A French AI radiology company is making its US debut, and it’s starting with one of the hardest problems in cancer care: keeping track of tumors over time. Raidium announced on July 16 that its imaging platform, Raidium Read (R.Read), is now available to US oncology research centers. The company says it has already deployed the tool at Moffitt Cancer Center in Tampa, Florida, one of the country’s top cancer research institutions, where it replaced an older radiomics system.
The timing isn’t random. Oncology imaging is under serious pressure. Research shows that catching lesions early can spare most patients from surgery, and the World Health Organization is updating its guidelines around early screening for precancerous lesions. At the same time, radiologists are overwhelmed. A 2025 Philips Future Health Index survey found that 41% of radiologists feel current AI tools don’t actually address their day-to-day needs. The tools exist, but they don’t fit the workflow.
That gap is what Raidium is trying to close. The company built its viewer from scratch as an AI-native product, rather than bolting AI features onto legacy software that hasn’t changed in two decades. R.Read is now available for clinical trial and oncology research use in the US, with full regulatory clearance for routine clinical practice still pending. Raidium is pursuing 510(k) clearance for a subset of features and expects to announce a decision before the end of 2026.
What the platform actually does
Oncology follow-up imaging is one of the most labor-intensive parts of a radiologist’s job. Every follow-up scan means manually finding lesions again, measuring them, comparing them to prior studies, and documenting everything. It’s repetitive, time-consuming, and prone to inconsistency between readers.
R.Read is built specifically to automate that process. Its core capabilities include:
- Whole-body lesion detection across organ types
- AI segmentation models that measure tumors automatically using RECIST criteria, the standard used in oncology trials
- Longitudinal lesion tracking that carries measurements forward across follow-up scans
- Reduced documentation burden while keeping the radiologist in control
- Streamlined reporting for tumor boards and oncologist communication
The company claims its platform can reduce reader-to-reader variability by a factor of three, which matters significantly in clinical trials where measurement consistency directly affects outcomes data.
Why radiology AI hasn’t delivered yet
AI in radiology looks impressive on paper. Industry estimates suggest more than 70% of all FDA-approved AI-enabled medical devices fall into the radiology category. But most of those tools are narrow and siloed. They solve one problem in one part of the workflow, and radiologists end up switching between multiple systems to get through a single case. That adds clicks, adds cognitive load, and often makes the job slower, not faster.
Raidium’s argument is that the underlying software environment needs to change, not just the AI models running inside it. CEO and co-founder Paul Herent, a physician, put it directly: standard PACS viewers have resisted meaningful change for twenty years, and most AI tools were simply too limited to challenge that. His view is that agentic AI, where the system can reason and act across a workflow rather than just flag a single finding, is what finally makes a unified reading environment possible.
Moffitt Cancer Center as the first test case
Moffitt is a meaningful early partner. It’s a National Cancer Institute-designated Comprehensive Cancer Center and one of the US institutions most focused on oncology research. Cesar Lam, a radiologist in Moffitt’s Diagnostic Imaging and Interventional Radiology department, said the platform opened up research projects his team wouldn’t have considered viable before.
That’s a useful signal for other oncology centers evaluating the tool. Clinical research environments are demanding. Data quality, measurement consistency, and auditability all matter in ways that don’t always apply in routine clinical reading. If R.Read holds up in that context, it builds a reasonable case for broader adoption once the 510(k) clearance comes through.
How centers can get access
Raidium says cancer centers can start using R.Read now for clinical trial and research purposes without needing to complete complex integrations first. That’s a deliberate choice to lower the barrier to entry. Full clinical practice use is still pending regulatory clearance.
The company is accepting registrations from qualified oncology centers on an invitation basis. Interested centers can register at raidium.eu/viewer. Raidium operates out of Paris and Silicon Valley, and its foundation model, called Curia, has a peer-reviewed paper forthcoming in the journal Radiology AI.