B1669
Title: Spatial distribution of white matter hyperintensities is related to cognition in MCI
Authors: Jordan Dworkin - Columbia University (United States) [presenting]
Elizabeth Sweeney - University of Pennsylvania (United States)
Abstract: The detection and characterization of white matter hyperintensities (WMH) on structural MRI is important across several neurological contexts. In multiple sclerosis, aging, and Alzheimer's disease, quantification has typically focused on whole-brain summaries of the total volume of the affected tissue. Recently, efforts to uncover more informative clinico-radiological relationships have motivated consideration of hyperintensities spatial distribution throughout the brain. Prior studies in stroke and traumatic brain injury have shown that connectome-based lesion-symptom mapping - a framework for quantifying lesions' spatial impacts by measuring the structural connections they are likely to impact - provides clinically relevant information. In aging and Alzheimer's disease, however, it is currently unknown (A) whether the spatial distribution of white matter hyperintensities is relevant for clinical outcomes, and (B) what level of spatial granularity best captures the relevant signal. Here, we conduct a comparison study to determine whether, and to what extent, spatial information improves prediction of cognitive performance. Cross-validated elastic net regression reveals that in healthy aging, spatial information is unrelated to cognitive performance when accounting for total WMH volume. Notably, spatial information significantly improves prediction in participants with mild cognitive impairment, with network-level dysconnectivity performing best among the tested models.
