|
nif:isString
|
-
Five healthy volunteers participated in the study (mean age: 29.4±2.8 years). All participants were right-handed as evaluated by the Edinburgh Handedness Inventory [42]. Participants gave their written informed consent prior to the experiment that was conducted in conformity with the Declaration of Helsinki and approved by the local Ethics Committee of the Faculty of Psychology and Neuroscience at Maastricht University.
Before the scanning session, participants were trained for half an hour outside the scanner to get familiarized with the motor task. During the MRI session (approx. 2 h) a block design with 20-s task blocks and alternating 20-s rest periods was employed. During task blocks participants were guided by a visual metronome that consisted of a flickering ‘R’ on the right side (indicating right index finger pace) and a flickering ‘L’ on the left side (indicating left index finger pace). This display was generated using the Presentation software package (Version 16, Neurobehavioral Systems Inc., Albany, CA, USA), and was projected onto a mirror mounted in the scanner in front of the participant’s head. During the task blocks of a localization run (11 min) participants were instructed to tap at a medium speed of 2 Hz with their left index finger or their right index finger only. During the following eight experimental runs (11 min each) the participants performed four different types of tapping sequences, which required an increasing demand on bimanual coordination: 1) moving only the right index finger (unimanual), 2) moving both index fingers in synchrony (bimanual synchronous), 3) moving both index fingers at the same pace in an alternating fashion (bimanual alternating), and 4) moving the left index finger in synchrony with the right index finger, but at half of the pace (bimanual unbalanced). Each of these tapping sequences was performed at four different right-finger tapping speeds (1, 2, 3, and 4 Hz), which resulted in 16 different experimental conditions (figure 1). Eight repetitions per condition were implemented for each participant (total of 128 task blocks), with the order being counterbalanced across the session. All finger movements during tapping were recorded using a button box, and the software Presentation. The imaging session concluded with the acquisition of the anatomical images.
The images were acquired at Maastricht Brain Imaging Centre (Maastricht University) on a 3T scanner, (Magnetom Allegra, Siemens Healthcare, Germany), equipped with a standard quadrature birdcage head coil. The participants were placed comfortably in the scanner and their heads were fixed with foam cushions to minimize task-related and other spontaneous motion. All participants were instructed to avoid any movement other than the finger tapping during scanning. Functional images were acquired with a repeated single-shot echo-planar imaging (EPI) sequence with a relatively short repetition time (TR = 1000 ms), adjusted flip angle (FA = 62°), standard echo time (TE = 30 ms), field of view (FOV = 224×224 mm), matrix size (64×64), and 17 slices (thickness = 4 mm, 1 mm gap), resulting in a voxel size of 3.5×3.5×5 mm3, ensuring full coverage of the visual, parietal and motor cortices with limited coverage of prefrontal cortex and the cerebellum. Anatomical images were collected with a sequence based on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (parameters: TR = 2250 ms, TE = 2.6 ms, FA = 9°, FOV 256×256 mm2, 256×256 matrix, 192 slices, slice thickness = 1 mm, duration = 8∶26 min).
The behavioral data were analyzed using custom code in MATLAB (R2010a; The MATHWORKS Inc., Natick, MA, USA) and SPSS Statistics (PASW Statistics 18; IBM Corporation, Armonk, NY, USA). As the participants’ instruction was to perform the required order of the left and right button presses as accurately as possible, the following types of errors were defined: a) unimanual tapping: a button press with the opposite index finger, b) bimanual synchronous and bimanual unbalanced tapping: a button press that was delayed more than 100 ms in reference to the button press of the opposite index finger, c) bimanual alternating tapping: each additional consecutive button press. Secondly, the actual individual tapping speed during each of the 128 task blocks was estimated by computing the inter-response intervals between the responses made with the right index finger. Both sets of behavioral measures were analyzed using a general linear model with linear contrasts to test for a linear increase along the dimensions of speed and demand on bimanual coordination. Statistical tests for an increase on demand of bimanual coordination were performed across all tapping sequences, as well as across the three bimanual sequences (testing for a linear increase from bimanual synchronous to alternating to unbalanced). All tests were performed on group-level (fixed effects, the individual block measures were concatenated into one data set), as well as on single-subject level.
