Automated detection of focal cortical dysplasia in epilepsy using machine learning on structural MRI
Abstract
Focal cortical dysplasia (FCD) represents a structural brain anomaly of developmental origin that significantly contributes to epilepsy. This proof-of-concept study evaluates the feasibility of machine learning (ML) algorithms to distinguish patients with FCD type II-associated epilepsy from healthy controls (HCs) using high-resolution structural MRI. Structural MRI data from a publicly available dataset comprising 167 subjects (both FCD-associated epilepsy patients and HCs) were analyzed. Neuroanatomical features were extracted with FreeSurfer software. Preprocessing steps included dataset reconstruction, feature scaling, dimensionality reduction, and encoding of demographic variables. Multiple supervised ML models—including Logistic Regression, Support Vector Machine, Gradient Boosting, and Random Forest—were trained and evaluated using nested cross-validation with 10 outer folds and 5 inner folds. Among the evaluated classifiers, Random Forest achieved the highest AUC (0.791). This model demonstrated balanced classification ability, as reflected by its ROC and Precision-Recall curves. While promising, these results reflect the model’s performance on the current dataset and require further validation. This study demonstrates the preliminary potential of ML techniques to distinguish FCD-associated epilepsy from HCs using structural MRI, providing supportive information for clinical review. However, further studies with larger, independent clinical cohorts are necessary to validate these results before considering any clinical implementation.
Related
- Multi-centre validation of MELD-graph for automated focal cortical dysplasia detection using 7T compared to paired 3T MRI
- Simulating the Radiological Appearance of Focal Cortical Dysplasia in MRI
- Balloon cells in type II focal cortical dysplasia: developmental origins, molecular signatures, and epileptogenic mechanisms.
- Epileptic hub in MRI-negative focal cortical dysplasia using advanced dynamic statistical parametric mapping and magnetoencephalography.
- Identification of microglia-associated hub genes and a CX3CR1/P2RY12 two-gene diagnostic model for cross-subtype epilepsy-associated cortical malformations: integrated transcriptomic analysis with external validation
BibTeX
@article{khoshakhlagh2026automated,
title = {{Automated detection of focal cortical dysplasia in epilepsy using machine learning on structural MRI}},
author = {Hamed KhoshAkhlagh and Fatemeh Rahimi and Mohammad Amin Khodadadegan and Moslem Solhirad and Arash Zare-Sadeghi and Abolhasan Rezaeyan},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-74891-9},
url = {https://www.nature.com/articles/s41598-026-74891-9},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Automated detection of focal cortical dysplasia in epilepsy using machine learning on structural MRI
AU - Hamed KhoshAkhlagh
AU - Fatemeh Rahimi
AU - Mohammad Amin Khodadadegan
AU - Moslem Solhirad
AU - Arash Zare-Sadeghi
AU - Abolhasan Rezaeyan
PY - 2026
DA - 2026/10/14
JO - Scientific Reports
DO - 10.1038/s41598-026-74891-9
UR - https://www.nature.com/articles/s41598-026-74891-9
AB - Focal cortical dysplasia (FCD) represents a structural brain anomaly of developmental origin that significantly contributes to epilepsy. This proof-of-concept study evaluates the feasibility of machine learning (ML) algorithms to distinguish patients with FCD type II-associated epilepsy from healthy controls (HCs) using high-resolution structural MRI. Structural MRI data from a publicly available dataset comprising 167 subjects (both FCD-associated epilepsy patients and HCs) were analyzed. Neuroanatomical features were extracted with FreeSurfer software. Preprocessing steps included dataset reconstruction, feature scaling, dimensionality reduction, and encoding of demographic variables. Multiple supervised ML models—including Logistic Regression, Support Vector Machine, Gradient Boosting, and Random Forest—were trained and evaluated using nested cross-validation with 10 outer folds and 5 inner folds. Among the evaluated classifiers, Random Forest achieved the highest AUC (0.791). This model demonstrated balanced classification ability, as reflected by its ROC and Precision-Recall curves. While promising, these results reflect the model’s performance on the current dataset and require further validation. This study demonstrates the preliminary potential of ML techniques to distinguish FCD-associated epilepsy from HCs using structural MRI, providing supportive information for clinical review. However, further studies with larger, independent clinical cohorts are necessary to validate these results before considering any clinical implementation.
SN - 2045-2322
ER -