Authors :
Presenting Author: Hua Xie, PhD – Children's National Hospital
Xiaotong Li, BS – Children's National Hospital
Priyanka Venkata Sita Illapani, MSc – Children's National Hospital
Leigh Sepeta, PhD – Children's National Hospital
Madison Berl, PhD – Children's National Hospital
William Gaillard, MD – Children's National Hospital
Nathan Cohen, MD – Children's National Hospital
Rationale:
Focal cortical dysplasia (FCD) is the most common etiology of drug-resistant epilepsy (DRE) in children. In particular, focal to bilateral tonic-clonic seizures (FBTCS) mark a high risk of DRE in FCD1, are associated with lower quality of life, worse surgical outcomes, and risk of sudden unexplained death in epilepsy. Evidence highlights the thalamus and its interacting cortical circuits in FBTCS generation and propagation as part of the epileptogenic network.2 While resting-state functional MRI (rsfMRI) offers a valuable tool3 for examining the integrity of thalamocortical circuitry and its interaction with FCD lesions, few studies have evaluated the thalamus-lesion connectivity signature of FBTCS in FCD.
Methods:
This is a retrospective imaging cohort from Children’s National Hospital from 1/2011 to 1/2025. Patients with 3T MRI-confirmed FCD underwent rsfMRI scan as part of routine clinical care. Forty-three pediatric epilepsy patients with FCD (14 with FBTCS) were included within the age range of 5-22 years at time of scan, and had a minimum of 18-month of follow-up. Age- and sex-matched typically-developing (TD) controls were included from CNH (n=16) and from Human Connectome Project-Development public dataset (n=100). The 3T resting-state data underwent standard preprocessing using fMRIprep and harmonization using NeuroCombat. Using the TD data, we employed a “winner-takes-all” approach to parcellate the thalamus into seven nonoverlapping parcels based on preferential connectivity to Yeo networks (Figure 1 Left). We calculated the thalamo-lesional functional connectivity between the thalamic parcels and FCD lesion. We used TD connectivity as a reference to generate z-scored FCD-thalamus connectivity profiles to facilitate comparison across FCD locations. The z-scored thalamo-lesional connectivity profiles were entered to train an XGBoost classifier to predict presence of FBTCS.
Results:
We performed 50 repetitions of 5x5-fold nested cross-validation. The classifier achieved a cross-validated accuracy of 0.71±0.04, sensitivity of 0.84±0.05, and specificity of 0.48±0.06. We examined the contribution of thalamic functional parcels in classification (Figure 1 Right). The SMN-thalamus parcel, the functional parcel preferentially connected to the somatomotor network (SMN), which overlaps with the ventral posterolateral and intralaminar nuclei, significantly contributed more to FBTCS classification than any other thalamic parcels (ps < 0.001).