Segregation and integration of subcortical brain network in temporal lobe epilepsy

Temporal lobe epilepsy (TLE) is characterized by significant abnormalities in perception, emotion, and behavior, but the underlying neural mechanisms remain incompletely understood. Here, we employed the nested-spectral partition (NSP) method to investigate the integration and segregation of resting-state subcortical brain functional networks in TLE patients, and explored the associations with clinical symptoms. We first revealed that the TLE group exhibited significantly reduced integration strength and increased segregation strength in the subcortical network, relative to the healthy control (HC) group, particularly in the basal ganglia, hippocampus and amygdala. Those abnormalities can significantly classify TLE patients from HCs with an AUC = 0.86. Then, we applied a meta-analysis to observe that those alterations of the subcortical network reflect the declines in high-order cognitive functions in TLE patients, especially in memory ability, which is further confirmed by the significantly decreased Wechsler Adult Intelligence Scale (WAIS) and Wechsler Memory Scale (WMS) scores. Finally, employing machine learning models, we successfully predicted WAIS/WMS scores wherein the amygdala, hippocampus, basal ganglia, cingulate gyrus, and thalamus were key contributors to WAIS scores, and the basal ganglia and cingulate gyrus were specifically associated with WMS scores. Compared to the graph theory measures, NSP-features were more powerful for classification and predictions. Our findings reveal the association between alterations in subcortical functional brain networks and cognitive declines in TLE patients, and offer novel insights into the neuropathological mechanisms underlying TLE.