Differential alterations of structural network in temporal lobe epilepsy with different seizure types are associated with cognitive and psychiatric status.

BACKGROUND

The fundamental pathophysiologic understanding of different seizure types in Temporal lobe epilepsy (TLE) remains unclear. This study aimed to assess the distinct alterations of structural network in TLE patients with different seizure types and their relationships with cognitive and psychiatric symptoms.

METHODS

Seventy-three patients with unilateral TLE, including 25 with uncontrolled focal to bilateral tonic-clonic seizures (FBTCS), 25 with controlled FBTCS and 23 with focal impaired awareness seizures (FIAS), as well as 26 healthy controls (HC), underwent the diffusion tensor imaging (DTI) scan. Network-based statistic (NBS) and graph theory analyses were employed to investigate the structural network and its topological properties. Partial correlation analyses were conducted to examine the relationships between clinical variables and disrupted network characteristics. Additionally, the support vector machine (SVM) algorithm was utilized for the classification of controlled and uncontrolled FBTCS.

RESULTS

Compared to HC, TLE seizure type subgroups presented differently aberrant SC within the frontostriatal network. Additionally, alterations in the rich club organization and global network metrics were observed only in FBTCS. Notably, a significant decrease in all nodal metrics of the right amygdala were observed within the uncontrolled FBTCS group compared to the other three groups. Additionally, the disrupted nodal properties were significantly correlated with the age of onset, duration of epilepsy and psychiatric symptoms in FBTCS. Furthermore, the classifier achieved notably high accuracy (98%) in distinguishing between controlled and uncontrolled FBTCS.

CONCLUSIONS

Our findings may contribute to elucidating the neuropathological mechanisms of different seizure types in TLE and their impacts on cognitive and psychiatric status. SVM algorithm combined with nodal properties holds promise for predicting the poor seizure control of FBTCS.