Abstracts

Rationale: Establishing imaging predictors of seizure recurrence in patients with a first seizure would potentially allow for earlier diagnosis and treatment, however few quantitative imaging studies have been conducted at this early stage. The objective of this study was to identify network markers of seizure recurrence in patients presenting with a first seizure using morphometric similarity analysis.

Methods: Sixty-four patients presenting with a first seizure underwent T1-weighted magnetic resonance imaging using a standard clinical protocol. Average last follow-up for the study was four years, 41 patients were seizure free (SF), and 23 patients had experienced seizure recurrence (SR) after an initial seizure. FreeSurfer was used for cortical parcellation and estimation of cortical thickness, surface area, grey matter volume, mean curvature and mean Gaussian curvature, which were used as morphometric features. To quantify similarity between cortical regions, morphometric similarity between each pair of regions was calculated using Pearson’s correlations between morphometric features to generate individual morphometric similarity matrices, which have been shown to be biologically meaningful (Seidlitz et al., Neuron, 2018;97;231-47). Graph metrics were calculated at a range of network thresholds and analysed using a generalized linear model, with age, sex and total intracranial volume as covariates. We investigated frequently applied network measures in epilepsy research, including betweenness centrality (BC), clustering coefficient (CC), local efficiency (LE) and nodal path length (PL) (Gleichgerrcht et al., Epilepsia, 2015,56;1660-8). Permutation tests and randomized null networks were generated to determine significance of network metrics.

Results: Patients with SR had higher BC in a network consisting of left and right parietal regions compared to patients with SF, who had higher BC between left temporal and right frontal regions. Higher CC and LE was observed in patients with SR in a network consisting of the left anterior occipital sulcus and preoccipital notch and transverse temporal sulcus, the left suborbital sulcus and right temporal pole also showed increased CC. Patients with SF had higher LE between the left angular gyrus and inferior temporal gyrus. We also found nodal PL was higher is patients with SR in a network consisting of left frontal and parietal regions, and higher in patients with SF in a network consisting of left and right temporal and occipital regions.

Conclusions: Our results show that morphometric similarity networks in patients with SR are altered compared to patients who only present with a single seizure. High CC and LE in patients with SR may suggest that while certain nodes are highly locally efficient, they are more isolated from the rest of the network, compared to patients with SF. Network segregation in patients with SR may represent a protective mechanism against the spread of seizures through cortical reorganisation during epileptogenesis (Pedersen et al., NeuroImage Clin, 2015;8;536-542).

Funding: Please list any funding that was received in support of this abstract.: This work was funded by a UK Medical Research Council (MR/S00355X/1) grant awarded to Simon S. Keller.