Alterations in Gray Matter Structural Networks in Amnestic Mild Cognitive Impairment: A Source-Based Morphometry Study

Abstract

Background: Amnestic mild cognitive impairment (aMCI), considered as the prodromal stage of Alzheimer's disease, is characterized by isolated memory impairment and cerebral gray matter volume (GMV) alterations. Previous structural MRI studies in aMCI have been mainly based on univariate statistics using voxel-based morphometry.

Objective: We investigated structural network differences between aMCI patients and cognitively normal older adults by using source-based morphometry, a multivariate approach that considers the relationship between voxels of various parts of the brain.

Methods: Ninety-one aMCI patients and 80 cognitively normal controls underwent structural MRI and neuropsychological assessment. Spatially independent components (ICs) that covaried between participants were estimated and a multivariate analysis of covariance was performed with ICs as dependent variables, diagnosis as independent variable, and age, sex, education level, and site as covariates.

Results: aMCI patients exhibited reduced GMV in the precentral, temporo-cerebellar, frontal, and temporal network, and increased GMV in the left superior parietal network compared to controls (pFWER < 0.05, Holm-Bonferroni correction). Moreover, we found that diagnosis, more specifically aMCI, moderated the positive relationship between occipital network and Mini-Mental State Examination scores (pFWER < 0.05, Holm-Bonferroni correction).

Conclusions: Our results showed GMV alterations in temporo-fronto-parieto-cerebellar networks in aMCI, extending previous results obtained with univariate approaches.

Keywords: Alzheimer’s disease; amnestic mild cognitive impairment; magnetic resonance imaging; source-based morphometry; structural network.

Fig. 1

Spatial maps of the 4 ICs showing significant group differences (aMCI versus Ct). For IC 5, 8, and 12: Light-dark blue colored regions show decreased GM in aMCI relative to Ct. Red-yellow-colored regions show increased gray matter in aMCI relative to Ct. For IC 2: Red-yellow-colored regions show increased gray matter in Ct relative to aMCI (or vice versa, decreased gray matter in aMCI relative to Ct). x,y,z, MNI coordinates of cluster maximum intensity. The color bar indicates the color mapping for the normalized component weights (Z-scores, thresholded at | 3 |). aMCI, amnestic mild cognitive impairment; Ct, controls; IC, Independent Component.

Fig. 2

Receiver operating characteristic (ROC) analysis of components that showed significant group differences (IC 2, 5, 8, and 12). ROC curve and the corresponding area under the curve (AUC) to differentiate between aMCI and Controls.

Fig. 3

Association between IC 6 and TMT B scores in the whole sample (pFDR < 0.05, pFWER Holm-Bonferroni = 0.26). IC, Independent Component; TMT B, Trail Making Test Part B.

Fig. 4

A) Significant positive correlations between IC 15 and diagnosis’ interaction and MMSE in the aMCI group (red, pFDR = 0.036, pFWER Holm-Bonferroni = 0.02). Negative correlations in the Ct group were not significant and are displayed in blue. B) Spatial map of IC 15. Light-dark blue colored regions show decreased gray matter in Ct relative to aMCI (or vice versa, increased gray matter in aMCI relative to Ct). Red-yellow colored regions showed decreased gray matter in aMCI relative to Ct. (Z-scores, thresholded at | 3 |). x,y,z, MNI coordinates of cluster maximum intensity. The color bar indicates the color mapping for the normalized component weights (Z-scores, thresholded at | 3 |). aMCI, amnestic mild cognitive impairment; Ct, controls; IC, Independent Component; MMSE, Mini-Mental State Examination.