Functional and anatomical images were pre-processed and analyzed using BrainVoyager QX (Version 2.3, Brain Innovation B.V., Maastricht, The Netherlands), custom code in MATLAB, and SPSS Statistics. The first five volumes of each functional run were discarded due to T1 saturation effects. The data was pre-processed using interscan slice-time correction, 3D rigid-body motion correction, and temporal high-pass filtering with a general linear model (GLM) Fourier basis set, and up-sampled to 3×3×3 mm3. Noise artifacts were removed from the data using a linear regression approach. Estimated head motion parameters (three translational, three rotational) to model motion artifacts [28], [30], a localized estimate of the white matter signal to model scanner artifacts [29], and the ventricular signal to model physiological artifacts, were included in the noise model [26], [27]. This combination of nuisance regressors has been recommended to be efficient in increasing the specificity [27], [28], [30], as well as the reliability of the results in functional connectivity analysis [31]. All anatomical and functional data were spatially normalized to Talairach space to enable a comparison between participants [43].
Group Level Functional Network Analysis: For the functional network analysis, eight a priori regions of interest (ROIs) were selected based on the reviewed literature and a meta-analysis on finger tapping [44]; left and right primary motor cortex (M1), left and right supplementary motor area (SMA), left and right dorsal premotor cortex (dPMC), left and right visual motion area (V5) (figure 2A). All ROIs were defined individually for each participant, based on the data from the localization run, by computing a GLM with the nuisance predictors and task predictors convolved with a standard two-gamma hemodynamic response function. The ROIs were defined by selecting the 20 most significant functional voxels from the activation cluster closest to the respective coordinates reported in the meta-analysis [44]. In a second step the average time courses of these ROIs were extracted from the pre-processed and spatially normalized data from the experimental runs. Pearson’s correlation coefficients of the activation-level changes in the selected ROIs were computed block-wise for the 128 experimental task blocks of each individual. Three different sets of correlations were calculated using three different time windows: 1) a wider task window that included task on- and offset responses to measure the overall task connectivity (26-s “full task” window, encompassing the rise and decline of the positive BOLD response from 2 seconds after task onset until 28 seconds after, when the decline is expected to level off), 2) a narrow task window omitting task on- and offset responses to compute the steady-state task connectivity during continuous task performance (12-s “steady-state” task window, encompassing only the plateau of the BOLD response from 10 seconds until 22 seconds after task onset, to exclude the initial onset and peak), and 3) a narrow rest window to measure rest connectivity (12-s “rest” window, starting 12s before task onset) (figure 2B). Fisher Z transformation was applied to all correlation values to improve the normality of the calculated correlation coefficients [45]. As a preliminary step to the main analysis on single-subject level, we submitted the block-wise correlations to the same group-level statistical analyses (fixed effects) as the behavioral data.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0085929.g002 Schematic representation of functional network.All fMRI-based measures were derived from a network of a priori selected regions of interest (M1 = primary motor cortex, dPMC = dorsal premotor cortex, SMA = supplementary motor area, V5 = visual motion area), which are depicted schematically in panel A. The time windows (grey boxes) used in the functional connectivity analysis are superimposed on the schematic BOLD responses of the two regions of interest (solid and dotted line) in panel B.
Based on the group-level analysis, the functional connection showing the strongest task modulation was selected in order to further investigate the feasibility of using functional connectivity measures as a neurofeedback measure on a single-subject level. Visual inspection of the single-block blood oxygenation level-dependent (BOLD) responses from the selected ROIs confirmed that the steady-state task- and rest-connectivity (12-s steady-state task window and 12-s rest window) were not contaminated by the task on- and offset responses. For the block-wise correlations the same statistical analyses as performed at the group-level were repeated at the single-subject level. Additionally, block-wise activation level measures were computed for all 128 task blocks (% signal change 12-s task vs. 12-s rest), and submitted to the same statistical analysis. Furthermore, we investigated two different types of brain-behavior link. First, to evaluate the sensitivity and specificity of the computed brain measures to detect if a task was performed uni- or bimanually, we set the intermediate value between the means from both types of tasks as a threshold. We then computed how well the actually performed task could be detected based on the single-block brain measures using this simple threshold approach (chance level being 50%). We tested for significance through paired t-tests. Second, to estimate the criterion validity of the brain measures for indicating which tapping speed was performed, we correlated the block-wise brain measures with the block-wise actual finger tapping speed. We tested for the significance of this correlation through linear regression. Finally, to directly test for differences between the different sets of functional connectivity and activation level-based measures statistically, we submitted them pairwise to a three-way (measure × tapping speed × demand on bimanual coordination) analysis of variance (ANOVA) for repeated measures on group (fixed effects) and single-subject level.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}