Joint cortical surface and structural connectivity analysis of Alzheimer's Disease

Abstract

Prior neuroimaging studies have demonstrated isolated structural and connectivity changes in the brain due to Alzheimer's Disease (AD). However, how these changes relate to each other is not well understood. We present a preliminary study to begin to fill this gap by leveraging joint independent component analysis (jICA). We explore how jICA performs in an analysis of T1 and diffusion weighted MRI by characterizing the joint changes of complex cortical surface and structural connectivity metrics in AD in subjects from the Baltimore Longitudinal Study of Aging. We calculate 588 region-based cortical metrics and 4,753 fractional anisotropy-based connectivity metrics and project them into a low-dimensional manifold with principal component analysis. We perform jICA on the manifold and subsequently backproject the independent components to the original data space. We demonstrate component stability with 3-fold cross validation and find differential component loadings between 776 cognitively unimpaired control subjects and 23 with AD that generalizes across folds. In addition, we perform the same analysis on the surface and connectivity metrics separately and find that the joint approach identifies both novel and similar components to the separate approaches. To illustrate the joint approach's primary utility, we provide an example hypothesis for how surface and connectivity components may vary together with AD. These preliminary results suggest jointly varying independent cortical surface and structural connectivity components can be consistently extracted from MRI data and provide a data-driven way for generating novel hypotheses about AD that may not be captured by separate analyses.

Keywords: Alzheimer’s Disease; ICA; Joint independent component analysis; cortical surface; diffusion MRI; magnetic resonance imaging; structural MRI; structural connectivity.

Figure 1.

Extraction of MRI metrics. On each subject’s T1-weighted image, we perform a cortical surface analysis, extracting the average mean curvature (MC), shape index (SI), sulcal depth (SD), cortical thickness (CT), shape complexity index (SCI), and local gyrification index (LGI) for each of 98 BrainCOLOR ROIs. This yields 588 SURF metrics. On each subject’s DWI image, we perform whole brain probabilistic tractography and calculate a structural whole brain connectome between the same 98 ROIs. Each edge in the connectome corresponds to the average FA across all streamlines connecting the two ROIs. We do not consider streamline direction, and we exclude self-connections. This yields 4,753 CONN metrics. We z-score each metric prior to further processing.

Figure 2.

Overview of ICA approaches. The general approach was to take the z-scored metrics, project them into a 20-dimensional manifold with PCA, run ICA, and backproject the resultant ICs back into data space. (a) For the joint approach, we concatenate the metrics in PCA space prior to performing ICA and split the ICs back into their SURF and CONN portions prior to backprojecting. (b and c) The non-joint ICA approaches for the SURF and CONN metrics do not require any data concatenation or splitting.

Figure 3.

Representative visualizations of joint component stability across 3-fold cross validation. We show the most and least correlated ICs across all matches.

Figure 4.

Loadings on joint ICs across 3-fold cross validation. We show the loadings for both the training and testing (withheld) sets of data. Statistical significance was determined with pair-wise Wilcoxon rank sum tests.

Figure 5.

Example of a hypothesis that can be generated with the joint approach. Joint IC 5 is a candidate for differential loading between control subjects (lower loadings) and those with AD (higher loadings). Visualization of a selected SURF metric (cortical thickness) and the CONN metrics from joint IC 5 suggest a hypothesis that can be generated from this approach is that AD is associated with decreased frontotemporal cortical thickness and inter-hemispheric structural connectivity.