Participants
Twenty participants (10 females) took part in the study. The mean age of the participants was 23.1 years (age range 19–30). Nineteen of the 20 participants were right-handed. Four of the 20 participants had to be excluded from the analyses due to excessive movement and technical scanner issues. The study had research ethics board approval (National Research Council Research Ethics Board, No. 2006–03 and the Capital Health Research Ethics Board, No. 2006–173) and participants provided informed consent prior to participation.
Experimental design
The participants performed a modified Sperry task [9, 10]. Specifically, we selected the “crossed” conditions, which have been shown to elicit maximum corpus callosum activation [11]. Face and word stimuli were presented to both the right and left hemispheres in order to elicit interhemispheric transfer. Stimuli were either intact or scrambled (i.e., scrambled faces and pseudo-words). Participants were asked to evaluate whether they saw an intact face, scrambled face, intact word, and scrambled word using an MR compatible response pad (four-button, forced choice). Response hand was also crossed for all trials (left hand for words and right hand for faces). Participants were asked to fixate on a point (“+”) that was continually displayed at the center of the screen for the duration of the experiment. All stimuli were presented laterally (>2.3° from fixation) and rapidly (100 ms) in order to initially stimulate only one hemisphere and avoid saccades.
E-Prime (Psychology Software Tools, Inc.) was used to present stimuli, which were displayed using back-projection to a screen mounted inside the magnet bore, and viewed through a mirror mounted on the head coil. Prior to the experiment, each participant performed a short practice task (with feedback) outside of the MRI scanner to ensure complete understanding of the task.
Figure 1 provides an overview of the design structure. We used a mixed fast/slow event-related design to examine both overlapping and isolated events. To minimize subject fatigue, the experiment was divided into two equal sessions (3 minutes and 48 s per session). In total across the two sessions, 80 stimuli were presented with 8 isolated events. Pilot study testing revealed that this design was optimal and further repetitions resulted in response reduction due to habituation (i.e., a confounding factor).
Stimuli were presented in blocks of 1–4 rapidly presented images (100 ms duration, 2 s inter-stimulus interval). The time between stimulus blocks was jittered pseudo-randomly (2000 ms, 4000 ms, or 16000 ms). To isolate events, we included four single events flanked by a 16 s rest period, in both of the trials. All 8 “isolates” were used to generate raw HRFs and all 80 events were used to estimate the finite impulse response (FIR; [17, 18]).
MRI acquisition
Data were acquired from a 4 T Varian INOVA whole body MRI system. Gradients were provided by a body coil (Tesla Engineering Ltd.) operating at a maximum of 35.5 mT/m at 120 T/m/s, and driven by 950 V amplifiers (PCI). A TEM head coil (Bioengineering Inc.) was employed.
Functional MRI data were obtained using an asymmetric spin echo (ASE) spiral sequence [13]. The ASE spiral sequence collects three images per slice per volume (maintaining equal BOLD contrast across the images, but increasing the T2 weighting). Seventeen axial slices (4 mm thick, no gap) were prescribed to cover the corpus callosum, as well as the regions extending superiorly and inferiorly. Other parameters for functional imaging were: 64x64 matrix, 220x220mm field of view, 1 shot, TR = 2 s, TR/TE/TE* = 2000/68/27 ms (TE = spin-echo center, TE* = asymmetric echo times). Following the functional MRI, a high resolution ASE spiral (128x128) registration intermediate image and a 3D MP FLASH whole brain anatomical image (72 2 mm axial slices) were collected, with TR/TI/TE = 10/500/5 ms.
Functional MRI Analyses
In order to increase the overall T2 weighting of the combined image, the three ASE images were combined using an inverted signal weighted averaging algorithm [7].
Group analysis
Statistical analyses were performed using a model-based approach (General Linear Model) in FMRIB Software Library (FSL) using fMRI expert analysis tool (FEAT) version 5.3 (FMRIB's Software Library). Pre-processing steps included: motion correction using MCFLIRT [19]; brain extraction using BET [20]; spatial smoothing using a Gaussian kernel of FWHM 5 mm; mean-based intensity normalization of all volumes by the same factor; and high pass temporal filtering (0.02 Hz). FILM with local autocorrelation correction was used for the time-series statistical analysis [21]. Z-statistic images were reported using a threshold for clusters determined by Z > 3.0 and a (corrected) cluster significance threshold of P = 0.05 [22]. Images were initially registered to the high-resolution spiral image (3 degrees of freedom; DOF), then to the high-resolution T1-weighted anatomical image (6 DOF) and were finally normalized to standard space (12 DOF) using FLIRT [19, 23].
To ensure group-level activation in the WM was not due to spatial smoothing from neighbouring GM activation, a follow-up analysis was performed using the identical approach as outlined above, but without the spatial smoothing (5 mm). Importantly, these results were demonstrated using a high threshold (Z > 4.0) in order to confirm that the activation cluster was localized within WM tissue.
ROI mask creation
The two fMRI sessions were first concatenated, and motion correction was used to align the images to the first image in the concatenated volume. Activation maps were created from the combined task versus rest condition following the GLM method. From these activation maps, region of interest (ROI) masks were created. To compare activation between white and gray matter, three ROI were selected: corpus callosum (white), parietal lobes (gray), and cingulate cortex (gray). The GM areas were chosen because interhemispheric tasks have consistently been shown to elicit activation in both the parietal lobes and the cingulate cortex [8, 15]. For each individual, all activation (Z > 3.0) within the largest cluster was masked for the corpus callosum (with only callosal activation clusters that were fully separated from GM were selected). ROI masks were then created using the strongest activation for both the parietal lobes and the cingulate cortex by matching to the size of the corpus callosum ROI (±5 %). Parietal ROI masks were comprised of both right and left hemisphere activation (each size-matched to the corpus callosum ROI).
Time course analyses
Both raw signal averaging and FIR analyses were calculated at the individual level to estimate the HRF. The individual HRF estimates were then used to calculate group mean and variance of the time courses.
For the raw signal-averaging estimate of the HRF, the data were first high pass filtered to remove low frequency drifts (0.04 Hz). For each voxel in each ROI, two TRs prior to and eight TRs following each isolate were extracted. These ten TRs (20 seconds) were averaged together for each ROI in all individuals to estimate the pre- and post-stimulus time course (HRF).
FIR analysis was performed using AFNI's 3dDeconvolve program [24]. All stimulus time-points were included to produce voxel-wise estimates of the HRF for eight TRs post-stimulus (16 seconds). A second order polynomial was included in the response model to account for low-frequency fluctuations. The FIR estimate of the HRF was then averaged for each ROI for each individual.
Statistical analyses
Additional statistical analyses were conducted using a repeated measures Analysis of Variance (ANOVA). These analyses were conducted to compare the peak response amplitudes (3 time points centered around the peak) between the corpus callosum, parietal, and cingulate areas (p < 0.05).