Brain metabolic characteristics distinguishing typical and atypical benign epilepsy with centro-temporal spikes

Atypical benign epilepsy with centro-temporal spikes (BECTS) have less favorable outcomes than typical BECTS, and thus should be accurately identified for adequate treatment. We aimed to investigate the glucose metabolic differences between typical and atypical BECTS using 18F-fluorodeoxyglucose positron emission tomography ([18F]FDG PET) imaging, and explore whether these differences can help distinguish. Forty-six patients with typical BECTS, 31 patients with atypical BECTS and 23 controls who underwent [18F]FDG PET examination were retrospectively involved. Absolute asymmetry index (|AI|) was applied to evaluate the severity of metabolic abnormality. Glucose metabolic differences were investigated among typical BECTS, atypical BECTS, and controls by using statistical parametric mapping (SPM). Logistic regression analyses were performed based on clinical, PET, and hybrid features. The |AI| was found significantly higher in atypical BECTS than in typical BECTS (p = 0.040). Atypical BECTS showed more hypo-metabolism regions than typical BECTS, mainly located in the fronto-temporo-parietal cortex. The PET model had significantly higher area under the curve (AUC) than the clinical model (0.91 vs. 0.70, p = 0.006). The hybrid model had the highest sensitivity (0.90), specificity (0.85), and accuracy (0.87) of all three models. Atypical BECTS showed more widespread and severe hypo-metabolism than typical BECTS, depending on which the two groups can be well distinguished. The combination of metabolic characteristics and clinical variables has the potential to be used clinically to distinguish between typical and atypical BECTS. • Distinguishing between typical and atypical BECTS is very important for the formulation of treatment regimens in clinical practice. • Atypical BECTS showed more widespread and severe hypo-metabolism than typical BECTS, mainly located in the fronto-temporo-parietal cortex. • The logistic regression model based on PET outperformed that based on clinical characteristics in classification of typical and atypical BECTS, and the hybrid model achieved the best classification performance.