Deep Learning-derived Bone Mineral Density and Longitudinal White Matter Microstructure and Cognitive Decline: Multi-Ethnic Study of Atherosclerosis.
Authors
Affiliations (12)
Affiliations (12)
- Department of Radiology and Radiologic Sciences, Johns Hopkins University, 601 N Caroline St, JHOC 3014, Baltimore, MD 21287.
- Department of Radiology, University of Pennsylvania, Philadelphia, Pa.
- Department of Epidemiology and Prevention, Wake Forest University School of Medicine, Winston-Salem, NC.
- Department of Radiology, University of Wisconsin School of Medicine and Public Health, Madison, Wis.
- Departments of Medicine and Epidemiology, Columbia University Irving Medical Center, New York, NY.
- Division of Cardiology, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Md.
- Neuroimage Analytics Laboratory and Biggs Institute Neuroimaging Core, Glenn Biggs Institute for Neurodegenerative Disorders, University of Texas Health Science Center at San Antonio, San Antonio, Tex.
- Department of Epidemiology, Cardiovascular Health Research Unit, University of Washington, Seattle, Wash.
- Department of Orthopaedic Surgery, Johns Hopkins University School of Medicine, Baltimore, Md.
- Lundquist Institute at Harbor-University of California Los Angeles School of Medicine, Torrance, Calif.
- Department of Internal Medicine, Section on Gerontology and Geriatric Medicine, Wake Forest University School of Medicine, Winston-Salem, NC.
- Artificial Intelligence in Biomedical Imaging Laboratory (AIBIL), Center for AI and Data Science for Integrated Diagnostics (AI2D), Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pa.
Abstract
Background White matter (WM) microstructural degeneration and WM hyperintensities (WMH) are imaging markers of brain aging. Low bone mineral density (BMD) and brain aging co-occur, but longitudinal evidence linking skeletal health to WM injury and cognitive decline remains limited. Purpose To determine whether deep learning-derived thoracic vertebral BMD (vBMD) from noncontrast chest CT is associated with longitudinal changes in WM integrity, WMH burden, and cognitive decline. Materials and Methods This secondary analysis of a prospective study conducted from September 2016 to March 2024 included participants without clinically recognized cardiovascular disease from the Multi-Ethnic Study of Atherosclerosis. Participants underwent noncontrast chest CT, multimodal brain MRI (diffusion tensor imaging and fluid-attenuated inversion recovery), and cognitive testing. Thoracic vBMD was quantified at baseline using a validated deep learning algorithm. Linear mixed-effects models related baseline vBMD to longitudinal changes in WMH volume, WM fractional anisotropy, and cognition, adjusting for demographics, apolipoprotein E-ε4, cardiometabolic risk factors, lifestyle, and medications. Results This study included 715 participants (median age [IQR], 69 years [65-75 years]; 397 men and 318 women). Lower baseline vBMD was associated with faster decline in total WM fractional anisotropy (<i>n</i> = 405; β = -0.036 SD/year/0.1 g/cm<sup>3</sup> vBMD decrease [95% CI: -0.065, -0.006]; nominal <i>P</i> = .02), although this did not survive false discovery rate correction (adjusted <i>P</i> = .07). Regionally, lower vBMD was associated with faster fractional anisotropy decline in the anterior limb of the internal capsule (β = -0.048 SD/year; adjusted <i>P</i> = .006) and faster WMH accumulation in the corpus callosum (<i>n</i> = 408; β = 12.8%/year; adjusted <i>P</i> = .04). Lower vBMD was also associated with faster decline in the global cognitive composite (<i>n</i> = 639; β = -0.025 SD/year; <i>P</i> = .002) and the Cognitive Abilities Screening Instrument (<i>n</i> = 675; β = -0.320/year; <i>P</i> < .001). Diabetes amplified associations with WMH progression (β = 4.99%; <i>P</i> = .03). Conclusion Lower vBMD from noncontrast chest CT was associated with modestly accelerated regional WM injury and faster cognitive decline, particularly among participants with diabetes, suggesting utility for detecting accelerated brain aging. ClinicalTrials.gov: NCT00005487 © RSNA, 2026 <i>Supplemental material is available for this article.</i>