Authors :
Presenting Author: Giridhar Kalamangalam, MD, DPhil – University of Florida
Sai Bavisetty, PhD – University of California Los Angeles
Subeikshanan Venkatesan, MBBS – University of Florida
Steven Roper, MD – University of Florida
Ioan Chelaru, PhD – University of Florida
Rationale:
The network conceptualization of focal epilepsy (Spencer 2002: Epilepsia 43; 219-27) is firmly established. Yet, fundamental questions remain regarding seizure generation by epileptogenic networks: What network characteristics predispose to seizure tendency? How are these characteristics modulated by sleep-wake state and antiseizure medications (ASMs)? What are the relevant passbands of interaction? We considered these questions in a multi-day study of pre-surgical temporal lobe epilepsy (TLE) patients undergoing SEEG.
Methods:
Forty days ( >900h) of continuous SEEG data in five patients were analyzed. Individual electrode contacts were localized in MNI space with YAEL (Wang et al, 2023: eNeuro 10(10);1-9) and assigned an enclosing anatomical parcel in the Yale Brain Atlas (YBA; McGrath et al, 2022: Sci Rep 12;18778). Data from electrode contacts without significant pathological activity were used to compute a continuous measure of sleep-wake state (Chelaru et al 2021: Neuron 109(24):3954-61) across the entire multi-day period. YBA parcels were coarsened into blocks and SEEG data averaged within blocks to yield 12-20 data streams per subject, representing distinct anatomical sites. Pearson correlations were computed between pairs of Hilbert amplitudes of moving 5s data windows in the five Berger bands. Results were assembled into supra-adjacency matrices (SAMs) of within-band and inter-band interaction (Brookes et al, 2016: Neuroimage 132; 425-38; Figure 1). Average network conductance – measuring graph synchronizability - was derived for each of the 15 combinations of frequencies (five within-band, and 10 inter-band) from the SAM time-series. Data analysis used custom scripts implemented on office workstations or the University of Florida supercomputing cluster.
Results:
Regardless of frequency band, pairwise conductances exhibited periodicities on multiple time scales (Figure 2), similar to previously described (Brohl et al, 2024: Front Netw Physiol 3:1338864). The slowest such periodicity was influenced but remained independent of global sleep-wake state fluctuation. ASM withdrawal was accompanied by progressive erosion of the autonomy of slow rhythms and greater entrainment to the circadian cycle, or the appearance of new periodicities, in addition to dissociations between conductance pairs and selective response amplification of others.
Conclusions:
This study is the first to deploy the formalism of multilayer networks (Buldu & Porter, 2017: Netw Neurosc 2(4);418-41) to SEEG in TLE. We observe oscillatory dynamics on multiple time scales in the stable network. Instability in the ASM-deprived, seizure-prone network was signaled by disruption of autonomous oscillations to global arousal rhythms, appearance of new rhythms and a heightened sensitivity of conductance between particular frequency bands. Further work will make these preliminary results more precise and chart a conceptual path between the novel phenomena observed and the genesis of runaway instability (seizure).
Funding:
Wilder family endowments to the University of Florida (GPK, SV, IMC) and NINDS R21NS128503 (GPK).