Michigan Alzheimer's Disease Research Center

This project a focused, neuroimaging study examining how decomposing diffusion MRI signals into distinct crossing-fiber components can reveal white matter alterations associated with amnestic mild cognitive impairment (aMCI) and Alzheimer’s dementia (AD). Using cross-sectional data from 179 clinically characterized participants enrolled in the University of Michigan Memory and Aging Project (98 cognitively unimpaired, 52 with aMCI, and 29 with AD), the team applied multi-shell, multi-tissue fixel-based analysis implemented in MRtrix3 to estimate fiber density (FD) separately for up to three fiber orientations within each voxel, as well as extracellular cerebrospinal fluid (CSF) signal. This approach was chosen to address a common limitation of conventional dMRI measures, which often aggregate signals across multiple axonal populations with differing orientations and thus can obscure fiber-specific microstructural changes.

For statistical analysis, the authors used multi-block partial least squares correlation (PLS-C) to derive latent variables linking multidirectional FD and CSF images with clinical diagnostic status. The PLS-C model returned three significant latent variables that collectively characterized the relationship between diagnostic severity and fiber-specific microstructure. The first latent variable (LV1) explained 80% of the variance distinguishing cognitively unimpaired, aMCI, and AD groups and reflected a stepwise pattern: a reduction in FD alongside increased CSF signal with greater disease severity. Importantly, LV2 and LV3 captured differential patterns in smaller crossing fibers that distinguished clinical diagnoses in ways that were not simply monotonic declines. In other words, aMCI and AD participants demonstrated distinct regional patterns of both increased and decreased FD in smaller crossing fibers compared with cognitively unimpaired individuals, and pairwise PLS-C comparisons indicated overlapping but not identical FD change patterns between the aMCI and AD groups.

These findings underscore the value of decomposing crossing-fiber signals in dMRI as a more sensitive and specific neuroimaging correlate of clinical diagnosis than aggregate measures. The results challenge a prevailing assumption that white matter integrity uniformly declines in a monotonic fashion with progression from unimpaired to aMCI to AD; instead, the study reveals region- and fiber-population-specific increases and decreases in fiber density that correspond to diagnostic categories. By highlighting the unique information available when resolving multiple fiber populations and extracellular CSF contributions, the work supports the potential utility of fixel-based multi-orientation FD metrics for improving mechanistic understanding and potentially enhancing imaging-based biomarkers for early diagnosis and differentiation between aMCI and Alzheimer’s dementia.