Retrospective Analysis of Interobserver Spatial Variability in the Localization of Broca's and Wernicke's Areas Using Three Different fMRI Language Paradigms.

PURPOSE

To determine interobserver spatial variability in language area localization using three commonly employed language tasks.

MATERIALS AND METHODS

With institutional review board approval, 125 fMRI time series from 50 different clinical language cases were retrospectively reviewed by three blinded readers who selected 3-dimensional points representing the perceived center of Wernicke's and Broca's areas using three language tasks (semantic decision, SD; sentence comprehension, SC; and silent word generation, WG). Point dispersion values were then calculated using the perimeter of the 3-dimensional triangle defined by the three readers' selections.

RESULTS

After resolving interobserver laterality disagreements, there was no difference in spatial variability between the three tasks (P = .069). The SD task had the fewest interobserver laterality disagreements (P = .028) and the SC task had fewer failed localizations for Broca's area (P = .050) and Wernicke's area (P = .013).

CONCLUSION

While there were no differences between interobserver spatial variability in language area localization between the three tasks, language task choice impacts the accuracy of fMRI language area identification because tasks vary in their rates of interobserver laterality disagreements and failed localizations. A combination of tasks including one with low laterality disagreements (eg, SD) and one with few failed localizations (eg, SC) may offer the best combination.