Brain Volumetrics: Turning MRI Into a Serial Biomarker of Brain Aging

A biomarker hiding inside a scan you already order

What if we could measure biological brain age from the same 3T MRI a radiologist already reads to rule out a stroke or a tumor? That question sounds like a rhetorical hook, but it is now a genuinely answerable clinical question, and the answer is more interesting than most longevity marketing gives it credit for.

Standard clinical brain MRI is a qualitative exercise. A radiologist looks at the scan and eyeballs atrophy on a 5-point visual rating scale — useful for excluding an acute structural problem, but not built to detect the slow, cumulative volume loss that defines biological brain aging. Quantitative volumetrics is a different layer entirely: post-processing software segments the same 3D T1 sequence into roughly 50 to 130 anatomical structures — hippocampus and its subfields, entorhinal cortex, cortical lobes, ventricles, white matter, thalamus — returns each structure’s volume in cubic millimeters, and compares it against age- and sex-adjusted normative percentiles, the same way we already read a bone density T-score against a reference population. The core reference dataset behind the most widely used platform is documented in the Cortechs.ai NeuroQuant normative database white paper, and the current FDA clearance for the platform is on file as 510(k) K241098.

The detail that changes how I think about ordering this is that no additional scan time is required. The same 3T T1 MPRAGE sequence any radiology group already acquires can be routed to a cloud volumetric engine and returned with a percentile-graded report in about five minutes, a milestone Cortechs’ own retrospective on a decade of FDA clearance describes plainly. This is closer to a lab add-on than an imaging upgrade — the acquisition is identical, only the post-processing changes.

That distinction matters because it reframes the adoption question. This is not a new scanner, a new contrast agent, or a new radiology fellowship. It is a software layer sitting on top of an MRI sequence radiologists have run for two decades, and the clinical validation behind it is real but uneven — a 2021 systematic review found that of 17 companies then marketing quantitative brain MRI reports, only 4 had published any clinical validation at all (Neuroradiology, 2021). The technology gap has narrowed since, but it is worth remembering that “quantitative” and “validated” are not synonyms, and I want physicians ordering this to know which platforms clear that bar.

What it actually measures that matters for longevity

Four structural signals carry essentially all of the epidemiologic weight in this literature, and it is worth being specific about each rather than treating “brain volumetrics” as one undifferentiated number.

Hippocampal volume is the single most-studied structural biomarker of brain aging and Alzheimer risk, a status confirmed again in a 2025 review of quantitative MRI in Alzheimer disease. What makes it clinically interesting rather than merely academic is timing: hippocampal atrophy is measurable years before memory loss is clinically detectable on a bedside cognitive exam, which is exactly the window a preventive medicine practice is built to act in. Cortical thickness and gray matter volume by lobe extend that same logic outward from the hippocampus to the rest of the cortex, giving a regional map of where atrophy is concentrated rather than a single global number.

White matter hyperintensity (WMH) burden is the vascular half of this story, and I think it gets underweighted in most longevity conversations that fixate on amyloid and tau. WMH reflects small-vessel cerebrovascular disease, and its trajectory captures a component of cognitive decline that a pure “amyloid plus tau” framing simply misses. Finally, ventricular volume and whole-brain volume normalized to intracranial volume serve as the global neurodegeneration proxies — the numbers that move when atrophy is diffuse rather than regional. Taken together, these four measurements are what a volumetric report is actually reporting on, and each has a distinct disease-relevant story attached to it.

The brain age gap — one number, big epidemiology

If there is a single number from this field that deserves to become as familiar to longevity clinicians as an A1c or an ApoB, it is the Brain Age Gap. The method is straightforward: feed the volumetric or structural MRI data into a machine-learning model trained on a large cohort of healthy adults, generate a predicted brain age, then subtract chronological age. The result — the BAG — is a single signed number that says whether a person’s brain looks older or younger than their calendar age on structural imaging.

