A chest scan can contain information beyond the reason it was ordered. A new study of bone density and cognitive change makes that possibility more interesting for aging research, while exposing the distance between extracting a measurement and knowing what to do with it.
Published in Radiology on September 22, research led by Sara Momtazmanesh links lower spinal bone density with faster decline in global cognition and selected measures of brain white matter. It is human observational evidence, not a treatment trial or proof that strengthening bone protects memory.
TENS Magazine’s analysis is that the study advances a research connection between organ systems, while leaving clinical deployment unresolved. Read alongside the earlier paper validating the measurement method and the underlying cohort’s documentation, it presents three distinct questions: can software measure reliably, does the measurement add useful information, and does acting on that information improve outcomes?
A measurement’s track record has boundaries
The technical foundation predates the brain-aging findings. A March 2025 Radiology paper by Quincy Hathaway and colleagues examined automated three-dimensional measurements of thoracic vertebral bone density from noncontrast chest CT. Its measurements were compared with manual segmentation, and the investigators also studied subsequent vertebral fractures.
That earlier study included 2,956 participants, with longitudinal fracture data available for 1,304. Adding the bone-density information to a model using fracture-risk factors improved discrimination in that sample. But there were only 16 incident vertebral fractures, and the authors called for further validation. They also identified unanswered questions about low-dose scans, contrast-enhanced scans and reference ranges.
The comparison matters because validation belongs to a specific task. Evidence that software can delineate vertebrae and contribute to fracture prediction does not automatically validate a cognitive-risk score. Reusing the measurement saves a technical step; it does not eliminate the need to establish each new use.
The brain findings are selective
The new analysis used the Multi-Ethnic Study of Atherosclerosis, or MESA. Among 2,086 eligible participants with bone-density measurements after exclusions, 715 contributed to one or more longitudinal outcome analyses. Their median age was 69. The global cognitive composite analysis included 639 people; the two longitudinal MRI outcome samples included 405 and 408.
Lower baseline bone density was associated with faster decline in the global cognitive composite. However, results were not uniformly positive across brain measurements. The overall white-matter fractional-anisotropy association did not remain statistically significant after false-discovery-rate correction, and total white-matter hyperintensity accumulation was not significantly associated with bone density.
Two regional associations survived correction: faster fractional-anisotropy decline in the anterior limb of the internal capsule, and faster accumulation of white-matter hyperintensities in the corpus callosum. The authors describe the regional findings as hypothesis-generating and requiring independent replication.
For TENS, that pattern changes the appropriate headline claim. A collection of regional signals can justify more research without establishing a general-purpose measure of accelerated brain aging. Cognitive test change, imaging change and a future dementia diagnosis are different outcomes; evidence for one cannot silently stand in for all three.
Follow-up is part of the technology
MESA’s official protocol describes an initial community-based cohort of 6,814 adults across six US centers, aged 45 to 84 and without known cardiovascular disease at enrollment. Its first examination ran from 2000 to 2002. The new study draws on later examinations, connecting bone measurements with repeated brain and cognitive assessments.
This history is central to the result. The AI measurement becomes informative because a long-running research system supplied repeat testing, participant follow-up and contextual data. A hospital archive containing many scans does not necessarily contain comparable outcomes or equally complete follow-up.
The investigators adjusted for multiple demographic, metabolic, lifestyle and medication factors and used weighting to address differential follow-up. Nevertheless, people retained in longitudinal research may differ from those who are lost to follow-up or too unwell to participate. The paper acknowledges that residual healthy-survivor bias may limit generalizability.
TENS’s interpretation is that the next implementation test must examine both the algorithm and the surrounding data system. Reliable measurements across scanners are one requirement. Knowing whose later outcomes are missing is another. Scaling image processing alone cannot resolve that second problem.
Useful evidence before a clinical claim
The study cannot determine whether bone changes drive brain changes or whether both reflect shared aging-related processes. It lacked several biological measures needed to investigate mechanisms, including bone-turnover markers and inflammatory mediators. Follow-up may also be too short to characterize later dementia fully.
A responsible next step for research would compare prediction with and without the CT-derived measurement in an independent population, then test whether any resulting decisions improve meaningful outcomes. That is an editorial assessment of the evidence needed, not a recommendation for readers to obtain scans or change treatment.
The opportunity is to learn more from images already collected. The unresolved obligation is to show that the additional information is reproducible, useful and beneficial when acted upon. Neither the new brain study nor the earlier fracture study establishes a lifespan or healthspan benefit from this approach.
