Functional connectome fingerprinting across the lifespan

Abstract

Systematic changes have been observed in the functional architecture of the human brain with advancing age. However, functional connectivity (FC) is also a powerful feature to detect unique "connectome fingerprints," allowing identification of individuals among their peers. Although fingerprinting has been robustly observed in samples of young adults, the reliability of this approach has not been demonstrated across the lifespan. We applied the fingerprinting framework to the Cambridge Centre for Ageing and Neuroscience cohort (n = 483 aged 18 to 89 years). We found that individuals are "fingerprintable" (i.e., identifiable) across independent functional MRI scans throughout the lifespan. We observed a U-shape distribution in the strength of "self-identifiability" (within-individual correlation across modalities), and "others-identifiability" (between-individual correlation across modalities), with a decrease from early adulthood into middle age, before improving in older age. FC edges contributing to self-identifiability were not restricted to specific brain networks and were different between individuals across the lifespan sample. Self-identifiability was additionally associated with regional brain volume. These findings indicate that individual participant-level identification is preserved across the lifespan despite the fact that its components are changing nonlinearly.

Keywords: Functional connectome fingerprinting; Lifespan; fMRI.

Figure 1.

Illustration of the methodology. (A) Illustration of the fingerprinting framework. Fingerprints are established by the correlation of the functional connectivity of participants between conditions. The correlation within the same individual constitutes self-identifiability, while the correlation between individuals constitutes others-identifiability (Amico & Goñi, 2018). If the self-identifiability is higher than any other others-identifiability for a given participant, they are successfully identified (Finn et al., 2015). (B) Yeo functional networks used in the analyses of the paper. V = visual, S = somatomotor, L = limbic, DA = dorsal attention, DM = default mode, SV = salience/ventral attention, F = frontoparietal, W = whole brain. (C) Illustration of the sliding-window approach to select subgroups of participants. Subsamples of participants (window size) were chosen iteratively by taking the participants from the cohort, ordered by age, and slowly moving along (step size) the lifespan. This method yields subsets of overlapping participants across the lifespan, offering a cross-sectional, semi-continuous overview of changes during aging. Window size and step size were varied to obtain different combinations of subsamples.

Figure 2.

Unique connectomes across the lifespan, networks, and tasks. Fingerprint identifiability in the pair of Rest and Task conditions. Panel (A) illustrates the fingerprint identification accuracy across the entire sample using within- and between-network edges. The blue color in the bar graphs and the percentages (with confidence intervals; alpha = 0.05) to the right of the graphs indicate the proportion of individuals correctly identified. Network acronyms on the y-axes match graphics in Figure 1B and represent the specific functional network used for identification. In panel (B), we used a between-individual, age-group sliding-window approach to plot how stable the fingerprint identification accuracy was across the lifespan for each network.

Figure 3.

Differences in self- and others-identifiability across the lifespan. Change in self-identifiability (colors) and others-identifiability (gray) are represented using either within-network edges (A) or between-network edges (B). Each graph represents a different network, following acronyms and color schemes of Figure 1B. The beta coefficient of the age term and its quadratic term are presented at the top of the graph. We also present the adjusted R² of the model and the p value of the nested likelihood ratio indicating the nonlinearity of the relationship. The p value of predictors surviving inclusion of covariates and execution of the bootstrapping are indicated by asterisks (***p < 0.001, **p < 0.01, *p < 0.05, °p < 0.10). The age at which the curve changed direction was calculated from Stimson’s equation (Stimson et al., 1978) and is illustrated on the graphs.

Figure 4.

Distribution of nodes predicting self-identifiability across the brain. For each age window (see sliding-window approach, Figure 1C), we plot the nodal density (sum of number of edges identified by the elastic net as being predictive of self-identifiability divided by the total number of nodes) using the Schaefer atlas (400 nodes). A higher nodal density indicates that the node had a higher proportion of edges contributing to self-identifiability. Average age in each window matches averages in Figure 2.

Figure 5.

Association between gray matter volume and self-identifiability. Scatterplots presenting the association between self-identifiability, derived using (A) within- and (B) between-network edges, and gray matter volume in three gray matter morphometric networks: frontal structural network (age-sensitive network), limbic structural network (Alzheimer’s/age-related network), and visual structural network (“control” network). Data points, regression slopes, and bubbles below the graph follow the color scheme of Figure 1B. The beta coefficient of the relationship between the self-identifiability and the brain volume is indicated beside each network bubble. The p value of each predictor surviving comparison with covariates and bootstrapping is denoted by asterisks next to the beta coefficient (***p < 0.001, **p < 0.01, *p < 0.05). Models surviving all confounders for all three morphometric networks were compared using Vuong’s test for non-nested models. A letter at the bottom right of the network acronym indicates that the association was stronger using that specific structural network compared with the other networks referred to by the letter (V = visual, F = frontal, L = limbic).