Resting-state MRI functional connectivity as a neural correlate of multidomain lifestyle adherence in older adults at risk for Alzheimer's disease

Abstract

Prior research has demonstrated the importance of a healthy lifestyle to protect brain health and diminish dementia risk in later life. While a multidomain lifestyle provides an ecological perspective to voluntary engagement, its association with brain health is still under-investigated. Therefore, understanding the neural mechanisms underlying multidomain lifestyle engagement, particularly in older adults at risk for Alzheimer's disease (AD), gives valuable insights into providing lifestyle advice and intervention for those in need. The current study included 139 healthy older adults with familial risk for AD from the Prevent-AD longitudinal aging cohort. Self-reported exercise engagement, cognitive activity engagement, healthy diet adherence, and social activity engagement were included to examine potential phenotypes of an individual's lifestyle adherence. Two adherence profiles were discovered using data-driven clustering methodology [i.e., Adherence to healthy lifestyle (AL) group and Non-adherence to healthy lifestyle group]. Resting-state functional connectivity matrices and grey matter brain features obtained from magnetic resonance imaging were used to classify the two groups using a support vector machine (SVM). The SVM classifier was 75% accurate in separating groups. The features that show consistently high importance to the classification model were functional connectivity mainly between nodes located in different prior-defined functional networks. Most nodes were located in the default mode network, dorsal attention network, and visual network. Our results provide preliminary evidence of neurobiological characteristics underlying multidomain healthy lifestyle choices.

Figure 1

The classification analysis pipeline.

Figure 2

The distribution of each lifestyle variable (a). Scatterplots of the correlations between each pair of lifestyle variables (b). correlation coefficients and significance were displayed on each correlation pair (***p < 0.001, **p < 0.01, *p < 0.05).

Figure 3

The total within sum of squares (a) and Jaccard index across different k (b), measures of cluster variability and stability respectively.

Figure 4

Scatterplots for cluster = 2 among pairs of lifestyle variables (a). The location of black cross (X) indicates the centroid of each cluster. Mean of each lifestyle variable for the two groups derived from k-means clustering (b). Positive scores represent greater adherence to healthy lifestyle variable. Participants in Cluster 2 showed higher levels of engagement across each health behaviour (AL group) compared to participants in Cluster 1 (NAL group).

Figure 5

Group differences in apathy (a), extraversion (b), and language (c) index change. Green represents the AL group, and red represents NAL group.

Figure 6

ROC for linear SVM classification model.

Figure 7

Important functional features. (a) Anatomical map of important features and their relative importance values. Greater node size indicates higher frequency of this node appearing in this list. Thicker edge indicates greater importance value of the connection. Red indicates greater connectivity values in AL group, and gray indicates greater connectivity values in NAL group. (b) Functional connectivity directionality in both groups. Red indicates positive mean connectivity across subgroups, and blue indicates negative connectivity across subgroups. (c) Frequency of nodes from each intrinsic resting-state network (i.e., Yeo 7 networks) in the list of important features. VN (Visual Network), SMN (Sensorimotor Network), DAN (Dorsal Attention Network), VAN (Ventral Attention Network), LN (Limbic Network), FPN (Frontoparietal Network), DMN (Default Mode Network).