Body size and intracranial volume interact with the structure of the central nervous system: A multi-center in vivo neuroimaging study

Abstract

Clinical research emphasizes the implementation of rigorous and reproducible study designs that rely on between-group matching or controlling for sources of biological variation such as subject's sex and age. However, corrections for body size (i.e., height and weight) are mostly lacking in clinical neuroimaging designs. This study investigates the importance of body size parameters in their relationship with spinal cord (SC) and brain magnetic resonance imaging (MRI) metrics. Data were derived from a cosmopolitan population of 267 healthy human adults (age 30.1 ± 6.6 years old, 125 females). We show that body height correlates with brain gray matter (GM) volume, cortical GM volume, total cerebellar volume, brainstem volume, and cross-sectional area (CSA) of cervical SC white matter (CSA-WM; 0.44 ≤ r ≤ 0.62). Intracranial volume (ICV) correlates with body height (r = 0.46) and the brain volumes and CSA-WM (0.37 ≤ r ≤ 0.77). In comparison, age correlates with cortical GM volume, precentral GM volume, and cortical thickness (-0.21 ≥ r ≥ -0.27). Body weight correlates with magnetization transfer ratio in the SC WM, dorsal columns, and lateral corticospinal tracts (-0.20 ≥ r ≥ -0.23). Body weight further correlates with the mean diffusivity derived from diffusion tensor imaging (DTI) in SC WM (r = -0.20) and dorsal columns (-0.21), but only in males. CSA-WM correlates with brain volumes (0.39 ≤ r ≤ 0.64), and with precentral gyrus thickness and DTI-based fractional anisotropy in SC dorsal columns and SC lateral corticospinal tracts (-0.22 ≥ r ≥ -0.25). Linear mixture of age, sex, or sex and age, explained 2 ± 2%, 24 ± 10%, or 26 ± 10%, of data variance in brain volumetry and SC CSA. The amount of explained variance increased to 33 ± 11%, 41 ± 17%, or 46 ± 17%, when body height, ICV, or body height and ICV were added into the mixture model. In females, the explained variances halved suggesting another unidentified biological factor(s) determining females' central nervous system (CNS) morphology. In conclusion, body size and ICV are significant biological variables. Along with sex and age, body size should therefore be included as a mandatory variable in the design of clinical neuroimaging studies examining SC and brain structure; and body size and ICV should be considered as covariates in statistical analyses. Normalization of different brain regions with ICV diminishes their correlations with body size, but simultaneously amplifies ICV-related variance (r = 0.72 ± 0.07) and suppresses volume variance of the different brain regions (r = 0.12 ± 0.19) in the normalized measurements.

Keywords: body height and weight; brain; in vivo human neuroimaging; intracranial volume; spinal cord; structural magnetic resonance imaging.

Fig. 1.

Cross-sectional area of spinal cord white matter correlates with body height and weight. CSA, cross-sectional area; SC, spinal cord; WM, white matter; GM, gray matter; r, Pearson correlation coefficient;ρ, Spearman correlation coefficient. All spinal cord measurements were averaged from cervical C3-4 levels. Regression lines (i.e., the dashed black lines) were estimated from all available data points. Plots with statistically significant correlation (p_FWE< 0.05) are highlighted with yellow background, and corresponding r andρvalues are highlighted with black bold font.

Fig. 2.

Pearson correlation coefficient maps showing interactions between body height and morphology of the central nervous system. (a) Representative image of brain and spinal cord (SC) anatomy. The brain scan shows cortical gray matter (GM), cerebral white matter (WM), subcortical GM structures, brainstem, and cerebellum. The axial SC scan shows the WM and GM anatomy at the C3/C4 level. Image orientation is described in (a): A, anterior; P, posterior; S, superior; I, Inferior; L, left; and R, right. (b) Pearson correlation coefficient between body height and (i) cortical GM volume, (ii) cerebral WM volume, (iii) subcortical GM structure volume, (iv) brainstem volume, (v) cerebellar volume, and (vi) cross-sectional area (CSA) of cervical SC WM at C3/C4 level. The colormap for the correlation values is shown in the left bottom corner of the figure. All correlations are significant (p_FWE< 0.05). Regarding the investigated list of structures, body height demonstrated the strongest correlation with the cortical GM volume. (c) Pearson correlation coefficient between the CSA of cervical WM at C3/C4 level and (i) cortical GM volume, (ii) cerebral WM volume, (iii) subcortical GM structure volume, (iv) brainstem volume, and (v) cerebellar volume. The colormap for the correlation values is shown in the left bottom corner of the figure. All correlations are significant (p_FWE< 0.05). The correlation map shows a descending gradient from the brainstem through subcortical GM structures and cerebral WM to cortical GM. The gradient may be driven by the increasing distance to the cervical SC level and decreasing relative volume of common tract pathways. The cerebellum shows the lowest, yet significant, correlation level. This finding may be explained by the fact that the cerebrum is more strongly and directly interconnected to the peripheral nervous system via SC than the cerebellum, with spinocerebellar tracts as the primary direct connections (Chandar & Freeman, 2014;Purves et al., 2001).

