Epilepsy is a neurological disorder that is characterized by epileptic seizures. Though these seizures are the most prominent features of epilepsy, the disorder is often accompanied by cognitive decline, such as memory deficits, language problems, intellectual impairments and loss of network efficiency¹. In this study, the cortical thickness is used to determine network efficiency. This method is based on the assumption that statistical correlations of cortical thickness across a group indicate connectivity, since axonally connected regions are believed to have trophic, developmental, and maturational concordances². The aim of this study is to assess if there are observable differences in the cortical thickness network efficiency of patients with epilepsy compared to healthy controls. Furthermore, it is examined whether impaired cognition is associated with the network efficiency.

Forty-four participants are included in this study, of which 35 have localization-related epilepsy (16 females, age 22-61y) and 9 are healthy controls (6 females, age 18-54y). Besides the full-scale intelligence quotient (FSIQ), also measures of intellectual decline, ΔIQ³, are determined for the patients (FSIQ: 97 ± 16, ΔIQ: -8 ± 6) and controls (FSIQ: 111 ± 18, ΔIQ: -4 ± 9). A significant difference between the groups is found for FSIQ, while no other significant differences are found. T1-weighted 3D fast gradient echo was acquired (TR = 9.91 ms, TE = 4.6 ms, TI = 3s, FA = 8°, voxel size 1 x 1 x 1 mm) for all the participants on a 3.0-T unit (Philips Achieva) with an 8-channel SENSE head coil. The mean cortical thickness is derived for 68 cortical regions using Freesurfer⁴. The obtained cortical thicknesses are corrected for possible effects of age, gender and total brain volume by means of linear regression across the participants, and per region.

Connectivity matrices are constructed using the Pearson’s correlation across a group of participants for each pair of regions, resulting in a correlation matrix R per group. These correlation matrices are thresholded and binarized to obtain adjacency matrices with 150 edges, which represent the graphs. However, no individual measures are obtained directly with this method. Raj et al⁵ previously described an other method to obtain a measure of network efficiency on an individual basis: By calculating the deviation from controls for each region pair while imposing cortical thinning individual disease progression graphs were obtained. However, by imposing cortical thinning, other aberrant cortical thickness values are not considered. Therefore a novel method is proposed here: For each patient a measure of deviation from the control group is calculated by including this patient in the control group and constructing a graph. This graph is then compared to the reference graph obtained for the native control group. For each control a similar approach was used, excluding the respective control from the control group instead.

Graphs with a small-world topology are considered efficient networks, with a high degree of local information processing and low wiring costs⁶. This so-called small-world topology can be evaluated by two graph features, a relatively high clustering coefficient and short minimum path length. From that a measure of small-worldness can be defined by a combination of both the clustering coefficient and minimum path length (both normalized with respect to 1000 random graphs with similar nodes and degree). A graphical representation of the proposed method is given in Figure 1.

For each participant, the normalized graph features are calculated for the graphs that represent the deviation from healthy controls. For these measures, we observe a significant lower normalized clustering coefficient, measures of small-worldness and deviation from controls in the patients compared to the controls (Table 1). Furthermore, from the normalized graph features calculated based on the native patient and control group it is observed that the patients have a higher minimum path length, lower clustering coefficient and lower measure of small-worldness. No significant correlation (p = 0.71) is found between intellectual decline and network deficiency in the patient group. The results of the new method show that the measures of deviation from the controls are significantly lower in the patients. This shows that the network efficiency in patients with epilepsy is lower compared to controls, which is in accordance to several previous studies^1,2. The impaired cognition is not associated with the decrease in cortical thickness network efficiency. This study introduces a new method to obtain individual measures of network efficiency based on cortical thickness measurements. It is shown that patients exhibit a less efficient brain network compared to healthy controls in terms of intercortical thickness relations.