Symptom-related damage patterns in mesial temporal lobe epilepsy with hippocampal sclerosis

This study was undertaken to elucidate the neuroanatomical and network substrates underlying diverse seizure symptoms in mesial temporal lobe epilepsy with hippocampal sclerosis (mTLE-HS) by applying multivariate support vector regression (SVR) to individualized structural symptom mapping and atrophy network mapping (ANM).

We retrospectively enrolled 216 patients with mTLE-HS and 460 healthy controls. Six non-mutually exclusive seizure symptoms were assessed. We performed left-right flipping of all images of right-sided mTLE-HS patients along the midline before processing the data. Individual atrophy gray matter maps were derived using normative modeling. Structural symptom mapping was performed using multivariate SVR lesion-symptom mapping. For each patient, indirect ANM maps were generated using a normative functional connectome and direct ANM maps using the patient's own resting-state functional magnetic resonance imaging. Correspondence between the two ANM approaches was assessed using individual-level whole-brain spatial correlations of unthresholded maps and group-level Dice coefficients of thresholded maps.

Individual structural atrophy patterns were spatially heterogeneous, whereas different seizure manifestations were associated with partially distinct structural patterns and more spatially convergent network-level associations. Focal to bilateral tonic-clonic seizures were associated with thalamic network involvement, gestural/limb automatisms with anterior cingulate-related networks, ictal vocalization symptoms with perisylvian language networks, dysmnesia with hippocampal-posteromedial networks, emotional phenomena with prefrontal-temporal networks, and autonomic phenomena with medial prefrontal and limbic networks. Furthermore, limited spatial correspondence was observed between the results of the direct and indirect atrophy network analyses, suggesting the indirect and direct ANM can be regarded as nonequivalent but potentially complementary approaches.

In mTLE-HS, heterogeneous seizure manifestations are associated with distinct patterns of atrophy and distributed functional network involvement. ANM provides a complementary framework for characterizing symptom-related network patterns, and differences between indirect/normative and direct/patient-specific connectome analyses may provide complementary information.