The pivotal analysis here is large and recent: a 2025 study in Nature Communications Medicine (see also PMC12552503) pooled 38,967 UK Biobank participants with ADNI and PPMI cohorts to link BAG directly to hard clinical endpoints, not just cross-sectional correlations.

Each 1-year increase in brain age gap raises all-cause mortality risk by 12% and Alzheimer risk by 16.5%, independent of age, sex, BMI, smoking, and alcohol.

What I want physicians to notice is the shape of the risk curve, not just the average hazard ratio. The relationship is nonlinear: risk stays essentially flat below a BAG of roughly 2 years, then rises sharply, with hazard ratios exceeding 2.0 once BAG reaches 6 to 8 years. A separate proteomic brain-age analysis of 45,113 UK Biobank participants found a similarly steep tail effect — a 6.15-fold Alzheimer risk in the extreme high-BAG group (Innovation in Aging, Oxford University Press, 2024). Put simply: for the first time we have a brain-organ-specific “biological age” number carrying the same epidemiologic weight as GrimAge or PhenoAge — except this one points directly at a modifiable organ, which is the whole reason I think this belongs in a longevity practice rather than only in a memory clinic.

The hippocampus alone tells a consistent story at smaller scale. A meta-analysis of 30 observational studies covering 13,187 individuals found that small baseline hippocampal volume predicted progression to Alzheimer disease with hazard ratios of 2.15 to 4.03 over five years (CNS Drugs, 2025). And on the vascular side, in the SPRINT-MIND cohort, WMH progression over 48 months independently predicted mild cognitive impairment or dementia even after adjusting for baseline WMH burden (Stroke and Vascular Neurology, 2022) — further evidence that the vascular structural signal is not redundant with the neurodegenerative one.

The modifiability case — what actually moves this biomarker

The reason I keep coming back to volumetrics rather than treating it as an interesting but inert number is that it responds — measurably, in randomized trials — to interventions longevity medicine is already recommending for other reasons.

Exercise has the cleanest dose-response data of any lever here. In the Erickson 2011 PNAS randomized trial (also indexed at PubMed 21282661), 120 older adults randomized to one year of moderate aerobic exercise gained 2.12% left hippocampal volume and 1.97% right, while the stretching-control group lost about 1.4% over the same year — a net 3.5% swing that roughly reverses one to two years of age-related volume loss in the single structure most tied to memory. The effect is thought to be mediated through BDNF, and the anterior hippocampus — the dentate gyrus and CA1 subfields specifically — is the most responsive subregion. A confirmatory trial in amnestic MCI found the same direction of effect with six months of aerobic training (British Journal of Sports Medicine RCT), and a 2023 CDC-published meta-analysis noted that effect sizes strengthen with intervention durations longer than six months (CDC 2023 meta-analysis) — a dosing detail worth telling patients directly, since six weeks of good intentions is not the studied window.

Diet shows a comparable magnitude of effect through a different mechanism. In UK Biobank longitudinal data, each 3-point increase in MIND diet score corresponded to roughly 20% slower gray-matter shrinkage — on the order of 2.5 years of delayed brain aging (BMJ Group, 2026; Alzheimer’s & Dementia, 2023). Higher MIND adherence tracked with larger hippocampus, thalamus, putamen, pallidum, and accumbens volumes and a lower WMH burden, while fried food and sweets were independently associated with faster hippocampal atrophy. The Lothian Birth Cohort study adds a confirmatory, longer-horizon data point: higher Mediterranean-diet adherence tracked with less total brain atrophy over three years in a cohort followed from age 73 to 76.

Blood pressure control may be the cleanest example in this entire article of a pharmacologic lever hitting a volumetric endpoint before a cognitive one. In the SPRINT-MIND MRI substudy, intensive blood pressure lowering to a systolic target under 120 reduced WMH progression and cut the combined rate of mild cognitive impairment plus probable dementia, a result summarized well by the Cleveland Clinic’s clinical summary of the trial. This is a lever most longevity practices are already pulling for cardiovascular reasons; the volumetric data simply gives it a second, independent justification.

