Identification of neural networks preferentially engaged by epileptogenic mass lesions through lesion network mapping analysis

Alireza M. Mansouri, Jürgen Germann, Alexandre Boutet, Gavin J.B. Elias, Karim Mithani, Clement T. Chow, Brij Karmur, George M. Ibrahim, Mary Pat McAndrews, Andres M. Lozano, Gelareh Zadeh, Taufik A. Valiante

Research output: Contribution to journalArticlepeer-review

15 Scopus citations


Lesion network mapping (LNM) has been applied to true lesions (e.g., cerebrovascular lesions in stroke) to identify functionally connected brain networks. No previous studies have utilized LNM for analysis of intra-axial mass lesions. Here, we implemented LNM for identification of potentially vulnerable epileptogenic networks in mass lesions causing medically-refractory epilepsy (MRE). Intra-axial brain lesions were manually segmented in patients with MRE seen at our institution (EL_INST). These lesions were then normalized to standard space and used as seeds in a high-resolution normative resting state functional magnetic resonance imaging template. The resulting connectivity maps were first thresholded (pBonferroni_cor < 0.05) and binarized; the thresholded binarized connectivity maps were subsequently summed to produce overall group connectivity maps, which were compared with established resting-state networks to identify potential networks prone to epileptogenicity. To validate our data, this approach was also applied to an external dataset of epileptogenic lesions identified from the literature (EL_LIT). As an additional exploratory analysis, we also segmented and computed the connectivity of institutional non-epileptogenic lesions (NEL_INST), calculating voxel-wise odds ratios (VORs) to identify voxels more likely to be functionally-connected with EL_INST versus NEL_INST. To ensure connectivity results were not driven by anatomical overlap, the extent of lesion overlap between EL_INST, and EL_LIT and NEL_INST was assessed using the Dice Similarity Coefficient (DSC, lower index ~ less overlap). Twenty-eight patients from our institution were included (EL_INST: 17 patients, 17 lesions, 10 low-grade glioma, 3 cavernoma, 4 focal cortical dysplasia; NEL_INST: 11 patients, 33 lesions, all brain metastases). An additional 23 cases (25 lesions) with similar characteristics to the EL_INST data were identified from the literature (EL_LIT). Despite minimal anatomical overlap of lesions, both EL_INST and EL_LIT showed greatest functional connectivity overlap with structures in the Default Mode Network, Frontoparietal Network, Ventral Attention Network, and the Limbic Network—with percentage volume overlap of 19.5%, 19.1%, 19.1%, and 12.5%, respectively—suggesting them as networks consistently engaged by epileptogenic mass lesions. Our exploratory analysis moreover showed that the mesial frontal lobes, parahippocampal gyrus, and lateral temporal neocortex were at least twice as likely to be functionally connected with the EL_INST compared to the NEL_INST group (i.e. Peak VOR > 2.0); canonical resting-state networks preferentially engaged by EL_INSTs were the Limbic and the Frontoparietal Networks (Mean VOR > 1.5). In this proof of concept study, we demonstrate the feasibility of LNM for intra-axial mass lesions by showing that ELs have discrete functional connections and may preferentially engage in discrete resting-state networks. Thus, the underlying normative neural circuitry may, in part, explain the propensity of particular lesions toward the development of MRE. If prospectively validated, this has ramifications for patient counseling along with both approach and timing of surgery for lesions in locations prone to development of MRE.

Original languageEnglish (US)
Article number10989
JournalScientific reports
Issue number1
StatePublished - Dec 1 2020

All Science Journal Classification (ASJC) codes

  • General


Dive into the research topics of 'Identification of neural networks preferentially engaged by epileptogenic mass lesions through lesion network mapping analysis'. Together they form a unique fingerprint.

Cite this