Prenuvo, the largest direct-to-consumer whole-body MRI chain, launched the Prenuvo Research Network on September 29. The network gives academic researchers controlled access to a database of more than 200,000 scans acquired with a standardized protocol. The company says the initiative brings together collaborators from more than 50 institutions, including City of Hope, the University of California, San Francisco (UCSF), and Moffitt Cancer Center, and that it opens with four studies on incidental findings and structural changes. The announcement came in a company press release and was reported by AuntMinnie.
What the network is and what it offers researchers
The network is built on the Prenuvo Imaging Biobank, a proprietary collection of whole-body exams performed on 1.5 T scanners with standardized sequences. According to the release, the company’s in-house Medical Affairs and AI research teams handle curation, standardization, and analysis, and give collaborators de-identified datasets and statistical summaries produced with AI tools. Prenuvo says the studies follow ethical and regulatory requirements, with review by the partner institution’s IRB where required.
Prenuvo pitches the network as an alternative to slow academic funding cycles and cites AAMC figures showing that the number of competitive NIH grants fell 63% to 66% in fiscal year 2026. Alex Exuzides, PhD, vice president and head of research and medical affairs, said the goal is to generate real-world evidence on the role of whole-body MRI in clinical care. On the academic side, Moffitt thoracic oncologist Lary Robinson, MD, said his group is studying people outside current lung screening criteria, such as never-smokers and women. UCSF neurosurgeon Jacob Young, MD, said the data could help define how often incidental lesions occur and what they look like. The company now runs 26 clinics in North America, Australia, and Europe.
The four launch studies, by the numbers
The studies that open the network are conference abstracts, not full papers. The numbers below come from the abstracts posted on the company’s website.
Lung and mediastinum (City of Hope). A retrospective analysis of 118,019 scans performed between January 2020 and June 2025 on Philips (75%) and Siemens (25%) systems, reported with the ONCO-RADS structured system. Readers described 3,547 nodules in 3,388 people, with a median size of 7 mm. In 232 scans (6.8% of the nodule group, per the abstract), the lesions were rated ONCO-RADS 4 or 5, meaning potentially or highly suspicious; only 37.3% of these patients had a smoking history. Of 77 mediastinal masses, 53 were considered concerning. The authors conclude that prospective research is needed to show whether this lowers mortality.
Brain findings (UCSF). Of 78,800 people scanned from January 2024 to June 2025 at 21 clinics, 5,637 (7.1%) had at least one finding of grade 3 or higher on an internal five-point scale, where 3 means low concern. The most common was intracranial aneurysm (2.2%), followed by small vessel ischemia (0.95%), indeterminate lesion (0.7%), and infarct (0.65%). Only 2.6% of the aneurysms were larger than 7 mm. Gliomas (0.02%) and metastases (0.01%) were rare.
Kidneys (University of Michigan and University of Cambridge). Among 117,826 people with a mean age of 48, 26.7% had simple cysts and 3.1% had complex cysts or solid lesions. Of those lesions, 52% measured under 1 cm and 92.8% under 3 cm; only 155 (4.2%) were 4 cm or larger. After multivariable adjustment, only age and family history stayed associated with the larger lesions. The authors themselves point out how hard it is to separate relevant tumors from the indolent majority.
Alcohol and body structure. A matched study of 1,134 scans (567 pairs) compared abstainers with people who reported two or more drinks a day and did not have alcohol use disorder. Using automated nnU-Net and FastSurfer segmentation, the drinkers had 10% larger livers, 5.7% larger kidneys, 15.6% more visceral fat, and lower volume in 24 brain regions. This is the same kind of AI-derived body composition biomarker we covered in whole-body MRI with AI that flags disease years ahead.
What the data show and what they cannot show yet
The database really is large, and for the epidemiology of incidental findings that matters: prevalence figures from tens of thousands of exams on a single protocol are rare. The brain study, for example, compares itself with the Rotterdam Scan Study, the Dutch population cohort that became the reference for incidental intracranial findings. The trouble lies in what these data cannot answer.
First, the population is self-selected. People who pay out of pocket for a whole-body MRI tend to have higher incomes, more health anxiety and, in some cases, symptoms or a family history that prompted the scan. That is a different population from cohorts such as UK Biobank or Germany’s NAKO, which recruit by sampling and link imaging with health records over years. Second, none of the four abstracts reports histologic confirmation or outcomes. We know how many lesions were rated suspicious, but not how many were cancer, how many biopsies followed, or how many complications the workups caused. Third, there is a structural conflict of interest: company researchers co-author the work and control access to the data.
The renal numbers show the incidentaloma problem clearly. More than a quarter of those scanned had a simple cyst, and more than nine in ten solid or complex lesions were under 3 cm, a size range usually managed with surveillance. In the brain, most aneurysms are small, and unruptured aneurysms under 7 mm generally carry a low rupture risk. Finding them leads to MR angiography follow-up, anxiety, and sometimes an intervention of uncertain benefit. The value of detecting aneurysms early is also discussed in our story on AI that finds 39% more brain aneurysms on CTA.
For the lung, the natural comparison is low-dose CT. NLST showed a 20% reduction in lung cancer mortality with CT compared with chest radiography, and whole-body MRI has nothing like that yet: there is no randomized controlled trial with a mortality endpoint. MRI is also less sensitive than CT for small nodules. Strategies to expand CT screening are covered in our report on lessons from WCLC 2026 on LDCT lung screening.
Where professional societies stand
Since 2023, the American College of Radiology has said there is not enough evidence to recommend total-body screening for people without symptoms or a family history suggesting disease, citing incidental findings, costs, and cascades of follow-up testing. No major medical society recommends the exam for healthy people. As we reported when radiologists pushed back on a podcaster’s whole-body MRI claims, the debate is also about resources: magnet time and radiologist time spent on asymptomatic people compete with the queue of patients who have a clinical indication.
That does not mean whole-body MRI has no role. It is an established tool for specific indications, such as surveillance in Li-Fraumeni syndrome, assessment of multiple myeloma (standardized by MY-RADS), and staging of metastatic bone disease. ONCO-RADS, which Prenuvo uses in its reports, was proposed in 2021 to standardize whole-body MRI reporting in cancer screening. The disagreement is about using the exam in people without elevated risk.
What it means for radiology practices
For radiologists, the four studies give a practical sense of what a consumer whole-body MRI service produces: about 7% of people with a brain finding that needs at least follow-up, roughly 3% with a complex or solid renal lesion, and a share of lung nodules rated suspicious. Each of these findings needs a follow-up pathway. Without one, the risk is the same as with any incidental finding, as the case of the missed 8 mm nodule that became stage 4 lung cancer showed. In Brazil and elsewhere in Latin America, where whole-body MRI is already sold in private check-up packages, the same workload questions apply.
For groups thinking about building imaging biobanks, Prenuvo’s model also raises governance questions. Reusing clinical exams for research requires a legal basis and transparency, under HIPAA in the US, GDPR in Europe, and LGPD in Brazil. De-identifying whole-body MRI is not trivial: head scans allow facial reconstruction, and DICOM headers carry metadata that has to be cleaned systematically.
The network may produce the large-scale epidemiology of incidental findings that the literature lacks. But prevalence is not benefit. The question that matters is whether whole-body MRI in healthy people lowers mortality without causing more harm than it prevents. That question remains open, and only prospective studies with clinical endpoints and independent oversight will answer it.
Source: AuntMinnie and Prenuvo press release (PR Newswire)
