Epileptogenic Zone Localization from SPECT Imaging Using Radiomics.

Area of Science:

• Medical Imaging
• Radiology
• Neurology

Background:

• Epilepsy is a severe brain disorder impacting millions globally.
• Refractory focal epilepsy often requires surgical intervention to remove epileptogenic zones (EZs).
• Accurate localization of EZs is critical for successful epilepsy surgery outcomes.

Purpose of the Study:

• To develop and evaluate radiomic predictive models for localizing EZs using SPECT imaging.
• To compare the performance of different radiomic feature extraction methods and models.
• To assess the potential of radiomics to support clinical decision-making in epilepsy surgery.

Main Methods:

• Retrospective analysis of 20 ictal and interictal SPECT datasets.
• Image preprocessing using Statistical Parametric Mapping (SPM12) including normalization, registration, and z-score calculation.
• Extraction of radiomic features using voxel-based and map-based methods from ictal and z-score images.
• Construction and performance evaluation of six radiomic models, compared against nuclear medicine physician readings.

Main Results:

• The voxel-based combined radiomic model achieved the highest sensitivity (0.954) and AUC (0.918).
• The map-based combined model also showed high performance with an AUC of 0.895.
• The voxel-based ictal model yielded the highest specificity (0.848). Physician readings had lower sensitivity (0.679) but higher specificity (0.980).

Conclusions:

• Radiomic analysis of SPECT images shows significant promise for improving EZ localization in epilepsy.
• Radiomic models can effectively augment the diagnostic capabilities of physicians.
• These findings suggest a valuable role for radiomics in guiding epilepsy surgery planning.