DATA-ANALYSIS IN BRAIN ACTIVATION STUDIES WITH MRI

The sensitivity of gradient-echo magnetic resonance imaging (MRI) to c hanges in cerebral blood oxygenation has been introduced for mapping f unctional brain activation. To benefit from the high spatial and tempo ral resolution of the respective dynamic MRI data sets, their analysis requires algorithms that are capable of both precisely delineating ta sk-related activation patterns and demonstrating functional connectivi ty of interacting areas. Here, we present various strategies for data evaluation by means of correlational analyses that surpass the quality of subtraction-based activation maps by improving both sensitivity an d robustness. On a pixel-by-pixel basis the approach correlates signal time courses with a reference function, reflecting the temporal seque nce of activated and control states. Extended versions employ the calc ulation of auto- or cross-correlation functions that increase sensitiv ity, but require periodic stimulations. Following individual correctio n for non-specific but correlated signal fluctuations, mapping of task -related coherent activation can be improved using neighborhood princi ples. Such refined strategies are expected to enhance the usefulness o f oxygenation-sensitive MRI for studying the functional anatomy of the human brain under both physiological and pathological conditions.