Abstracts

Rationale: Human high gamma-frequency activity (HFA: 45–95Hz) has been used to map episodic memory (EM) in drug resistant epilepsy (DRE) patients underwent invasive EEG (IEEG). Early electrical stimulation attempts to enhance EM guided by HFA regions have met with encouraging modest results1. Understanding the nodal segregation and hubness of the HFA regions in relation to healthy controls (HC) will provide an understanding of brain organization that could prove crucial to identifying targets that will support stimulation success or failure in DRE. Methods: Thirty-seven DRE patients (prior to IEEG implantation) and 94 matched HCs underwent a 5 minute rsfMRI in 3T MRI scanner. The patients underwent presurgical neuropsychology assessment (California Verbal Learning Test, CVLT, Logical Memory, LM) and IEEG implantation with concomitant word list recall test (quantified using percent recall, p_rec). Regions with significant power changes in HFA band during ‘successful encoding’ were calculated. RsfMRI nodal graph indices of the HFA ROIs were then calculated (Fig1). Results: The IEEG HFA-ROIs mapped to 27 parcels (left (L) pars orbitalis, pars triangularis, frontal, cingulate, postcentral, parietal, fusiform, temporal and right (R) superior frontal, precentral, superior parietal, lateral occipital, fusiform, inferior temporal gyri) (Fig 1). The regions that showed significantly reduced segregation and hubness indices in DRE compared to HC are displayed in Figure 2.Among the regions that showed a significant group difference, several significant correlations emerged with the behavioral data in the DRE group. CVLT rate of forgetting positively correlated with increased CC (r=.36, p=.02) and LE (r=.4, p=.01) of L inferior temporal gyrus (L ITG). CVLT total learning (r=-.36, p=.02) and delayed recall (r=-.4, p=.01) was negatively correlated with LE of L superior frontal gyrus (L SFG). CC of R fusiform (R Fus) was positively correlated with p_rec (r=.35, p=.049). BCn and EC of L pars triangularis (L PT) were negatively correlated with LM (r=-.5, p=.004). Conclusions: This is one of the first reports of an association between HFA IEEG-defined memory regions, network organization, and behavioral measures of EM. HFA memory-relevant regions in DRE patients possess abnormal network organization compared to HC’s, with no strong regional specificity. Among the HFA nodes, a small set of left fronto-temporal regions showed heightened segregation that was associated with poor memory encoding and increased levels of forgetting. In contrast, heightened levels of hubness in an inferior frontal region was related to lower recall. In the setting of DRE, nodes with lower levels of segregation and hubness appear more adaptive and more likely to support enhanced memory. Our data suggest that nodal properties characterizing network segregation and hubness may provide added value to algorithms for selecting targets of electrical stimulation to enhance memory.1              Ezzyat, Y. et al. Direct brain stimulation modulates encoding states and memory performance in humans. Current Biology 27, 1251-1258 (2017). Funding: Nil