Neurofunktionelle MRT bei hohen Feldern

Clinical/methodical issue Functional magnetic resonance imaging (fMRI) examinations are limited in their sensitivity due to the low activation-induced signal change. Within short tolerable scan times the spatial resolution is thus limited. Standard radiological methods fMRI is a reliable tool in neuroscience as well as for clinical applications such as presurgical mapping of brain function. Methodical innovations The fMRI sensitivity improves greatly (more than linearly) with increasing magnetic field strengths. For many years this was the main driving force in the push towards higher field strengths, such as 7 T. Performance The sensitivity gain is greatest for high spatial resolution and fMRI with very high sub-millimeter resolution becomes feasible. Current results demonstrate that the localization of the blood oxygenation level dependent (BOLD) signal is better than previously assumed. Achievements High-field fMRI not only allows quantitative improvements but also opens the way to new information content, such as columnar and layer-dependent functional structures of the cortex. This may pave the way for further information, e.g. the directionality of cortico-cortical connections; however, these possibilities also pose new challenges. New methods for processing such high resolution data are required which do not require spatial smoothing and preserve the high information content. Practical recommendations Common spatial resolutions of 2–3 mm are still very well suited for examinations at 3 T where they benefit from the low signal void, lower geometrical distortion and reduced acoustic noise. To achieve higher resolution at 7 T parallel imaging and geometric distortion correction are essential and permit the best congruence with structural data. The echo time at 7 T should be adjusted to about 20–25 ms. Data processing for single subjects or patients should be performed with little or no smoothing to retain resolution. Group studies could achieve good correlation with local normalization. New methods for information extraction, such as multivariate pattern analysis may allow combination of group data without the need for voxel-based congruence