Abstracts

RATIONALE: Language lateralization and precise localization of certain language areas is often required for presurgical planning. Although fMRI is able to deliver reliable language maps it is not yet routinely performed, partly because of its time consuming post-processing. This study was designed to compare results obtained from a real-time analysis tool (FIRE) and the 'gold standard' for off-line analysis (SPM99) in a clinically relevant task. METHODS: We examined twelve healthy right handed subjects on a conventional whole-body 1.5 T magnet. Over a period of 13 minutes we continuously measured 248 functional data sets with a T2*-weighted EPI sequence (TE=50 ms, TR=3125 ms). During this time a semantic task (ON) and a letter matching task (OFF) alternated every 25 seconds. In the ON phase volunteers had to decide whether two visually presented words have similar meaning or not. The task of the OFF period was to match two consonant strings for identity. Online analysis included detrending of the original data, realignment on a k-space basis, and a pixel-by-pixel comparison of the mean differences (t-test). All pixels of the statistical map with a z-score above 2.5 were displayed. For off-line analysis a standard protocol of SPM99 with a box-car design convolved with the hemodynamic response function was utilized. Statistical maps were overlaid on the surface rendering and on corregistered axial slices. RESULTS: Compared with the control condition, performance of the semantic task produced activations in multiple left hemispheric regions, in accordance with previous reports. The statistical overlay images of the real-time analysis were visually compared with the post-processed data. Real-time data were slightly noisier whereas SPM was able to suppress noise more effectively. A cluster comparison of activated areas also revealed a good match. Global and regional lateralization indices were affected by noise but also showed good agreement between both analysis. CONCLUSIONS: Functional imaging in real-time (FIRE) delivers reliable data with respect to lateralization of language dominance and localization of activated areas.