Methods
Experimental Design
Task
The Experiment consisted of a FFA/PPA localizer task. Participants had to detect matching images in a 1-back-task (CogAtlas ID:
http://www.cognitiveatlas.org/task/n-back_task). Six different image categories were presented (faces, houses, landscape scenes, body parts, objects and scrambled figures) in four block-design-runs. For further information about the experiment we refer to the original study by
Sengupta 2016. For the purpose of this study only a subset of the original data (consisting of only three runs per subject) was used.
Subjects
In total 16 healthy participants conducted the experiment. Since only a subset of the original data (
Sengupta 2016) was used, five participants were excluded from further analysis. The remaining seven male and four female subjects were right handed and aged between 20 and 40. All had normal or corrected vision, the integrity of their visual function was assessed at the Visual Processing Laboratory, Ophthalmic Department, Otto-von-Guericke University, Magdeburg, Germany. The execution of the experiment was approved by the ethics comission of the University of Magdeburg and all participants consented voluntarily.
FMRI-analysis
fMRI Acquisition
For image acquisition a 3.0-Tesla-Scanner (Manufacturer: Philips Medical Systems, Model: Achieva dStream, Software Version 5.1.7.0) equipped with a 32 channel head coil was used. No special accomodations were made. Prior to the experiment a T1-weighted structural image was taken for every participant. For the functional data T2*-weighted echo-planar images with the following settings were used: gradient-echo, 2s repetition time, 30 ms echo time, 90 ° flip angle, 1943 Hz/px bandwith in phase encoding direction, parallel acquisition with sensitivity encoding (SENSE, parallel reduction factor 2). 35 axial slices (thickness 3.0 mm) with 80×80 voxels (3.0×3.0 mm) of in-plane resolution, 240 mm field-of-view (FoV), anterior-to-posterior phase encoding direction) with a 10% inter-slice gap were recorded in ascending order.
Data Records
The recorded data was converted into NIfTI-format and saved in compliance with the Brain Imaging Data Structure (BIDS version 1.0.0-rc3) specifiactions. All images were anonymized using a custom defacing-tool (
http://github.com/hanke/mridefacer.git). For further information about software, conversion and data structure, refer to Sengupta et al. (2016). The raw data used in this analysis is available online (
http://kumo.ovgu.de/~mih/demo_fmri.tar.gz). The scripts for preprocessing and statistical analysis can be found in the attachments.
Pre-processing
Data
pre-processing and statistical analyses were performed with tools of the FSL
package, version 5.0.9 (
Smith 2004;
Jenkinson 2012).
The fMRI data were corrected for head
motion by aligning all image volumes with the middle volume of each run with
FSL's MCFLIRT tool (Jenkinson 2001; Jenkinson 2002). After extraction of the brain from
surrounding tissue with the BET tool (Smith 2002), the functional data
were spatially smoothed with a Gaussian kernel at a full-width–at-half-maximum
of 6 mm. Slice timing correction (regular up) was performed using FEAT (version 6.00). A
temporal high-pass filter was applied to the time series with a cutoff period
of 80 s to remove low frequency confounds. For each participant, FLIRT (Jenkinson 2001) was used to align the functional images
to the individual structural T1 weighted image using a rigid body transformation
with 6 degrees-of-freedom (DOF). All non-brain tissue was removed from the structural images with the BET tool (Smith 2002) using robust center estimation beforehand. The high-resolution structural brain image of each participant
was co-registered to the MNI152 standard template (2 mm isometric voxel
resolution) by applying a 12 DOF affine transformation with FLIRT.
GLM analysis
The general
linear model (GLM) analysis was carried out with the FEAT tool version 6.00. The
six regressors representing the stimulus categories were modeled as boxcar
function defined by stimulus onset and duration and convolved with a
double-gamma function. The temporal derivative of these regressors was included
in the GLM analysis to account for small timing differences (
Friston 1998). Figure 1 displays an example design matrix for one run of one subject. FILM prewhitening
was applied to account for serial autocorrelation (
Woolrich 2001). Motion
correction parameters were included in the model to reduce nuisance effects.
The same temporal filtering parameters used during preprocessing were also
applied to the GLM to account for low frequency components.