Similar protocols

Protocol publication

[…] fMRI data were pre-processed with SPM8 software ( All functional images across pre- and post-drug sessions were spatially realigned to the first scan of the first run to correct for head movement. The T1 structural image was also co-registered to the mean aligned functional image. Realigned functional images were then normalized into a standard space using the transformation parameter from each individual’s T1-weighted scan to the standard T1 template. For multivariate fMRI analysis of olfactory and visual categorization scans, we did not perform subsequent spatial smoothing in order to preserve the voxel-wise fidelity of the signal. Images from visual localizer scans were smoothed for generating functional visual object recognition ROIs. [...] Following GLM estimation, we extracted 36 β pattern vectors (one vector for each of the 6 odors/images and each of the 6 runs) from all voxels within anatomically defined bilateral ROIs, manually drawn on the mean image of normalized T1 scans of all subjects, using MRIcron software ( A human brain atlas was used to help delineate the anatomical borders of anterior and posterior piriform cortex (APC and PPC), amygdala, and hippocampus (). The boundary of anterior and posterior hippocampus was delineated at the uncal apex (y = -21 in MNI space) (). The delineation of olfactory OFC was guided by an olfactory fMRI meta-analysis (). The entorhinal cortex was drawn with reference to an MR volumetric analysis of the human entorhinal cortex (Insausti et al., undefined). Visual ROIs of LOC and fusiform were defined by the independent functional localizer scan, as described above.For multivariate pattern analysis, because we focused on information encoded in distributed fMRI patterns, the pattern vectors from the left and right hemisphere of each ROI were individually scaled to have a mean of 0 and standard deviation of 1, and then concatenated together for bilateral ROI pattern analysis. This assures that the mean signal and any lateralization of activity does not account for information coding.The LIBSVM (Library for Support Vector Machines, implementation was used to decode category information from patterns within a given ROI at baseline (). We trained the SVM classifier to separate pairs of odors of different categories (e.g. C1 vs. M1) using all six runs, and then tested the SVM by classifying odor patterns of corresponding categories but different identities (C2 vs. M2). Because the training set and testing set contain odors of different identities, significant above-chance decoding is only possible when the patterns code category-specific information independent of the identities.Based on the regions identified by the SVM classifier in the baseline (pre-drug) session, pattern correlation analysis was then applied to these data, in an effort to characterize changes in pattern separation from pre- to post-drug sessions. Pattern dissimilarity (correlation distance) between presented stimuli was estimated by computing the linear correlation coefficient between vectors of β patterns across pairs of different runs and subtracting from 1 (thus, maximal similarity = minimal distance = 0). All possible pair-wise comparisons were calculated, and then averaged across same-odor distances (e.g., C1 in run 1 vs. C1 in run 2), within-category distances (e.g., C1 in run 1 vs. C2 in run 2), and across-category distances (e.g., C1 in run 1 vs. M1 in run 2). To control for potential drift and variations from one odor to the next in a session or between the sessions, we subtracted same-odor correlation distances from within-category and across-category correlation distances. The adjusted within- and across-category odor distances were used in group-level statistical analysis to test for drug effect. […]

Pipeline specifications

Software tools SPM, MRIcron, LIBSVM
Applications Miscellaneous, Magnetic resonance imaging, Functional magnetic resonance imaging
Chemicals Baclofen, gamma-Aminobutyric Acid