Epilepsy Lateralization by Elastic Net Regularization of a Diffusion Tensor Imaging Biomarker using Machine Learning

Elastic net logistic regression using DTI microstructure plus clinical and demographic variables discriminates left-sided from bilateral or generalized epilepsy with AUC = 0.88 in 334 adults.

• Why it matters: Accurate localization of the epileptogenic focus is critical for presurgical planning in drug-resistant epilepsy, yet existing imaging signals may be insufficient alone.
• What they did: In a retrospective cohort of 334 adults at the University of Kansas Health System, FD and MD from 36 bilateral white matter tracts were combined with clinical/demographic variables and used in supervised machine learning, with elastic net logistic regression evaluated by AUC, accuracy, sensitivity, and specificity.
• The result: The model achieved AUC values from 0.76 to 0.88 across tasks, including AUC = 0.83 for unilateral versus bilateral or generalized epilepsy and AUC = 0.76 for right- versus left-sided epilepsy.

AI summary of the abstract · may contain errors · not medical advice · read the paper.