Health Sciences Building (HSB), Room 10W77, located at 60 N. 36th Street, one block north of Market Street.
BIOMED PhD Thesis Defense
Title:
Altered Intrinsic Thalamic and Subcortical Functional Connectivity in Temporal Lobe Epilepsy with Focal-to-Bilateral Tonic-Clonic Seizures
Speaker:
Stacy N. Hudgins, PhD Candidate
School of Biomedical Engineering, Science and Health Systems
Drexel University
Advisor:
Hasan Ayaz, PhD
Professor
School of Biomedical Engineering, Science and Health Systems
Drexel University
Details:
Epilepsy is a debilitating neurological disorder affecting over 50 million people worldwide. In temporal lobe epilepsy (TLE), seizures typically arise from a mesial temporal epileptogenic zone and remain focal, yet a substantial subgroup develops focal-to-bilateral tonic-clonic seizures (FBTCS), which confer heightened risk for sudden unexpected death in epilepsy (SUDEP), cognitive decline, and poor surgical outcomes. The intrinsic thalamic and subcortical network mechanisms that support seizure generalization—particularly their dependence on seizure onset laterality and clinical utility as biomarkers—remain poorly defined.
This dissertation tested the hypothesis that TLE patients exhibit laterality-dependent disruptions in intrinsic thalamic and broader subcortical functional organization that distinguish FBTCS+ from FBTCS− patients and index seizure generalization propensity. Resting-state functional MRI data were acquired from 166 TLE patients (120 FBTCS+, 46 FBTCS−; 71 right TLE, 95 left TLE) and 119 healthy controls. Functional network connectivity (FNC) and graph-theoretic analyses quantified clustering coefficient (segregation), degree centrality (hubness), local efficiency (integration), and intrinsic connectivity contrast (ICC; cross-hemispheric balance) across 16 bilateral thalamic regions (Aim 1) and 27 bilateral regions per hemisphere spanning 7 subcortical structures (54 total ROIs) encompassing thalamus, hippocampus, amygdala, caudate, putamen, globus pallidus, and nucleus accumbens (Aim 2). Support vector machine (SVM) classification evaluated the discriminative value of connectivity features versus clinical features (age, sex, seizure onset age, duration, education) for FBTCS status prediction (Aim 3).
Aim 1 (intrathalamic network reorganization) revealed reduced cross-hemispheric ventral anterior (VA) thalamic connectivity in TLE versus controls, with more pronounced ipsilateral VA dysfunction in right compared to left TLE. FBTCS+ patients demonstrated reduced right dorsal posterior local efficiency. Critically, only right-TLE FBTCS+ patients exhibited impaired cross-hemispheric connectivity from ipsilateral VA that worsened with illness duration, implicating a laterality-specific mechanism of thalamic disruption in seizure generalization.
Aim 2 (subcortical network topology) identified six bilateral subcortical subnetworks in TLE, with right TLE exhibiting more extensive topological alterations than left TLE. FBTCS+ patients showed reduced clustering coefficient (CC; segregation) in the left ventroposterior lateral thalamus, left ventral anterior caudate, and right ventroposterior putamen (β ≤ −0.087, p ≤ .036), consistent with thalamo-striatal desegregation. Right TLE patients with FBTCS history presented with more pronounced subnetwork derangements that included increased intrinsic connectivity contrast (ICC; cross-hemispheric connectivity) in bilateral nucleus accumbens, a pattern heightened with recent FBTCS and absent in left TLE. Temporal stratification of FBTCS+ patients into current (≤1 year) versus remote (>1 year) subgroups revealed that nucleus accumbens cross-hemispheric abnormalities tracked with recent seizure burden (state-dependent marker), while thalamo-striatal segregation reductions persisted regardless of FBTCS recency (trait-like marker).
Aim 3 (machine learning classification) provided proof-of-concept evidence that SVM classification using subcortical connectivity features captured complimentary FBTCS-related variance partially independent of standard clinical features, with cross-validated AUC values ranging from 0.519 to 0.769 across models. Overall discrimination was modest and laterality dependent. In left TLE, connectivity features outperformed clinical features (AUC 0.648 vs. 0.519); in right TLE, clinical features significantly outperformed connectivity (AUC 0.769 vs. 0.685; DeLong Z = 2.48, p = .013), and combined models did not improve upon clinical features alone (combined vs. clinical, RTLE: Z = 4.14, p < .001). Ventroposterior thalamus and nucleus accumbens emerged as top connectivity discriminators. Because connectivity features were pre-selected from full-sample rmMANOVA contrasts and z-score normalization was applied to the pooled SVM cohort prior to fold partitioning, reported AUCs reflect apparent rather than fully held-out performance and may overestimate generalization. The classification analysis is therefore presented as exploratory.
Taken together, subcortical connectivity abnormalities in TLE involved organized limbic-thalamo-striatal subnetworks rather than isolated regional effects, with state-dependent Nucleus Accumbens effects dissociable from trait-like thalamo-striatal markers. Connectivity features provided complementary, proof-of-concept discriminative information with laterality-dependent performance: connectivity outperformed clinical features in left TLE, while clinical features outperformed connectivity in right TLE, and combined models did not uniformly improve classification beyond clinical variables. These intrinsic brain signatures motivate continued investigation as candidate FBTCS risk-stratification biomarkers and implicate targets for personalized medical treatment, pending nested-cross-validation replication and prospective external validation.