Characterization of high-resolution Gradient Echo and Spin Echo EPI for fMRI in the human visual cortex at 7 T

• Published in: Magnetic resonance imaging 2017-07, Vol.40, p.98-108
• Main Authors: Rua, Catarina, Costagli, Mauro, Symms, Mark R, Biagi, Laura, Donatelli, Graziella, Cosottini, Mirco, del Guerra, Alberto, Tosetti, Michela
• Format: Article
• Language: English
• Subjects: Radiology
• ISSN: 0730-725X
• DOI: 10.1016/j.mri.2017.04.008

Abstract The increased signal-to-noise ratio (SNR) offered by functional Magnetic Resonance Imaging (fMRI) at 7T allows the acquisition of functional data at sub-millimetric spatial resolutions. However, simply reducing partial volume effects is not sufficient to precisely localize task-induced activation due to the indirect mechanisms that relate brain function and the changes in the measured signal. In this work T2* and T2 weighted Echo Planar Imaging (EPI) schemes based on Gradient Recalled Echo (GRE) and Spin Echo (SE) were evaluated in terms of temporal SNR, percent signal change, contrast to noise ratio (CNR), activation volume, and sensitivity and specificity to gray matter. Datasets were acquired during visual stimulation at in-plane resolutions ranging between 1.5 × 1.5 mm2 and 0.75 × 0.75 mm2 targeting the early visual cortex. While similar activation foci were obtained in all acquisitions, at in-plane resolutions of 1.0 × 1.0 mm2 and larger voxel sizes the T2 weighted contrast of SE-EPI allowed the identification of the activation site with better spatial accuracy. However, at sub-millimetric resolutions the decrease in temporal SNR significantly hampered the sensitivity and the extent of the activation site. On the other hand, high resolution T2* weighted data collected with GRE-EPI provided higher CNR and sensitivity, benefiting from the decreased physiological and partial volume effects. However, spurious activations originating from regions of blood drainage were still present in GRE data, and simple thresholding techniques were found to be inadequate for the removal of such contributions. The combination of 2-class and 3-class automated segmentations, performed directly in EPI space, allowed the selection of active voxels in gray matter. This approach could enable GRE-EPI to accurately map functional activity with satisfactory CNR and specificity to the true site of activation.