Multivariate analysis of white matter crossing fibers reveals correlations with aMCI and dementia diagnoses
Andrew R. BenderC. WeissmanNorman ScheelYoujin JungJessica S. DamoiseauxScott PeltierBenjamin M. Hampstead
0
Citation
0
Reference
10
Related Paper
Abstract:
Abstract Background Diffusion magnetic resonance imaging (dMRI) permits characterizing differences in white matter microstructure associated with amnestic mild cognitive impairment (aMCI) and Alzheimer’s dementia (AD). However, most dMRI measures aggregate signals across multiple axonal fiber populations with varying spatial orientations, which limits the sensitivity and specificity of clinical diagnosis. To overcome this shortcoming, we estimated fiber density (FD) measures, independently from crossing fiber populations, and extracellular cerebral spinal fluid (CSF). We hypothesized that aMCI and AD diagnoses are associated with differential patterns of FD changes in larger and smaller diameter fiber populations. Method We evaluated cross‐sectional dMRI data from 179 clinically characterized participants enrolled in the University of Michigan Memory and Aging Project. Image processing leveraged the MRtrix3 multi‐shell multi‐tissue fixel‐based analysis framework to estimate FD, separately for three fiber orientations, and CSF. Data analysis used multi‐block partial least squares correlation (PLS‐C) to estimate factors from multidirectional FD and CSF images correlated with differences between cognitively unimpaired (CU; n = 98) and those diagnosed with aMCI (n = 52) or AD (n = 29). Result The PLS‐C model yielded three significant latent variables (LVs; Fig. 1), reflecting patterns of both significant positive and negative associations between diagnosis and FD in crossing fibers. LV1 explained 80% of the differences from CU to aMCI to AD, demonstrating a stepwise reduction of FD and increased CSF with greater disease severity. However, LV2 and LV3 showed FD differences in smaller crossing fibers, distinguishing clinical diagnoses. Intriguingly, participants in the aMCI and AD groups showed different regions with increased or decreased FD in smaller crossing fibers, relative to CU participants. Pairwise PLS‐C models showed aMCI and AD diagnoses were associated with similar patterns of FD changes in smaller crossing fibers in overlapping regions. Conclusion The present study found distinctive patterns of white matter alterations that systematically differ across diagnostic severity in MCI and Alzheimer’s dementia. These results highlight the value of decomposing signals from crossing fibers as a sensitive neuroimaging correlate to clinical diagnosis. These findings challenge the common perspective that MCI and Alzheimer’s dementia are associated with monotonic declines in white matter integrity.Univariate
Cite
Citations (0)
"Multivariate Analysis—V: Proceodlngs of the Fifth International Symposium on Multivariate Analysis." Technometrics, 22(3), pp. 447–448
Cite
Citations (0)
Three multivariate prognostic models based on Cox's regression were tested in terms of how they predicted prognosis in a material of 134 patients with breast cancer. The multivariate models all incorporated tumor size, mitotic activity index (MAI), and axillary lymph node status in their formulas, and were originally produced through studies on different patient materials. The predictive behavior of MAI was also tested separately in the same material. The multivariate models gave roughly parallel predicted percentages of survival at two years (CV=5.2%), but showed clearly greater variation later (12.3% and 24.4% at 5 and 9 years, respectively). The results were more uniform between the multivariate models, than between the prediction by MAI and the multivariate models. The variation between repeated estimates was smaller within multivariate models than within the estimation of one of their components (MAI). We found the use of the multivariate models easy. However, traditional hospital practice does not necessarily favor the use of multivariate models, although they seem to group patients more reliably than single prognostic features.
Univariate
Cite
Citations (3)
Univariate methods are very helpful when utilized appropriately within the research analysis. However, there are many occasions in which only multivariate methods will satisfy an optimal assessment. In this case, multivariate methods will permit the researcher to incorporate many variables within a single research analysis. This work reviews the use of multivariate methods and how to apply them in clinical medicine.
Univariate
Cite
Citations (1)
Multivariate measurement systems analysis is usually performed by designing suitable gauge R&R experiments ignoring available data generated by the measurement system while used for inspection or process control. This article proposes an approach that, using the data that are routinely available from the regular activity of the instrument, offers the possibility of assessing multivariate measurement systems without the necessity of performing a multivariate gauge study. It can be carried out more frequently than a multivariate gauge R&R experiment, since can be implemented at almost no additional cost. Therefore the synergic use of the proposed approach and the traditional multivariate gauge R&R studies can be a useful strategy for improving the overall quality of multivariate measurement systems and is effective for reducing the costs of a multivariate MSA performed with a certain frequency.
Cite
Citations (1)
Cite
Citations (1)
Multivariate statistical analyses are appropriate whenever a study involves two or more outcome variables. Because multiple-outcome models reflect social reality more accurately than do conventional single-outcome or univariate models, multivariate analysis should be studied and practiced more extensively than it is. In this article, several reasons for doing multivariate analysis are presented, and two common errors in statistical analysis are discussed. Examples are presented to show how a single multivariate analysis can produce different results than do separate univariate analyses, and to illustrate the relationship between ANOVA and canonical correlation analysis.
Univariate
Canonical correlation
Statistical Analysis
Cite
Citations (98)
Univariate
Quantile
Cite
Citations (2)
Current approaches to the issue of interpreting the output of multivariate analyses are examined and two broad classes of approach distinguished: univariate and multivariate. Most researchers opt for interpretations that deal with variables one at a time rather than in combination. This approach is appropriate for univariate but not for multivariate techniques which are essentially tools for investigating combinations of psychological attributes which may determine outcome scores or category membership. Assessing the relative importance of individual variables is also inappropriate. A number of multivariate approaches that deal with combinations of variables are then examined. These include the use of simultaneous test procedures (STPs), rotation, and the simplification of canonical variates.
Univariate
Canonical correlation
Cite
Citations (18)
Abstract Hydrological events are described through a number of dependent features. To be correctly treated, the latter should be considered jointly in a multivariate framework. The aims of the multivariate hydrological frequency analysis are given and the justifications of adopting the multivariate setting in hydrology are discussed. The main steps of the analysis are described. Some extensions and perspectives are presented.
Cite
Citations (8)