Maturation of task-induced brain activation and long range functional connectivity in adolescence revealed by multivariate pattern classification
- a Department of Psychiatry and Neuropsychology, Faculty of Health, Medicine and Life Sciences, Maastricht University, The Netherlands
- b Department of Neuropsychology and Psychopharmacology, Faculty of Psychology and Neuroscience, Maastricht University, The Netherlands
- c Centre Brain and Learning and AZIRE Research Institute, Faculty of Psychology and Education, VU Amsterdam, The Netherlands
- Received 15 September 2011
- Revised 18 November 2011
- Accepted 19 December 2011
- Available online 8 January 2012
Abstract
The present study uses multivariate pattern classification analysis to examine maturation in task-induced brain activation and in functional connectivity during adolescence. The multivariate approach allowed accurate discrimination of adolescent boys of respectively 13, 17 and 21 years old based on brain activation during a gonogo task, whereas the univariate statistical analyses showed no or only very few, small age-related clusters. Developmental differences in task activation were spatially distributed throughout the brain, indicating differences in the responsiveness of a wide range of task-related and default mode regions. Moreover, these distributed age-distinctive patterns generalized from a simple gonogo task to a cognitively and motivationally very different gambling task, and vice versa. This suggests that functional brain maturation in adolescence is driven by common processes across cognitive tasks as opposed to task-specific processes. Although we confirmed previous reports of age-related differences in functional connectivity, particularly for long range connections (> 60 mm), these differences were not specific to brain regions that showed maturation of task-induced responsiveness. Together with the task-independency of brain activation maturation, this result suggests that brain connectivity changes in the course of adolescence affect brain functionality at a basic level. This basic change is manifest in a range of tasks, from the simplest gonogo task to a complex gambling task.
Highlights
► Multivariate pattern analysis discriminates fMRI maps of 13, 17 and 21 year-olds. ► Age-affected voxels are distributed throughout the brain. ► Age-distinctive activation patterns generalize across different cognitive tasks. ► Long range functional connectivity increases from 13 to 17 and 21 years of age. ► There is no direct relationship between maturation of connectivity and activation.
Keywords
- Adolescence;
- Development;
- fMRI;
- Functional connectivity;
- Multivariate pattern classification analysis
Figures and tables from this article:
Fig. 1. Overview of image analysis steps. At the first level of analysis (left side of figure) preprocessed gonogo data of a single participant are statistically analyzed with general linear model estimation. This estimation is the starting point both for creating functional activation maps and for generating low frequency time series data for functional connectivity analysis. The latter are generated from the GLM residuals by regressing out white matter and CSF signal and band-pass filtering (0.01–0.1 Hz) (upper line of figure). The functional activation maps are generated as condition specific percent signal change maps from the estimated betas (middle line of figure). Similar percent signal change maps were available from a previous study (Keulers et al., 2011) for the conditions of a gambling task (lower line in figure). At the second level (right side of figure) the single participant data were integrated into several group level analyses. First, a univariate estimation of age effects was performed on the gonogo percent signal change maps (GLM; see Univariate analysis and Task and developmental effects revealed by the univariate voxel-wise analysis sections). These age effects were also estimated in a multivariate pattern analysis from the same percent signal change maps (MVP1; see Multivariate pattern classification between age groups, Age classification performance and Pattern of brain areas contributing to age group classification sections). In addition, the gonogo percent signal change maps are used together with the gambling percent signal change map in a multivariate pattern analysis as a test of the specificity of age discriminating patterns (MVP2; see Multivariate pattern generalization across cognitive domains and Generalizability of age-distinctive functional patterns sections). A third multivariate pattern analysis was performed on the functional connectivity strength maps, also to look for age effects (MVP3; see and sections). Lastly, a post hoc analysis of the overlap in the weight maps resulting from MVP1 and MVP3 was performed (see Relation to developmental effects in low frequency functional connectivity section).
Fig. 2. Differences in brain activation between 13 and 21 year-olds during gonogo performance revealed by the univariate voxel-wise analysis. Age group differences are presented separately for each task condition: Go (red), Go oddball (blue) and Nogo (green) trials. F-Maps are thresholded at the less stringent uncorrected significance level of p = .001 (F = 11.23) for the sake of visualization. Age-related differences are overlaid on the age-independent task-responsive activations thresholded at the False Discovery Rate (FDR) corrected significance level: task positive responding areas (white, T = 2.44), task negative responding areas (dark gray, T = 2.23), and inhibition specific areas (Nogo > Go oddball, yellow, T = 2.84) are visualized.
Fig. 3. Multivariate pattern classification accuracies per task condition and pair-wise age group comparison in relation to the progress of the recursive feature elimination process. Each graph shows the mean accuracies over the folds on the data set at each iteration of the recursive process for discriminating between respectively 13 and 17 year-olds (A), 13 and 21 year-olds (B), and 17 and 21 year-olds (C). The three lines in each graph represent the classification result for activation maps derived from different task conditions: Go (black), Go oddball (dotted) and Nogo (gray) trials. The horizontal line at 50% accuracy represents chance level. The arrows indicate the best iteration, i.e., with the highest accuracy over folds, for each task condition. This best iteration accuracy is reported as the classification result in the text and Table 1.
