Predicting Cognitive Performance in Older Adults from White Matter Hyperintensities with the Lesion Quantification Toolkit

Poster Session C - Sunday, April 14, 2024, 5:00 – 7:00 pm EDT, Sheraton Hall ABC

Arthur P. Hamilton¹ (arthurhamilton@cmail.carleton.ca), Kaiah Sotebeer², John G. Grundy², Cassandra Morrison¹, John A. E. Anderson¹; ¹Carleton University, ²Iowa State University

Cognitive decline can cause substantial functional impairment in older adults, but the rate of cognitive decline varies significantly between individuals. Prior research has associated cognitive decline in older adults with white matter hyperintensities (WMHs) using fluid-attenuated inversion recovery (FLAIR) MRI imaging. The recently released Lesion Quantification Toolkit (LQT) provides a measure of gray matter disruption by WMHs, which quantifies a given gray matter parcel’s level of disconnection caused by WMHs intersecting white matter tracts. An open question is whether this measure is useful in the context of normal cognitive aging. The present study sought to predict cognitive performance in cognitively normal older adults using the gray matter disruption measure. We processed FLAIR images from datasets from York University, Cornell University, and the multisite Alzheimer’s Disease Neuroimaging Initiative, and then analyzed the combined dataset (n = 303) with partial least squares path modeling. We specified causal paths from 1) gray matter disruption to cognition, 2) education to cognition, 3) age to cognition, and 4) age to gray matter disruption. Only the path from gray matter disruption to cognition was reliable (path coefficient: -0.30; 95% CI: -0.45 to -0.19). The effect was driven most heavily by subcortical regions, including the bilateral putamen, the bilateral pallidum, and cerebellar regions. The results demonstrate the value of parcellated WMH measures for predicting cognitive performance in older adults, with potential implications for early identification of pathological aging. Future analyses will examine a more diverse set of cognitive reserve and cognition variables unique to each dataset.

Topic Area: METHODS: Neuroimaging