Abstracts

Rationale: Epilepsy has been traditionally considered as a cortical dysfunctional disorder; however, accumulating neuroimaging evidence has indicated the active involvement of subcortical nuclei (Badawy et al., Neurology 2013). Magnetic resonance fingerprinting (MRF) offers rapidly acquired quantitative maps to probe in-vivo brain tissue properties in a non-invasive way (Ma et al., Nature 2013). Here, we use MRF to characterize normal-appearing subcortical nuclei in patients with medically intractable focal epilepsy.

Methods: A 3D MRF protocol with 1mm isotropic resolution (Ma et al., JMRI 2019) was acquired from 48 patients with unilateral medically refractory focal epilepsy and 39 age-and-gender-matched healthy controls (HCs). T1 and T2 tissue property maps were reconstructed for each subject. Ten subcortical nuclei were segmented as regions of interest (ROIs) using Freesurfer, including bilateral thalamus, caudate, putamen, pallidum and nucleus accumbens. To avoid partial volume effects, MRF T1 and T2 values were sampled from the central 60% voxels of each ROI. The mean T1 and T2 value, as well as their coefficient of variation (CV) were compared between the patients and HCs at group level using two-sample t-tests. Subgroup and correlation analyses were performed to examine the relationship between significant MRF measures and clinical characteristics. Using significantly abnormal MRF measures in the subcortical ROIs, a support vector machine (SVM) model was built and tested with 5-fold and 10-fold cross-validation to separate the patients from HCs at individual level.

Results: MRF revealed increased T1 mean value in the ipsilateral thalamus and nucleus accumbens; increased T1 CV in the bilateral thalamus, bilateral pallidum, and ipsilateral caudate; and increased T2 CV in the ipsilateral thalamus in patients as compared to HCs (p < 0.05, false discovery rate corrected, Figure 1). These differences did not show significant correlations with epilepsy duration, seizure onset age, seizure frequency, and pathological findings. No significant differences were seen when sub-grouped by MRI lesion, epilepsy localization and seizure outcome. Trained by the combination of significantly abnormal subcortical MRF measures, the SVM classifier produced 78.2% average accuracy to separate individual patients from HCs, with area under the receiver operating characteristic curve (AUC) of 0.827. The 5-fold and 10-fold cross-validation showed stable AUC and identical classification performance (Figure 2).