GLP-1 receptor agonists are the section I find most clinically interesting right now, because the 2025 data actively diverges by molecule. In the ELAD trial, published in Nature Medicine in December 2025, liraglutide given to 204 patients with mild-to-moderate Alzheimer disease over 52 weeks produced nearly 50% less brain volume loss across temporal, frontal, parietal, and total gray matter, along with an 18% slower rate of cognitive decline on the ADAS-Exec measure — though the trial’s primary endpoint, cerebral glucose metabolism, did not separate from placebo. A useful plain-language summary is available from Imperial College London. Set against that positive signal, the EVOKE and EVOKE+ trials tested oral semaglutide in roughly 3,800 patients and, in November 2025, reported a negative result on the CDR-SB primary endpoint, with only a small, clinically inconsequential 10% improvement in plasma p-tau181 — a divergence recapped well by Clinical Trials Arena’s AD/PD 2026 coverage. The clean explanation on offer is pharmacokinetic: injectable liraglutide crosses the blood-brain barrier more effectively than the stability-optimized oral semaglutide formulation. What I want colleagues to notice is the methodological point — brain volumetrics is what surfaced the divergence between these two agents before cognitive endpoints did. That is exactly the job a serial imaging biomarker is supposed to perform.

Sleep and hormones deserve a shared paragraph because both illustrate that this measurement is dynamic, not fixed, and can register harm as readily as benefit. Just 36 hours of sleep deprivation produces measurable gray-matter volume changes that partially recover after a single night of recovery sleep (Frontiers in Psychiatry, 2018) — a useful reminder to standardize sleep state before any scan meant to be compared longitudinally. On the hormone side, the WHIMS-MRI substudy found that postmenopausal women on conjugated equine estrogens, with or without medroxyprogesterone, lost more gray matter in the anterior cingulate and orbitofrontal cortex compared with placebo. “More longevity drug” is not automatically “more brain preservation,” and volumetrics is one of the few tools that can catch that kind of harm directly rather than inferring it.

The honest negative case belongs here too. The FINGER trial’s 2-year multidomain MRI substudy did not find a significant hippocampal or cortical thickness advantage for the intervention arm, though there was a trend toward hippocampal preservation that did not reach significance (p = 0.085). The PREVENT-dementia LIBRA sub-study similarly found that a 24-month lifestyle intervention did not move hippocampal volume, WMH, or free-water measures. I read both of these as a timing problem rather than a null biology problem: 24 months may simply be too short a window in a relatively low-risk midlife population, and volumetric change, while real, is slow. That is precisely why baseline-plus-longitudinal is the correct framing for this biomarker, not a single before-and-after snapshot.

How this fits into a longevity program

The framing I use with patients mirrors exactly how we approach DEXA and trabecular bone score for skeletal health — measure the organ serially, don’t just guess from a single point value. Practically, that means five things. First, establish a baseline in midlife, roughly 35 to 55, while the brain is still near peak volume and before atrophy becomes visible on a qualitative read. Second, repeat serially every two to three years to track the slope rather than the point value — the slope is the actual biomarker, not any single percentile. Third, insist on the same scanner and same software for every repeat scan, because inter-vendor and inter-software variability is real and can easily be misread as biological decline (Scientific Reports, 2025 scan-rescan reliability study). Fourth, layer volumetrics with cognitive testing, APOE genotyping, and plasma p-tau217 — volumetrics is the structural layer of this picture, and the molecular biomarkers are what supply mechanism. Fifth, pair the measurement with the actionable levers already in the longevity toolkit: aerobic zone-2 exercise, a MIND or Mediterranean dietary pattern, systolic blood pressure under 120 in appropriate patients, sleep quality, and — with the 2025 caveats above squarely in view — potentially GLP-1 therapy in metabolically at-risk patients.

