, patients’ values were adjusted based on normative maps derived from the open-access MNI iEEG sleep atlas (https://mni-open-ieegatlas.research.mcgill.ca/).
Results:
Spindle rates were significantly reduced in the SOZ relative to the IZ and NZ during N2 and N3 sleep (ps< 0.001; effect sizes = 0.69-0.74). Moreover, rhythmic power in the alpha, sigma, and gamma bands significantly differed across the epileptic network during both sleep stages, with alpha and sigma power being lower in the SOZ relative to the IZ and NZ (ps< 0.001; effect sizes = 0.72-0.81), and gamma power being higher in the SOZ relative to the NZ (p< 0.001; effect size = 0.76). Slow wave rates or rhythmic delta power did not differ across the epileptic network. The H exponent differed between the three zones during N2 sleep only, with higher H values (steeper scale-free slope) in the SOZ relative to the NZ (p=0.001; effect size = 0.58).
Conclusions:
We found that specific sleep oscillatory and non-oscillatory iEEG features are altered by the epileptic network. Specifically, reduced sleep spindles, rhythmic alpha and sigma power, higher rhythmic gamma power, and a steeper scale-free slope could distinguish the SOZ from non-SOZ areas. These metrics constitute promising interictal iEEG sleep biomarkers of the SOZ in patients with DRE, complementing the existing biomarkers that are heavily weighted towards the presence of abnormal pathological EEG activities.
Funding:
CIHR Banting Postdoctoral Fellowship, CIHR (PJT-175056), FRQS Chercheur-boursier clinicien Senior.