Abstracts

Rationale: High-frequency oscillations (HFO) emerged as a focus of study when first recorded using hybrid micro-macroelectrodes in 1999. A significant advancement came in 2006 when fast-ripples (FR), a subtype of HFO, were detected with clinical macrocontacts in epileptic patients, positioning HFO as potential biomarkers of the epileptogenic zone (EZ). However, the use of microwires remains limited, with only about 17 studies published. Macrocontacts record a much larger volume than microwires, raising the question: do FR recorded by micro- and macro-electrodes convey the same information? The scarcity of studies recording both micro- and macroelectrodes simultaneously highlights the need for true multiscale analyses. To address this, we employed intracranial electrodes equipped with tetrodes.


Methods: Tetrodes, which extend between macrocontacts, provide a multiscale view of fast oscillations in local field potentials (LFP) at both millimeter and micrometer scales, offering insights into the local origin of FR. For accurate comparisons, we focused on LFP recorded simultaneously on two macrocontacts adjacent to the tetrodes. We used a semi-automatic detection algorithm based on a convolutional neural network, Halizya, to identify events, which were visually reviewed and classified as FR, artifacts, neuronal spikes, or FR-like events. Since 2015, we have implanted 65 patients with hybrid electrodes equipped with tetrodes, recording during rest, sleep, or while watching the same TV show. Our initial analysis focused on 35 patients, using 10-minute epochs analyzed by Halizya. FR were characterized using three LFP components: the filtered temporal signal, the Hilbert envelope of the filtered temporal signal, and the spectral power density. We extracted measures of amplitude, duration, FR index, spectral mode, and entropy for each FR.


Results: FR were detected in 24 of the 35 patients. Both macrocontacts and microwires recorded different types of FR. Microwires detected FR not seen on macrocontacts (Figure 1) and more FR within the EZ and irritative zones (Figure 2B). MicroFR were longer (mean duration: 15.6 ms) than macroFR (14.7 ms, p< 0.001). MicroFR had a lower amplitude (z-score: 5.8) compared to macroFR (6.9, p< 0.001). MicroFR exhibited higher frequencies (Figure 2A), both in peak (spectral mode: 314.9 Hz) and proportion (FR-index: 0.32), than macroFR (259.8 Hz, p< 0.001; 0.18, p< 0.001). Additionally, microFR had higher entropy (0.6) compared to macroFR (0.7, p< 0.001), indicating a less complex and more predictable signal. These differences were consistent across patients and lesion types and did not vary between epileptogenic and irritative zones.


Conclusions: Are micro-FR and macro-FR the same? Not entirely. Microwires capture more local generators not detectable by macrocontacts, recording FR of higher frequency and lower entropy. More FR are detected within and outside the EZ with microwires. These findings question the clinical value of FR detected at both scales, which is under investigation in a multicenter prospective trial (NCT06105645).

Funding: This research was funded by the French National Research Agency (DYNEUMICS, ANR-21-CE17-0029) and the Ligue Française contre l'Epilepsie.