White matter signal abnormality quality differentiates mild cognitive impairment that converts to Alzheimer's disease from nonconverters

Abstract

The objective of this study was to assess how longitudinal change in the quantity and quality of white matter signal abnormalities (WMSAs) contributes to the progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). The Mahalanobis distance of WMSA from normal-appearing white matter using T1-, T2-, and proton density-weighted MRI was defined as a quality measure for WMSA. Cross-sectional analysis of WMSA volume in 104 cognitively healthy older adults, 116 individuals with MCI who converted to AD within 3 years (mild cognitive impairment converter [MCI-C]), 115 individuals with MCI that did not convert in that time (mild cognitive impairment nonconverter [MCI-NC]), and 124 individuals with AD from the Alzheimer's Disease Neuroimaging Initiative revealed that WMSA volume was substantially greater in AD relative to the other groups but did not differ between MCI-NC and MCI-C. Longitudinally, MCI-C exhibited faster WMSA quality progression but not volume compared with matched MCI-NC beginning 18 months before MCI-C conversion to AD. The strongest difference in rate of change was seen in the time period starting 6 months before MCI-C conversion to AD and ending 6 months after conversion (p < 0.001). The relatively strong effect in this time period relative to AD conversion in the MCI-C was similar to the relative rate of change in hippocampal volume, a traditional imaging marker of AD pathology. These data demonstrate changes in white matter tissue properties that occur within WMSA in individuals with MCI that will subsequently obtain a clinical diagnosis of AD within 18 months. Individuals with AD have substantially greater WMSA volume than all MCI suggesting that there is a progressive accumulation of WMSA with progressive disease severity, and that quality change predates changes in this total volume. Given the timing of the changes in WMSA tissue quality relative to the clinical diagnosis of AD, these findings suggest that WMSAs are a critical component for this conversion and are a critical component of this clinical syndrome.

Keywords: Alzheimer's Disease Neuroimaging initiative; Alzheimer's disease; FreeSurfer; Hippocampus; Mild cognitive impairment; White matter signal abnormality.

Figure 1

MCI-C individual alignment to point of AD conversion on a total-study timeline. Bold numbers on the x-axis are months. Light numbers represent the number of data sets from MCI-C individuals at each time point. Each gray bar represents the set of all MCI-C individuals who converted at the same time point on the ADNI timeline, and each dividing line represents a data collection point during the ADNI study. The red dashed line indicates the time of AD conversion. Each MCI-C was then matched for age and sex with an MCI-NC, and this MCI-NC was aligned with its MCI-C counterpart on the total-study timeline.

Figure 2

A) Flowchart of major steps involved in automatic WMSA labeling pipeline. B) Visual representation of Mahalanobis distance, using a single WMSA voxel and 100 randomly selected NAWM voxels from an ADNI data set. Red point indicates a single WMSA voxel's position in T1/T2/PD intensity space, and blue dots indicate positions of NAWM voxels. The WMSA voxel's MD is measured from NAWM by extracting y_v, μ_H and Σ_H⁻¹ and plugging them into Equation 1. C) Comparison of the automatic WMSA labeling with only the MMGCA step to the manual labels (C.3) and comparison using the automatic labeling using MMGCA + MD refinement (C.4). This single subject is a representation of the general trend seen in the ADNI data, where the MMGCA labeling frequently failed to label connected sections of WMSA voxels that were detected by a manual labeler, and these sections were corrected with the MD refinement step. We additionally note that the “false” positives seen in C.4 are addressed in the Validation section of the Methods, and actually reflect damaged tissue that was undetected by the human rater. Subtle differences in tissue signal intensity in these regions can be seen in the T1-weighted image (C.1).

Figure 3

Cross sectional group differences in total WMSA volume and WMSA/total WM volume ratio. **Significantly different from OC, MCI-NC, and MCI-C (p < 0.0001). Error bars are standard error of the mean.

Figure 4

Time courses of WMSA/total WM volume ratio in MCI-C and MCI-NC individuals. Error bars are standard error of the mean. Red vertical line indicates time of AD conversion in MCI-C group.

Figure 5

Time courses of hippocampal volume change in MCI-NC and MCI-C individuals. **Significant interaction in rate of change (p < 0.01). Error bars are standard error of the mean. Red vertical line indicates time of AD conversion in MCI-C group.

Figure 6

Time courses of enduring WMSA MD from enduring NAWM (top) and incident WMSA from enduring NAWM (bottom) in MCI-NC and MCI-C individuals. *Significant interaction (p < 0.05), **(p < 0.01), ***(p < 0.0001). Error bars are standard error of the mean. Red vertical lines indicate time of AD conversion in MCI-C group.

Figure 7

Hypothetical model of the trajectories of WM damage progression over the course of MCI development in populations that do and do not convert to AD. The width of each bounded region corresponds to the percent of the total WM that is damaged (WMSA /total WM volume). The dashed line in the middle of each region corresponds to the mean MD of all WMSAs from NAWM (degree of damage within lesions). At 18 months prior to AD conversion, the MCI-C group exhibits a faster increase in WMSA development than the MCI-NC group. After AD conversion in the MCI-C group, volume differences start to be seen between the groups, and the MCI-C group exhibits a larger volume increase.