Two Distinct Networks Discovered in Mesial Temporal Lobe Epilepsy: Connectomic Implications for Functional Neurosurgery

Introduction: Functional neurosurgical interventions aim to modulate or eliminate abnormal neuronal activity, guided by the brain’s functional architecture. Normal brain architecture is organized into canonical networks that are also reflected in neuropathology, vis-à-vis the Network Degeneration Hypothesis. However, no canonical network fully captures the effects of any one disease. Recent evidence suggests that focal epilepsy forms distributed pathologic networks (doi:10.1227/NEU.0000000000001880_331). Yet, even in well-characterized conditions like mesial temporal lobe epilepsy (MTLE), disease-specific networks remain ill-defined.

Methods: A comprehensive literature search identified 45 structural and 29 resting-state-functional imaging experiments, collectively reporting 588 coordinate-foci of pathology (n=1599 MTLE subjects). Patterns of coordinate co-occurrence (within and across experiments) were computed using independent component analysis (FSL-MELODIC.v3.14). Spatial cross-correlations were computed to assess the stability of and overlap between network-patterns. Functional interpretations were obtained by regional behavioral analysis (Mango.v4.1; rii.uthscsa.edu/mango/plugin_behavioralanalysis.html) of healthy task-activation data (20,998 experiments; n=103,977 subjects) from the BrainMap database (brainmap.org). Effect-modification was assessed (χ²-homogeneity).

Results: Two stable (R=0.890) MTLE-networks were identified, bearing little resemblance to canonical architecture (R < 0.3). Network-patterns overlapped at the hippocampus/MDN-thalamus but were otherwise spatially distinct (R=0.042). Network-1 (supramarginal, frontal/precentral, fusiform, temporal, and occipital gyri) was associated with verbal (speech-action:Z=8.23; speech-cognition:Z=6.85), motor (Z=6.06), and visual (Z=4.30) tasks. Network-2 (frontal, temporal and cingulate cortices, pulvinar, anterior-nucleus, and contralesional-parahippocampus) was associated with limbic (explicit-memory:Z=5.42; positive-emotion:Z=4.54) and executive-function (attention:Z=4.65; reasoning:Z=3.15) tasks. Structural and functional data contributed equally to both networks (χ²:p>0.05).

Conclusion : This study identified two distinct MTLE-networks, congruent with semiotic observations of communication deficits/stereotypic behaviors/visual hallucinosis (Network-1) and dyscognitive effects/transient amnesia/social-emotional deficits (Network-2). These cortical networks may underlie discrete behavioral phenomena manifested in MTLE and should be studied as potential routes of seizure propagation. Further investigation must determine their precise functions as they relate to the peri-ictal continuum; doing so will support pre-operative mapping/biomarker development and enable the pursuit of new network-targeted interventions.