What the study found
A multidisciplinary team led by Sara Momtazmanesh, M.D., a postdoctoral fellow in the Department of Radiology and Radiological Sciences at Johns Hopkins University, ran a secondary analysis of the Multi-Ethnic Study of Atherosclerosis (MESA) cohort, published in RSNA's journal Radiology. Using a deep learning algorithm developed in the lab of senior author Shadpour Demehri, M.D., professor of radiology at Johns Hopkins, the team calculated baseline thoracic vertebral volumetric bone mineral density (vBMD) directly from the noncontrast chest CT scans of 2,086 MESA participants.
Of those, 715 participants also had longitudinal brain MRI and cognitive testing available, giving the researchers a cohort to track over time — 408 for white matter hyperintensity progression and 405 for fractional anisotropy, a measure of white matter microstructural integrity.
The critical detail is where the bone density measurement came from: not a dedicated DEXA scan, but the same noncontrast chest CT already performed for lung cancer screening, coronary artery calcium scoring, or pulmonary nodule follow-up. The algorithm extracted it opportunistically, with no added radiation dose and no added scan cost.
The numbers behind the headline
Lower baseline vBMD was associated with a faster decline in global cognition (β = −0.025 SD/year; P = .002) and on the Cognitive Abilities Screening Instrument (β = −0.320/year; P < .001). On brain MRI, lower vBMD tracked with faster accumulation of white matter hyperintensities in the corpus callosum (β = 12.8%/year; adjusted P = .04) and a steeper decline in fractional anisotropy in the anterior limb of the internal capsule (β = −0.048 SD/year; adjusted P = .006) — brain regions tied to executive function, memory, and attention.
The association was not uniform across the cohort: diabetes amplified the link between low bone density and white matter hyperintensity progression (β = 4.99%; P = .03), pointing toward a shared metabolic pathway rather than two unrelated processes.
Bone and brain, not cause and effect
The researchers were careful to frame this as association, not causation. "This study is not about cause and effect, but rather the observation of a metabolic syndrome that may cause both bone and brain degeneration," Dr. Demehri said, pointing to shared drivers such as insulin resistance, dyslipidemia, and menopausal hormonal change. He also described it as the first longitudinal secondary analysis linking baseline vertebral bone density to changes in white matter structure, white matter hyperintensity progression, and cognition together in the same cohort.
The chest CT is already collecting this data
Vertebral bone density joins a growing list of clinically meaningful findings that are already visible on a chest CT ordered for something else entirely. None of these require a new scan or a new radiation dose — they require a report that reliably captures what the image already contains.
| Opportunistic finding | Seen on | Why it matters |
|---|---|---|
| Vertebral bone density | Chest CT (screening, cardiac, nodule follow-up) | Predicts cognitive decline and white matter injury |
| Breast lesions | Chest CT (any indication) | 70% of scans with later-confirmed cancer went unreported at initial read |
| Renal masses | Abdominal CT/MRI (any indication) | Can be malignant; outcome depends on report precision |
| Coronary artery calcium | Chest CT (lung screening) | Predicts cardiovascular risk |
Each row is a version of the same problem: the signal is in the pixels, but it only reaches a patient's chart if the reporting workflow is built to look for it every time, not only when it's the reason for the scan.
Where structured, AI-assisted reporting fits
Opportunistic screening research like this only translates into practice if reporting can consistently extract secondary findings without slowing down the primary read. That is the case for building incidental-finding extraction — bone density, breast lesions, renal masses, coronary calcium — into structured CT reporting by default rather than leaving it to whether the finding happens to be on a radiologist's mind that day. Foundation-model based reporting is built for exactly this kind of comprehensive, structured draft: it surfaces what's visible across the whole study, xAID's in-house radiologist reviews every preliminary, and the report reaches the client's reading radiologist ready-to-sign.
Frequently asked questions
What did the study find about chest CT and brain aging?
A 2026 Radiology study by Johns Hopkins researchers found that lower thoracic vertebral bone mineral density, measured on routine noncontrast chest CT scans, was associated with faster cognitive decline and accelerated white matter injury on brain MRI. The analysis used data from the Multi-Ethnic Study of Atherosclerosis (MESA).
How was bone density measured without an extra scan?
Researchers used a deep learning algorithm to calculate thoracic vertebral volumetric bone mineral density directly from noncontrast chest CT scans that had already been performed for lung cancer screening, coronary artery calcium scoring, or pulmonary nodule follow-up — with no additional radiation dose or scan cost.
Does lower bone density cause faster brain aging?
The researchers were explicit that the study shows association, not causation. Senior author Shadpour Demehri, M.D., said the findings likely reflect a shared metabolic syndrome — including insulin resistance, dyslipidemia, and menopausal change — that drives degeneration in both bone and brain rather than one causing the other.
What does this mean for CT reporting?
It adds to a growing list of clinically meaningful signals already sitting inside routine CT scans — alongside incidental renal masses, breast lesions, and lung nodules — that go unreported unless extraction is a structured, standard part of the read. Opportunistic screening only pays off if the CT report reliably captures what the images already contain.
Source: Momtazmanesh S, et al. "Deep Learning–derived Bone Mineral Density and Longitudinal White Matter Microstructure and Cognitive Decline: Multi-Ethnic Study of Atherosclerosis." Radiology (2026). DOI: 10.1148/radiol.260656. As reported by Radiology Business and News-Medical. Figures are rounded as reported.