Poster No:
1969
Submission Type:
Abstract Submission
Authors:
hwiyoung lee1, chixiang chen1, Peter Kochunov2, L. Elliot Hong2, shuo chen1
Institutions:
1University of Maryland, Baltimore, School of Medicine, Baltimore, MD, 2University of Texas Houston, Houston, TX
First Author:
hwiyoung lee
University of Maryland, Baltimore, School of Medicine
Baltimore, MD
Co-Author(s):
chixiang chen
University of Maryland, Baltimore, School of Medicine
Baltimore, MD
shuo chen
University of Maryland, Baltimore, School of Medicine
Baltimore, MD
Introduction:
Research into brain aging has garnered increasing interest due to the rapid growth of the elderly population worldwide. Cognitive decline stands out as the hallmark of brain aging.
Neuroimaging plays a central role in capturing both functional and structural changes in the brain associated with aging and cognitive functions. From a neurobiological perspective, the aging process triggers downward alterations in brain functions and structures, resulting in a decline in neurocognitive performance.
While the three-way relationship among age, the brain, and cognitive function is important in understanding cognitive aging, the majority of current research focuses on marginal associations. The limited availability of methods addressing the neurobiological aspect of cognitive aging motivates our research.
Our research aims to understand whether and to what extent the age-effect on cognitive decline can be explained by neuroimaging measures. To do this, we propose a new multivariate mediation model. The method is applied to U.K. Biobank (UKBB) data to understand the progressive loss of cognitive function with aging.
Methods:
Our proposed multivariate mediation model is built on the linear structural equation model (LSEM) framework. We introduce a novel objective function. The proposed objective function is designed to maximally uncover the effect of mediation pathways by maximizing the mediation proportion. To do this, we devise an aggregate mediation factor as a sparse linear combination of multiple neuroimaging mediators and show that maximizing the mediation proportion is equivalent to optimizing the weight vector of the linear combination. We apply a sparsity-inducing penalty to identify the active brain regions involved in cognitive aging. We also developed a numerical algorithm to implement our method.
We apply the method to a subset of the UKBB dataset, comprising 37,441 healthy participants aged between 40 and 70. Age is treated as the exposure variable, and two types of neuroimaging data are considered as mediators: 1) fractional anisotropy (FA) of white matter, derived from diffusion tensor imaging (DTI) data, and 2) cortical thicknesses (CT) calculated using MRI T1 data. Specifically, by following the ENIGMA-DTI analysis pipeline^1, we calculated the FA of 40 white matter tracts, and we used FreeSurfer to extract CT from 34 cortical gray matter regions in each hemisphere defined according to the Desikan-Killiany atlas^2 . In our analysis, the g-factor (cognitive score) serves as the outcome variable. To assess the robustness of the result from the complete dataset, we conducted an extensive validation analysis.
Results:
We identified 30 FAs involved in the mediation pathway of age-induced cognitive decline, and CT variables were not identified. To assess the consistency, we calculated the selection probabilities from the resampled data (validation analysis). The selection probabilities of FAs are generally high, with 18 FAs having a probability of 1, and another 19 FAs having probabilities between 50% and 100%. In contrast, all CTs have selection probabilities near zero, with the highest probability at 2%.
The identified white matter tracts have been frequently discussed in the literature on cognitive neuroscience. For example, we identify the cingulum^3 in both the cingulate gyrus regions and the hippocampal regions.
The estimated mediation proportion obtained from the entire dataset is 97%. Based on the resampling, the median of the mediation proportion for the independent testing dataset was 92% (the first and third quantiles were 89%, and 95%, respectively).
Conclusions:
These results confirm that first given the high mediation proportion value, we conclude that age-related cognitive decline can be almost completely mediated by neuroimaging mediators. Second, white matter integrity plays a more crucial role than cortical thickness in explaining age-induced cognitive decline. This is well aligned with previous neurobiological finding^4.
Modeling and Analysis Methods:
Methods Development 2
Multivariate Approaches 1
Keywords:
Aging
Cognition
Statistical Methods
White Matter
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
1|2Indicates the priority used for review
Provide references using author date format
1. Kochunov, P. (2015), 'Heritability of fractional anisotropy in human white matter: a comparison of Human Connectome Project and ENIGMA-DTI data', Neuroimage vol. 111, pp. 300–311.
2. Desikan, R. S. (2006), 'An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest', NeuroImage vol. 31, pp. 968-980.
3. Lin, Y. C. (2014). 'Cingulum Correlates of Cognitive Functions in Patients with Mild Cognitive Impairment and Early Alzheimers Disease: A Diffusion Spectrum Imaging Study', Brain Topography vol. 27, pp. 393–402.
4. Ziegler, D. A. (2010). 'Cognition in healthy aging is related to regional white matter integrity, but not cortical thickness', Neurobiology of Aging vol. 31, pp. 1912-1926.