Sieci neuronowe na wysokich częstotliwościach mogą wskazać tkankę powodującą epilepsję i przewidzieć wynik operacji u pacjentów opornych na leki
High-frequency directed networks can identify epileptogenic tissue and predict surgical outcome in drug-resistant epilepsy
W skrócie
Badacze przeanalizowali zapis aktywności mózgu u 20 pacjentów z epilepsją oporną na leki, badając połączenia między neuronami na bardzo wysokich częstotliwościach (powyżej 80 Hz). Odkryli, że połączenia mózgowe zmieniają się w charakterystyczny sposób w miejscach, gdzie powstają napady, i mogą pomóc lekarzom dokładnie określić, którą część mózgu usunąć podczas operacji. Badanie wykazało, że usunięcie określonych obszarów mózgu zidentyfikowanych tą metodą było udane w 83% przypadków.
Oryginalny abstract (angielski)
In drug-resistant epilepsy (DRE), functional connectivity (FC) analyses highlight the role of network hubs as potential surgical targets, yet most studies focus on conventional frequency bands. This study extends FC investigations to higher frequencies (>80 Hz) to explore their role in tracing epileptogenicity. We retrospectively analysed intracranial electroencephalography (iEEG) recordings from 20 DRE patients. FC was computed using orthogonalized amplitude envelope correlation (oAEC) and direct directed transfer function (dDTF) across eight frequency bands (delta to fast ripples) in data segments with and without high-frequency oscillations (HFOs). Network measures were subsequently calculated using graph theory. We compared these measures between segments with and without HFOs, and between electrodes covering resected and non-resected tissue separately in patients with good and poor outcomes, assessing their utility in identifying the epileptogenic zone (EZ) and predicting surgical outcome. Our results revealed a decrease in oAEC FC during HFO presence (p < .01) in the high gamma and ripple bands. Additionally, we observed increased oAEC nodal strength and clustering, and decreased dDTF outward strength and clustering inside the resection of good outcome patients, predominantly at higher frequencies (>beta). A logistic regression classifier localised the EZ with up to 80% accuracy (Area Under the Curve: 0.57-0.87), while resection of "isolates" predicted outcome with up to 83% accuracy. We conclude that, during the interictal period, high-frequency directed networks can trace epileptogenicity in DRE, with our findings contributing to the growing body of literature supporting the integration of FC-based biomarkers into presurgical evaluation workflows.