I want to be direct about the purpose of measuring in the first place. The point of a baseline-and-slope approach is not to generate an anxiety-inducing headline number for a patient to carry around. It is to earn the intervention — to know, with actual data rather than assumption, whether the exercise prescription, the dietary change, or the blood pressure target is doing anything to the organ we actually care about. That is the same discipline we apply to DEXA and TBS in our bone health work, and I don’t think brain aging deserves a lower evidentiary bar just because the organ is harder to picture.

None of this requires new equipment or a new scan indication. If a patient already has a clinical reason for a brain MRI — a headache workup, a concussion follow-up, a family history conversation that prompts imaging — adding volumetric quantification to that acquisition costs nothing in scan time and gives us a number worth tracking going forward.

This matters for a referring physician because not every vendor selling a percentile-graded brain report clears the same regulatory or validation bar. When I recommend volumetrics to a colleague or a patient, I want to know the specific platform their radiology group uses and whether it has a 510(k) clearance and a published normative dataset behind it — the difference between a validated clinical tool and an unvalidated consumer report is exactly this kind of detail.

The consumer landscape (because your patients are already asking)

Patients are not waiting for their physicians to bring this up. Prenuvo explicitly runs NeuroQuant on its imaging and includes brain volumetric analysis as part of its Executive membership tier, and Ezra offers a $3,999 MRI scan bundling skeletal and neurological assessment, including a brain age readout. This is a market signal, not a medical one, but it is a real one: patients will show up asking about volumetrics whether or not their primary care physician is ready for the conversation. The framing that helps me hold this steady is that this technology is already regulated, already reimbursed as part of a standard dementia workup, and producible on any 3T scanner — the actual gap is ordering behavior, not technology.

The honest limitations

1. Incidentalomas. Roughly 2.7% of asymptomatic brain MRIs show an incidental finding (number needed to scan to find one: 37), and high-resolution sequences push that figure to 4.3%; about 1.4% of findings are potentially serious — meningiomas, aneurysms, arachnoid cysts among them (BMJ, 2009; BMJ, 2018). Downstream anxiety and the follow-up imaging cascade are real costs, and patients need that counseling before the scan, not after an unexpected finding.
2. Scan-rescan reliability varies meaningfully by structure and by software; small structures such as the amygdala and entorhinal cortex carry wider coefficients of variation than whole-brain measures (Scientific Reports, 2025).
3. Normative databases skew white, Western, and educated. The percentile report should come with an equity caveat spoken aloud, particularly with patients from underrepresented backgrounds.
4. Hydration status and medications can shift measured volumes, so patients should be scanned in a consistent physiological state each time.
5. No cost-effectiveness data yet exists for using volumetrics as a screening tool in asymptomatic longevity patients. This remains cash-pay; insurers will not cover routine longitudinal brain volumetry outside a diagnosed indication.
6. Volumetrics is a lagging indicator. Atrophy is downstream of amyloid, tau, and vascular pathology, and blood-based biomarkers such as p-tau217 will catch some of that pathology earlier. The more accurate framing is that structural change is the phenotypic ground truth against which molecular signals are ultimately tested — not a replacement for them.

The bottom line

Volumetrics is the missing organ-level metric in longevity medicine — a serial biomarker of biological brain aging that is FDA-cleared, inexpensive enough to add to a 3T brain MRI a patient already has a reason to obtain, and modifiable by the same levers longevity clinicians are already pushing: exercise, diet, blood pressure control, sleep, and in select patients, metabolic therapy. It is not a magic screening test to hand every healthy 40-year-old with no risk factors. It is a slope-tracking tool for patients whose vascular, metabolic, cognitive, or family history makes brain aging a legitimate clinical question — and who want to earn their intervention rather than pay for the concept of one.

At Pravida we approach brain volumetrics the same way we approach DEXA and TBS — the value is in the trajectory, not the first number. If your MRI is already ordered for something else, or if your risk profile makes a baseline worth doing, the data supports adding volumetric quantification to that acquisition. What you do with the number afterward is the actual medicine.

Frequently asked questions

Is brain volumetrics the same thing as a “brain age test” I’ve seen advertised online?

