Temporal lobe epilepsy lateralisation and surgical outcome prediction using diffusion imaging.

Area of Science:

• Neurology
• Radiology
• Artificial Intelligence

Background:

• Medically refractory temporal lobe epilepsy (TLE) poses challenges for presurgical diagnosis.
• Diffusion-weighted imaging (DWI) offers insights into white matter integrity.

Purpose of the Study:

• To develop a supervised machine learning (ML) technique using DWI to improve presurgical workup for TLE.
• To classify seizure onset laterality and predict surgical outcomes in TLE patients.

Main Methods:

• A cohort of 151 subjects (62 TLE patients, 89 controls) was analyzed.
• Supervised ML models were trained using diffusion-weighted metrics.
• Feature dimensionality was reduced using community detection for interpretability.

Main Results:

• The ML model achieved high accuracy in classifying unilateral vs. bilateral seizure onset (AUC 1.000).
• Accurate classification was also observed for seizure-free vs. not seizure-free outcomes (AUCs 0.800 and 0.775 for left and right TLE, respectively).
• The model distinguished TLE patients from healthy controls (AUC 0.745) and left vs. right TLE (AUC 0.662).

Conclusions:

• This DWI-based ML technique effectively classifies key clinical decisions in TLE presurgical workup.
• The findings highlight the role of white matter pathology in TLE and augment existing network connectivity research.
• This approach can enhance the utility of diffusion imaging in managing TLE.