Normative Modeling of Structural and Diffusion MRI for Epilepsy Diagnosis and Classification

Background and Objectives: Normative modeling expresses an individual's brain measures as deviations from a large healthy reference and has been advanced as a route to precision imaging biomarkers in epilepsy, where structural abnormality is distributed rather than confined to the presumed focus. Whether this framework can inform clinical classification from clinically acquired T1-weighted and diffusion MRI is untested. We asked whether normative deviations discriminate between (1) epilepsy and functional seizures; (2) focal and generalized epilepsy; and (3) drug-resistant and drug-responsive epilepsy. Methods: We retrospectively studied patients with epilepsy or functional seizures without comorbid epilepsy who underwent 3-Tesla epilepsy-protocol MRI at the University of Pennsylvania. Morphometric deviation scores (cortical thickness, subcortical volume) were derived from CentileBrain (n>37,000 healthy references). White matter fractional anisotropy and mean diffusivity profiles were referenced to harmonized Human Connectome Project cohorts. Lobar composites, hemispheric asymmetry indices, and whole-tract means were compared with rank-sum tests; discrimination was summarized as area under the ROC curve (AUC) with bootstrap confidence intervals and Benjamini-Hochberg correction. Whole-brain deviation burden was examined as a secondary analysis. Minimum detectable effects were estimated by Monte Carlo simulation at 80% power (α=0.05). Results: We included 273 patients with epilepsy (203 focal, 37 generalized, 33 unclassified; 97 drug-resistant, 176 drug-responsive) and 32 with functional seizures. In the primary analyses, no regional or tract-level normative measure distinguished epilepsy from functional seizures, focal from generalized epilepsy, or drug-resistant from drug-responsive epilepsy; AUCs did not exceed 0.62, with upper confidence limits up to 0.72. In the secondary analysis, whole-brain morphometric deviation burden was modestly higher in epilepsy than in functional seizures (AUC 0.66, 95% CI 0.57-0.74) but did not separate the within-epilepsy comparisons. The study could detect AUC ≥0.65 for epilepsy versus functional seizures, ≥0.64 for focal versus generalized, and ≥0.60 for drug-resistant versus drug-responsive epilepsy. Discussion: Univariate normative deviation summaries from clinical MRI do not currently distinguish these groups well enough to guide care. Whole-brain deviation burden, the only summary measure that differed after correction, was detectable at the group level but insufficient for individual-level classification. Progress may require network-targeted analysis of patient-specific deviations and multimodal integration.

Authors

Institutions

Publication Details

Journal
medRxiv
Published
2026-11-17
DOI
https://doi.org/10.64898/2026.11.14.26368080
Primary Topic
Epilepsy research and treatment
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Normative Modeling of Structural and Diffusion MRI for Epilepsy Diagnosis and Classification

James J. Gugger, Alfredo Lucas, Ethan Eisenberg, Nishant Sinha et al.
3 links
medRxiv
Epilepsy research and treatment
preprint

Normative Modeling of Structural and Diffusion MRI for Epilepsy Diagnosis and Classification

James J. Gugger, Alfredo Lucas, Ethan Eisenberg, Nishant Sinha, Erin C. Conrad, Marc Jaskir, Daniel J. Zhou, Kathryn A. Davis, Joel M. Stein, Zack Goldblum, Chetan Vadali, Mariam Josyula, Sandhitsu Das, Julian Gal, Tamjid J Imtiaz, Raina Hutchinson
preprint en

Abstract

Background and Objectives: Normative modeling expresses an individual's brain measures as deviations from a large healthy reference and has been advanced as a route to precision imaging biomarkers in epilepsy, where structural abnormality is distributed rather than confined to the presumed focus. Whether this framework can inform clinical classification from clinically acquired T1-weighted and diffusion MRI is untested. We asked whether normative deviations discriminate between (1) epilepsy and functional seizures; (2) focal and generalized epilepsy; and (3) drug-resistant and drug-responsive epilepsy. Methods: We retrospectively studied patients with epilepsy or functional seizures without comorbid epilepsy who underwent 3-Tesla epilepsy-protocol MRI at the University of Pennsylvania. Morphometric deviation scores (cortical thickness, subcortical volume) were derived from CentileBrain (n>37,000 healthy references). White matter fractional anisotropy and mean diffusivity profiles were referenced to harmonized Human Connectome Project cohorts. Lobar composites, hemispheric asymmetry indices, and whole-tract means were compared with rank-sum tests; discrimination was summarized as area under the ROC curve (AUC) with bootstrap confidence intervals and Benjamini-Hochberg correction. Whole-brain deviation burden was examined as a secondary analysis. Minimum detectable effects were estimated by Monte Carlo simulation at 80% power (α=0.05). Results: We included 273 patients with epilepsy (203 focal, 37 generalized, 33 unclassified; 97 drug-resistant, 176 drug-responsive) and 32 with functional seizures. In the primary analyses, no regional or tract-level normative measure distinguished epilepsy from functional seizures, focal from generalized epilepsy, or drug-resistant from drug-responsive epilepsy; AUCs did not exceed 0.62, with upper confidence limits up to 0.72. In the secondary analysis, whole-brain morphometric deviation burden was modestly higher in epilepsy than in functional seizures (AUC 0.66, 95% CI 0.57-0.74) but did not separate the within-epilepsy comparisons. The study could detect AUC ≥0.65 for epilepsy versus functional seizures, ≥0.64 for focal versus generalized, and ≥0.60 for drug-resistant versus drug-responsive epilepsy. Discussion: Univariate normative deviation summaries from clinical MRI do not currently distinguish these groups well enough to guide care. Whole-brain deviation burden, the only summary measure that differed after correction, was detectable at the group level but insufficient for individual-level classification. Progress may require network-targeted analysis of patient-specific deviations and multimodal integration.

medRxiv
University of Washington (US), University of Rochester (US), University of Pennsylvania (US), Temple University (US)
Epilepsy Society, National Institutes of Health
Good health and well-being
Epilepsy research and treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon