Abstract .
MTLE is the most common type of epilepsy referred for surgery. Most MTLE seizures begin unilaterally and
respond well to resection of the epileptogenic tissue. Surgical success, however, is predicated on correct
identification of seizure-onset laterality. Non-invasive methods for determining onset laterality are seldom
definitive since MTLE seizures rapidly spread beyond the epileptogenic zone to extratemporal regions.
Functional network modeling is a promising method for establishing non-invasive biomarkers; in MTLE, this
method detects modulations in functional connectivity (edges) between pathologic regions (nodes) involved in
extratemporal seizure spread. However, many modeling techniques define nodes by arbitrary atlas-based
approaches that introduce bias in assessments of connectivity that is detrimental to model accuracy. An
incomplete understanding of MTLE connectivity (particularly with extratemporal regions of pathology) has
contributed to why non-invasive imaging biomarkers have yet to be approved for use in individual MTLE patients.
One robust approach for developing functional network models uses meta-analysis to define nodes and then
computes functional connectivity between those data-driven regions. This approach has been successfully used
to construct a clinically validated biomarker for network disruption in multiple sclerosis. In 2013, meta-analytic
modeling was first applied to MTLE. First, atrophy was identified in the medial temporal lobe and thalamic medial
dorsal nucleus (MDN). Then, a model of the functional connectivity between these atrophic regions was
constructed and used to predict seizure-onset laterality (86% sensitivity, 100% specificity). Until recently,
insufficient literature limited this meta-analytic model’s accuracy and clinical utility. However, the body of MTLE
literature has grown considerably (over double), presenting an exciting opportunity to expand this model and,
ideally, improve prediction accuracy of seizure-onset laterality to clinically useful levels. The proposed strategy
will address the hypothesis that: MTLE is a network-based disorder exhibiting lateralized changes in function at
rest; these changes will predict seizure-onset laterality with >97% sensitivity without loss of specificity. In this
proposal, MTLE connectivity will be studied in three specific aims by constructing 1) meta-analytic models of
health and disease connectivity, 2) group-wise models of connectivity in MTLE patients and healthy controls
using primary resting-state fMRI data, and 3) per-subject models of left- and right- lateralized MTLE patients
using rs-fMRI. Crucially, all models will be data-driven using nodes defined by meta-analysis. Using both sparse
and rich modeling approaches for each level of analysis (each aim), the proposed studies will identify paths of
MTLE seizure propagation (Aim 1), quantify the effects of MTLE (versus healthy controls) on network connectivity
(Aim 2), and validate a non-invasive diagnostic biomarker to predict seizure-onset laterality, per-subject, in MTLE
(Aim 3). Completion of these aims will further establish a pipeline for developing noninvasive clinical biomarkers
for other epilepsies while providing critical training for a future independent neurosurgical investigator.