Not exactly, though the two are related. Consumer brain age products typically report a single derived number — the brain age gap discussed above — without necessarily disclosing which normative database, which segmentation software, or which validation cohort generated it. Clinical volumetric platforms such as NeuroQuant report dozens of individual structure volumes against an FDA-cleared normative reference, and a brain age gap can be calculated from that same underlying data. The clinical version is the more transparent and more clinically actionable of the two, because it lets a physician see which specific structures are driving any overall age signal rather than handing over one opaque number.

How much does adding volumetric analysis to an existing MRI actually cost?

Because the volumetric processing rides on a scan a patient is often already obtaining for another indication, the marginal cost is the cloud-processing fee charged by the volumetric vendor rather than a new imaging charge. In a purely elective, cash-pay longevity context — without an underlying diagnostic indication — expect the total cost of scan plus volumetric report to run in the range consumer whole-body MRI programs like Prenuvo and Ezra already charge publicly. Insurance will typically cover the base MRI when there is a qualifying diagnosis, but will not currently reimburse the volumetric add-on itself when ordered purely for longevity screening.

Who is actually a good candidate for a baseline brain volumetric scan today?

I think about this in terms of risk profile rather than age alone. A reasonable candidate is a midlife patient with a first-degree relative with dementia, a cardiometabolic risk profile that includes hypertension or insulin resistance, a subjective cognitive complaint that hasn’t yet crossed into a formal diagnosis, or simply a patient who already has an MRI ordered for another reason and wants the additional data at no extra scan time. I do not think a healthy 35-year-old with no family history and no risk factors needs this as a routine screening test today — the evidence supports it as a targeted tool for people with a legitimate clinical question, not a universal longevity panel item.

References

1. Cortechs.ai NeuroQuant Normative Database white paper
2. FDA 510(k) clearance K241098
3. Cortechs.ai, 10 years of FDA clearance retrospective
4. Neuroradiology, 2021 systematic review of quantitative brain MRI validation
5. Quantitative MRI in Alzheimer disease, 2025 review
6. Nature Communications Medicine, 2025 brain age gap study
7. PMC12552503, brain age gap analysis
8. Innovation in Aging, Oxford University Press, 2024 proteomic brain-age analysis
9. CNS Drugs, 2025 hippocampal volume meta-analysis
10. Stroke and Vascular Neurology, 2022 SPRINT-MIND WMH analysis
11. Erickson et al., PNAS, 2011 exercise-hippocampus RCT
12. PubMed 21282661
13. British Journal of Sports Medicine aMCI exercise RCT
14. CDC 2023 exercise-brain volume meta-analysis
15. BMJ Group, 2026, MIND diet and brain aging
16. Alzheimer’s & Dementia, 2023 MIND diet study
17. Lothian Birth Cohort Mediterranean diet and brain atrophy study
18. JAMA, SPRINT-MIND MRI substudy
19. Cleveland Clinic summary of SPRINT-MIND substudy
20. Nature Medicine, 2025 ELAD trial (liraglutide)
21. Imperial College London summary of GLP-1/Alzheimer data
22. Novo Nordisk EVOKE/EVOKE+ trial announcement
23. Clinical Trials Arena, AD/PD 2026 recap
24. Frontiers in Psychiatry, 2018 sleep deprivation and gray matter study
25. WHIMS-MRI substudy of hormone therapy and gray matter
26. FINGER trial 2-year multidomain MRI substudy
27. PREVENT-dementia LIBRA sub-study
28. Scientific Reports, 2025 scan-rescan reliability study
29. Prenuvo, What We Offer
30. Ezra, MRI scan offerings
31. BMJ, 2009 incidental findings meta-analysis
32. BMJ, 2018 incidental findings update

Dr. Trevor Turner is a physician and co-founder of Pravida Health, a regenerative medicine and longevity practice in Atlanta, Georgia. He writes about the intersection of clinical medicine, emerging performance and longevity interventions, and how physicians should weigh dose, safety, and evidence quality before recommending a new protocol to patients.