Diffusion spectrum imaging-based machine learning for temporal lobe epilepsy lateralization.

Area of Science:

• Neuroimaging
• Machine Learning
• Epilepsy Research

Background:

• Accurate preoperative lateralization of temporal lobe epilepsy (TLE) is crucial for surgical planning but challenging with subtle or MRI-negative lesions.
• Conventional MRI limitations necessitate advanced neuroimaging techniques for precise localization of the epileptogenic zone.

Purpose of the Study:

• To develop and evaluate a diffusion spectrum imaging (DSI)-based machine learning approach for noninvasive TLE lateralization.
• To overcome the limitations of conventional MRI in identifying subtle or MRI-negative TLE lesions.

Main Methods:

• Retrospective analysis of DSI scans from 49 unilateral TLE patients and 25 healthy controls.
• Extraction of local connectome fingerprints and quantitative anisotropy (QA) features.
• Training a support vector machine (SVM) for patient classification and epileptogenic hemisphere identification using 10-fold cross-validation.

Main Results:

• The DSI-based SVM achieved high accuracy in distinguishing TLE from healthy controls (97.3%) using both fingerprint and QA features.
• The fingerprint model demonstrated 100% accuracy for lateralization among TLE patients, outperforming QA features (91.8%).
• Three-class classification (left TLE, right TLE, HC) yielded accuracies of 78.4% (fingerprint) and 73.0% (QA).

Conclusions:

• DSI-derived metrics combined with machine learning provide accurate and noninvasive TLE lateralization.
• This approach reliably detects epileptogenic zones, including those with subtle abnormalities, enhancing presurgical decision-making.
• The DSI-based method holds significant potential for improving patient outcomes in TLE management.