Support vector machine classification and characterization of age-related reorganization of functional brain networks

Abstract

Most of what is known about the reorganization of functional brain networks that accompanies normal aging is based on neuroimaging studies in which participants perform specific tasks. In these studies, reorganization is defined by the differences in task activation between young and old adults. However, task activation differences could be the result of differences in task performance, strategy, or motivation, and not necessarily reflect reorganization. Resting-state fMRI provides a method of investigating functional brain networks without such confounds. Here, a support vector machine (SVM) classifier was used in an attempt to differentiate older adults from younger adults based on their resting-state functional connectivity. In addition, the information used by the SVM was investigated to see what functional connections best differentiated younger adult brains from older adult brains. Three separate resting-state scans from 26 younger adults (18-35 yrs) and 26 older adults (55-85) were obtained from the International Consortium for Brain Mapping (ICBM) dataset made publically available in the 1000 Functional Connectomes project www.nitrc.org/projects/fcon_1000. 100 seed-regions from four functional networks with 5mm(3) radius were defined based on a recent study using machine learning classifiers on adolescent brains. Time-series for every seed-region were averaged and three matrices of z-transformed correlation coefficients were created for each subject corresponding to each individual's three resting-state scans. SVM was then applied using leave-one-out cross-validation. The SVM classifier was 84% accurate in classifying older and younger adult brains. The majority of the connections used by the classifier to distinguish subjects by age came from seed-regions belonging to the sensorimotor and cingulo-opercular networks. These results suggest that age-related decreases in positive correlations within the cingulo-opercular and default networks, and decreases in negative correlations between the default and sensorimotor networks, are the distinguishing characteristics of age-related reorganization.

Figure 1

Shown are the 100 seed-regions used in this study taken from Dosenbach et al. (2010). Fronto-parietal network is in yellow; sensorimotor in blue; default in red; cingulo-opercular in green.

Figure 2

Pie charts illustrating the percentage of the total weight that came from each (a) scenario observed and (b) network studied.

Figure 3

Illustration of the consensus features that decreased positive correlation with age (top) and the consensus features that increased positive correlation with age and decreased negative correlation with age (bottom). Connections are scaled by their respective feature weight, with thicker connections representing greater feature weight.

Figure 4

Pie charts illustrating the breakdown of the feature weight coming from (a) connections that are more positively correlated with age, (b) connections that are less negatively correlated with age, and (c) connections that are less positively correlated with age as being between network or within network, and the breakdown of these connection types based on the contributing network or networks.

Figure 5

Pie charts illustrating the percentage of feature weight coming from between network connections and within network connections for the (a) fronto-parietal network, (b) cingulo-opercular network, (c) sensorimotor network, and (d) the default network. Shows the percentage of the feature weight coming from within and between network connections that either increased positive correlation with age, decreased negative correlation with age, or decreased positive correlation with age is also shown for each network.

Figure 6

Connections that were more positively correlated in older adults than younger adults were significantly shorter than both connections that were less negatively correlated in older adults (t(76) = -2.826, * p < 0.01)) and connections that were less positively correlated in older adults (t(84) = -3.112, **p < 0.005)). Based on Euclidian distance. Error bars are the standard error of the mean.