Common functional connectivity alterations in focal epilepsies identified by machine learning.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Focal epilepsies are a diverse group of neurological disorders.
• Understanding shared functional network characteristics is crucial for diagnosis and localization.
• Resting-state functional magnetic resonance imaging (fMRI) offers a non-invasive method to assess brain connectivity.

Purpose of the Study:

• To identify common functional connectivity (FC) patterns in focal epilepsies of varying etiologies.
• To differentiate epilepsy patients from healthy controls using FC measures.
• To lateralize the seizure focus hemisphere in epilepsy patients.

Main Methods:

• Utilized resting-state fMRI data from 103 adult and 65 pediatric focal epilepsy patients and 109 controls.
• Employed whole-brain FC measures including parcelwise connectivity matrices, mean FC, and degree of FC.
• Trained support vector machine models with cross-validation for patient-control classification and seizure onset hemisphere lateralization.

Main Results:

• FC measures linked to default mode and limbic networks were most important for distinguishing patients from controls.
• Somatosensory, visual, default mode, and basal ganglia regions were important for lateralization.
• The classification model achieved 75.6% accuracy (AUC = 0.83) for patient-control distinction.
• Lateralization accuracy reached 64.0% (AUC = 0.69) using a 400-parcel connectivity matrix.

Conclusions:

• Machine learning models identified common FC alterations in heterogeneous focal epilepsy populations.
• These findings suggest shared functional alterations extend beyond the seizure onset zone.
• FC measures demonstrate potential for distinguishing epilepsy patients and lateralizing seizure focus.