Morphological integration of the human brain across adolescence and adulthood

Abstract

Brain structural covariance norms capture the coordination of neurodevelopmental programs between different brain regions. We develop and apply anatomical imbalance mapping (AIM), a method to measure and model individual deviations from these norms, to provide a lifespan map of morphological integration in the human cortex. In cross-sectional and longitudinal data, analysis of whole-brain average anatomical imbalance reveals a reproducible tightening of structural covariance by age 25 y, which loosens after the seventh decade of life. Anatomical imbalance change in development and in aging is greatest in the association cortex and least in the sensorimotor cortex. Finally, we show that interindividual variation in whole-brain average anatomical imbalance is positively correlated with a marker of human prenatal stress (birthweight disparity between monozygotic twins) and negatively correlated with general cognitive ability. This work provides methods and empirical insights to advance our understanding of coordinated anatomical organization of the human brain and its interindividual variation.

Trial registration: ClinicalTrials.gov NCT00001246.

Keywords: cerebral cortex; development; lifespan.

Conflict of interest statement

The authors declare no competing interest.

Figures

Fig. 1.

Workflow for AIM. (A) First, we estimate cortical thickness at k regions across the cortex in each subject, regressed for age, sex, and Euler number. (B) We then create a linear model relating cortical thickness between pairs of regions and extract orthogonal residuals from those fits (examples shown in teal). (C) This gives a symmetric k regions × k regions matrix for each individual, where values are orthogonal distance from covariance norms (i.e., anatomical imbalance) for each pair of regions. (D) These anatomical imbalance estimates can be averaged across all pairs of regions (i.e., the upper or lower triangle of matrices depicted in C) to give a summary estimate of anatomical imbalance per individual. (E) Anatomical imbalance estimates can also be averaged by region, giving a cortical map of average anatomical imbalance, which is displayed here for the cross-sectional NIH data set.

Fig. 2.

Anatomical imbalance declines across development. (A) Decline in global anatomical imbalance across development in four developmental data sets. Each point is a unique individual. (B) Cortical maps of imbalance decline (quantified as the coefficient from a model regressing z-scored age on z-scored average regional imbalance) for each data set. Spatial correlations with NIH map: rNIH,PNC = 0.38, rNIH,ABIDE = 0.34, and rNIH, NSPN = 0.15. (C) Cortical map of fixed effects, inverse variance weighted meta-analytic effect sizes for the effect of age on regional anatomical imbalance, computed from the four maps in B.

Fig. 3.

Anatomical imbalance decline is patterned across functional networks. (A) Functional networks of Yeo, Krienen et al. (27). (B) Testing for significant enrichment of anatomical imbalance decline with development in each functional network against a null distribution generated by “spin”-based spatial permutation of the meta-analytic age/imbalance effect map (Fig. 2C, Materials and Methods, **P < 0.01, ***P < 0.001). (C) Age effects on anatomical imbalance for each functional network. (D) Relationship between slopes and intercepts of fits from C (with intercept referring to value at age 5 y). We show point estimates (dots) and SEM across regions in each network (lines) for slope and intercept coefficients.

Fig. 4.

Anatomical imbalance is dynamic across the lifespan. (A) Global anatomical imbalance of 21,711 individuals from seven data sets that tessellate the lifespan. CT estimates were harmonized across cohorts with ComBAT, and the imbalance/age relationship was modeled with smoothing splines. (B) Cortical map of imbalance/age standardized effect sizes observed in individuals 50 y of age and older from the Cam-CAN cohort. Standardized effect size was quantified as the coefficient from a model regressing scaled age on scaled average regional imbalance. (C) Testing the spatial correlation of anatomical imbalance change between the meta-analytic developmental imbalance/age map (Fig. 1C) and the Cam-CAN imbalance/age map (B) against a null hypothesis generated by a “spin”-based spatial permutation of the meta-analytic developmental map.

Fig. 5.

Individual anatomical imbalance estimates are associated with functional outcomes and subtle developmental insults. (A) Relationship between age- and sex-regressed individual anatomical imbalance estimates and full-scale IQ in NIH, ABIDE, and NSPN data sets. (B) Meta-analysis of effects from A. (C) Comparison of individual global anatomical imbalance estimates between heavier and lighter cotwins in the ABCD data set (Materials and Methods). Bolded points indicate means for each group.