Fig. 3.

Mean diffusivity and magnetization transfer ratio in spinal cord white matter correlates with body weight. GM, gray matter; WM, white matter; SC, spinal cord; FA, fractional anisotropy; MD, mean diffusivity; MTR, magnetization transfer ratio; r, Pearson correlation coefficient;ρ, Spearman correlation coefficient. All spinal cord measurements were averaged from cervical C2-5 levels. Black dashed regression lines were estimated from the Siemens and Philips scanners’ data points. Red dotted regression lines were estimated from the GE scanner’s data points. Plots with statistically significant correlation (p_FWE< 0.05) are highlighted with yellow background, and corresponding r andρvalues are highlighted with black bold font.

Fig. 4.

Brain morphology moderately correlates with body size but weakly with age. GM, gray matter; WM, white matter; Vol, volume; SubCort, subcortical; r, Pearson correlation coefficient;ρ, Spearman correlation coefficient. Regression lines (i.e., the dashed black lines) were estimated from all available data points. Plots with statistically significant correlation (p_FWE< 0.05) are highlighted with yellow background, and corresponding r andρvalues are highlighted with black bold font.

Fig. 5.

Cortical morphology correlates with body size, age, and cross-sectional area of the spinal cord white matter. CSA, cross-sectional area; SC, spinal cord; WM, white matter; GM, gray matter; PrecentralG, precentral gyrus; PostcentralG, postcentral gyrus; Vol, volume; r, Pearson correlation coefficient;ρ, Spearman correlation coefficient. Regression lines (i.e., the dashed black lines) were estimated from all available data points. Plots with statistically significant correlation (p_FWE< 0.05) are highlighted with yellow background, and corresponding r andρvalues are highlighted with black bold font. (a) Graphs demonstrate correlations with body size and age. (b) Graphs demonstrate correlation with CSA measured in the SC region as averages from cervical C3-4 levels.

Fig. 6.

Brain morphology correlates with spinal cord morphology. CSA, cross-sectional area; SC, spinal cord; WM, white matter; GM, gray matter; Vol, volume; SubCort, subcortical; r, Pearson correlation coefficient;ρ, Spearman correlation coefficient. All SC measurements were averaged from cervical C3-4 levels. Regression lines (i.e., the dashed black lines) were estimated from all available data points. Plots with statistically significant correlation (p_FWE< 0.05) are highlighted with yellow background, and corresponding r andρvalues are highlighted with black bold font.

Fig. 7.

Exploratory visualization using biplot projections of principal components. (a) Biplot projection of 1^stand 2^ndprincipal components (PCs); (b) biplot projection of 1^stand 3^rdPCs; (c) biplot projection of 1^stand 4^thPCs; (d) biplot projection of 1^stand 5^thPCs. Variable vectors are visualized in each biplot projection with a color-coding characteristic for a corresponding variable group. Variable name abbreviations and variable color codings are described as follows.Variable abbreviations:ICV, intracranial volume; MD, mean diffusivity; RD, radial diffusivity; MTR, magnetization transfer ratio; SC, spinal cord; WM, white matter; GM, gray matter; CSA, cross-sectional area; Vol, volume; PrecentralG, precentral gyrus; PostcentralG, postcentral gyrus.Variable color coding:ICV, orange; demography, green; cerebral volumes, light blue; cortical thicknesses, yellow; SC morphometry, magenta; SC WM microstructure, red.How to read a biplot:The overall domain of each component axis is <-1,1>. Each variable is characterized as a vector of magnitude in the range of <0,1> in the biplot space. Angle 0° between the component axis and variable vector with magnitude 1 (or between two variable vectors both with magnitude 1) is proportional to Pearson correlation coefficient 1. Under the same vector magnitude circumstances, an angle of 180° equals Pearson correlation coefficient -1, and angles of 90° and 270° equal Pearson correlation coefficient 0. The lower magnitude of variable vectors proportionally decreases the overall linear dependence between vector angles close to 0° or 180°, respectively. Similarly, angle deviation from 0° or 180° also decreases the level of linear dependence between pairs of vectors in the biplot.