# Multimodal integration of magnetic resonance imaging and intracranial electroencephalographic abnormalities in temporal lobe epilepsy surgery

## TL;DR

This study shows that combining MRI and iEEG data improves identifying brain regions causing seizures in epilepsy patients, leading to better surgical outcomes.

## Contribution

The study introduces a quantitative multimodal approach integrating MRI and iEEG data to enhance epileptogenic tissue localization and outcome prediction in TLE surgery.

## Key findings

- MRI abnormalities in GM and SWM localized to the hippocampus and inferior temporal regions correlate with seizure-free outcomes.
- Adding iEEG abnormalities significantly improves outcome differentiation (AUC = 0.92, AUPRC = 0.89).
- MRI abnormalities are more likely to colocalize with iEEG implantation sites in patients with favorable outcomes.

## Abstract

Precise localization of epileptogenic tissue is critical for successful surgery in drug‐resistant temporal lobe epilepsy (TLE) but is challenging in those requiring intracranial electroencephalography (iEEG). A range of modalities are used for localization, including magnetic resonance imaging (MRI) and EEG, which are typically integrated qualitatively by the clinical team.

This study quantitatively performed retrospective analysis of three modalities in 40 individuals with TLE who underwent subsequent resective surgery: preoperative diffusion‐weighted MRI, T1‐weighted MRI, and iEEG. Brain abnormalities in gray matter (GM) volume, superficial white matter (SWM) mean diffusivity, and interictal iEEG band power were derived by comparison to 97 MRI controls and 247 subjects with iEEG. We hypothesized that combined abnormalities in GM and SWM could differentiate postsurgical outcomes and adding iEEG abnormalities would improve outcome differentiation.

MRI (union of GM and SWM) abnormalities were primarily concentrated in the ipsilateral hippocampus and inferior temporal regions. Resection of these abnormal regions effectively differentiated seizure‐free outcomes (area under the curve [AUC] = .76, area under the precision–recall curve [AUPRC] = .78, p < .01), corroborating previous results from larger TLE cohorts. Adding iEEG abnormalities improved outcome differentiation (AUC = .92, AUPRC = .89, p < .01; z = 2.01, p < .05). MRI abnormalities were more likely to colocalize with iEEG implantation sites (z = 6.26, p < .01) and iEEG abnormalities (z = 4.34, p < .01) in individuals with favorable outcomes (International League Against Epilepsy [ILAE] class 1 and 2), but not in those with less favorable outcomes (ILAE class 3+).

Combining quantitative MRI‐derived GM and SWM abnormalities with interictal iEEG data improves localization of epileptogenic tissue and postsurgical outcome differentiation. Multimodal approaches may offer added value for surgical planning in complex situations.