Abstract
OBJECTIVE: To develop language functional MRI (fMRI) methods that accurately predict postsurgical naming decline in temporal lobe epilepsy (TLE). METHODS: Forty-six patients with TLE (25 left) and 19 controls underwent two overt fMRI paradigms (auditory naming and picture naming, both with active baseline conditions) and one covert task (verbal fluency). Clinical naming performance was assessed preoperatively and 4 months following anterior temporal lobe resection. Preoperative fMRI activations were correlated with postoperative naming decline. Individual laterality indices (LI) were calculated for temporal (auditory and picture naming) and frontal regions (verbal fluency) and were considered as predictors of naming decline in multiple regression models, along with other clinical variables (age at onset of seizures, preoperative naming scores, hippocampal volume, age). RESULTS: In left TLE patients, activation of the left posterior inferior temporal gyrus during auditory naming and activation of left fusiform gyrus during picture naming were related to greater postoperative naming decline. Activation LI were the best individual predictors of naming decline in a multivariate regression model. For picture naming, an LI of higher than 0.34 gave 100% sensitivity and 92% specificity (positive predictive value (PPV) 91.6%). For auditory naming, a temporal lobe LI higher than 0.18 identified all patients with a clinically significant naming decline with 100% sensitivity and 58% specificity (PPV: 58.3%). No effect was seen for verbal fluency. INTERPRETATION: Auditory and picture naming fMRI are clinically applicable to predict postoperative naming decline after left temporal lobe resection in individual patients, with picture naming being more specific.