BACKGROUND: Functional magnetic resonance imaging (fMRI) is a less invasive way of mapping brain functions. The reliability of fMRI for localizing language-related function is yet to be determined.
OBJECTIVE: We performed a detailed analysis of language fMRI reliability by comparing the results of 3-T fMRI with maps determined by extraoperative electrocortical stimulation (ECS).
METHODS: This study was performed on 8 epileptic patients who underwent subdural electrode placement. The tasks performed during fMRI included verb generation, abstract/concrete categorization, and picture naming. We focused on the frontal lobe, which was effectively activated by these tasks. In extraoperative ECS, 4 tasks were combined to determine the eloquent areas: spontaneous speech, picture naming, reading, and comprehension. We calculated the sensitivity and specificity with different Z score thresholds for each task and appropriate matching criteria. For further analysis, we divided the frontal lobe into 5 areas and investigated intergyrus variations in sensitivity and specificity.
RESULTS: The abstract/concrete categorization task was the most sensitive and specific task in fMRI, whereas the picture naming task detected eloquent areas most efficiently in ECS. The combination of the abstract/concrete categorization task and a 3-mm matching criterion gave the best tradeoff (sensitivity, 83%; specificity, 61%) when the Z score was 2.24. As for intergyrus variation, the posterior inferior frontal gyrus showed the best tradeoff (sensitivity, 91%; specificity, 59%), whereas the anterior middle frontal gyrus had low specificity.
CONCLUSION: Despite different tasks for fMRI and extraoperative ECS, the relatively low specificity might be caused by a fundamental discrepancy between the 2 techniques. Reliability of language fMRI activation might differ, depending on the brain region.