Abstracts

Rationale: About 30% of temporal lobe epilepsy (TLE) patients continue to experience recurrent seizures after anterior temporal lobectomy (ATL)¹. Known indicators of poor seizure outcomes include cortical atrophy²; however, the dynamics of atrophy progression post-ATL and its correlation with surgical outcomes remain poorly understood. Here we employ an individualized network diffusion model (NDM)³ to simulate progressive morphometric changes in TLE patients post-ATL, to assess each unresected region’s likelihood to serve as an epicenter for such pathological progression.


Methods:
Forty-four TLE patients underwent T1 scans before and ~1 year after ATL. Patients were categorized into seizure free (Engel I, L/R=14/15) and seizure recurrence (Engel II-IV, L/R = 7/8) groups. Using FreeSurfer, we extracted regional volume for each scan for each patient, and calculated the percentage of postoperative change for each region. By applying individualized resection mask to a healthy white matter connectome template, we generated preserved structural connectivity (pSC) matrices for each patient. Using these pSC matrices, we simulated the progression of postoperative atrophies with NDM from each region, to test whether the atrophies diffuse through the pSC to form the observed atrophy patterns from certain epicenters in the brain. The correlation between the simulated and the actual atrophy patterns determines the likelihood of a region being considered as an epicenter, hence referred to as epicenter probability (Fig 1).




Results:
At the group level, we identified several brain regions with an epicenter probability greater than zero (P< 0.05, FWE-corrected). Notably, the highest probabilities were detected in the ipsilateral frontal lobe and extended into various contralateral regions (Fig 2A). Within these regions, recurrence group showed higher epicenter probability particularly in the bilateral frontal lobes and contralateral thalamus (P< 0.05, FWE-corrected, Fig 2B). On an individual basis, we generated null distributions for epicenter probabilities by randomizing each patient's pSC matrix 1,000 times. Regions surpassing the 95th percentile of these distributions were designated as individual epicenters. We then analyzed the distribution of epicenters relative to areas affected by ATL—those disconnected by resection versus those unaffected (Fig 2C). Our analysis revealed a significant interaction (F(1,42)=4.9, P=0.03), where patients with recurrent seizures demonstrated a higher prevalence of epicenters in the affected areas than in the unaffected ones (Fig 2D).




Conclusions:
Our findings indicate that post-ATL atrophies in TLE may follow a specific progression trajectory. Patients with recurrent seizures tend to have more epicenters, primarily emerging from areas disconnected by the resection. This study highlights the importance of examining postoperative changes as potential keys to understanding seizure recurrence.


Refs:

1. He, X et al. (2017) Neurology 88, 2285–93

2. Keller, S et al. (2015) Annals of Neurology 77, 760-74

3. Chopra, S et al. (2023) JAMA Psychiatry 80, 1246-57




Funding: NSFC82271491