ONCOPILOT is Raidium's dedicated tumor-analysis solution within the Raidium Read platform. It leverages the underlying AI foundation model to perform high-precision 3D tumor segmentation, automatically track tumors over time, analyze the total tumor burden, and can even offer predictions related to patient survival based on the AI's internal representations of the medical data.

How does Raidium Read benefit radiologists?

Raidium Read aims to benefit radiologists by automating time-consuming and complex tasks such as precise tumor measurement and segmentation. This automation can reduce manual workload, increase the consistency and accuracy of measurements, and potentially free up radiologists' time to focus on higher-level interpretation, complex cases, and direct patient interaction.

What is FDA 510(k) clearance?

FDA 510(k) clearance is a premarket submission made to the U.S. Food and Drug Administration to demonstrate that a medical device is at least as safe and effective as, i.e., substantially equivalent to, a legally marketed device (predicate device) that is not subject to Premarket Approval (PMA). It is a critical regulatory step required for medical devices to be legally marketed and sold for clinical use in the United States.

Which cancer center is using Raidium Read?

Moffitt Cancer Center, one of the leading oncology research institutions in the United States, is currently using Raidium Read for its clinical trials and research purposes. This early adoption provides real-world validation and feedback for the platform.

Image: courtesy of Thenextweb

techJuly 19, 2026By Veridact EditorialUpdated Jul 19

Raidium's AI-Native Radiology Viewer Aims to Reshape Oncology Imaging, Pending Critical FDA Decision

A French startup, Raidium, has introduced an AI-native radiology platform, Raidium Read, to the U.S. market, with Moffitt Cancer Center already integrating it into its clinical trial and research workflows. The platform, built from the ground up with artificial intelligence at its core, automates complex tasks like tumor measurement and segmentation, seeking to enhance precision and efficiency in cancer diagnosis and tracking. The company anticipates receiving FDA 510(k) clearance for the platform by the end of 2026, a crucial step for broader clinical adoption.

Outlook

Raidium's entry into the U.S. market with Moffitt Cancer Center as an early adopter signals a shift in how medical imaging software is designed and utilized. Unlike previous AI applications that often served as add-ons, Raidium Read is an 'AI-native' viewer, meaning artificial intelligence is fundamental to its architecture, from image interpretation to report generation. This approach suggests a move towards more integrated and intelligent imaging workflows, potentially reducing manual burdens on radiologists and increasing the consistency of tumor analysis. The forthcoming FDA 510(k) clearance by year-end is the next critical milestone, which, if granted, would pave the way for wider commercial deployment and adoption across other U.S. cancer centers and healthcare systems.

Background

Medical imaging has long been a cornerstone of cancer diagnosis and treatment planning, but the sheer volume and complexity of images present significant challenges. Radiologists routinely spend hours manually measuring tumors, comparing scans over time, and generating detailed reports. Traditional 'radiomics applications' have attempted to automate parts of this process, but often operate as separate tools, requiring radiologists to switch between different systems. This fragmentation can introduce inefficiencies and variability in measurements, which are critical for tracking disease progression and treatment response.

Raidium, a company with roots in Paris and Silicon Valley, positions its Raidium Read platform as a fundamental departure from these legacy systems. It integrates a 'world-class foundation model research lab' directly into a unified clinical viewer. A foundation model is a large AI model trained on a vast amount of data, capable of adapting to a wide range of tasks and understanding complex patterns within medical images. This allows Raidium Read to perform high-precision 3D tumor segmentation, automatically track changes over time (longitudinal tracking), calculate total tumor burden, and even offer survival prediction based on its internal representations, a feature called ONCOPILOT.

Dr. Cesar Lam, a radiologist at Moffitt Cancer Center, confirmed that the platform has already initiated an 'operational shift' within the center's Diagnostic Imaging and Interventional Radiology Department. He noted that Raidium's unified approach makes it possible to pursue research projects that were previously considered unfeasible, transforming how oncology clinical research is conducted by providing a tool specifically designed for complex analysis.

Precedents

The integration of artificial intelligence into medical imaging is not a new concept. For years, AI tools have been developed to assist radiologists with tasks such as anomaly detection, lesion classification, and even preliminary diagnoses for specific conditions like lung nodules or breast cancer. However, many of these early AI solutions were often point-solutions, designed to address a single problem or integrate into existing, often outdated, Picture Archiving and Communication Systems (PACS) as an add-on. This meant radiologists might use an AI tool for one specific task, then revert to their traditional viewer for others, creating disjointed workflows.

The industry has seen a gradual evolution from these siloed AI applications towards more integrated platforms. Companies like Gleamer and Intrasense, also from France, have developed AI-powered software for medical imaging, with Gleamer's BoneView, for instance, aiming for semi-automated diagnosis in standard radiology. However, Raidium's 'AI-native' approach marks a significant architectural shift. Instead of bolting AI onto existing infrastructure, Raidium has built its viewer from scratch with AI as the core operating principle. This mirrors a broader trend in software development where 'AI-first' or 'AI-native' design is becoming the standard, aiming to embed intelligence seamlessly into every function rather than adding it as a layer.

Historically, FDA clearance for medical imaging AI has focused on demonstrating safety and efficacy, often against human performance benchmarks. The 510(k) pathway, which Raidium is pursuing, is for devices that are substantially equivalent to a legally marketed predicate device. This process is generally faster than a Premarket Approval (PMA) but still requires rigorous data submission and validation. The success of previous AI imaging tools in securing FDA clearance provides a precedent, but each new platform, especially one as integrated and 'native' as Raidium Read, presents unique validation challenges.

The introduction of Raidium Read and its adoption by a leading institution like Moffitt Cancer Center represents more than just another AI tool entering the market. It signals a potential redefinition of the radiologist's workstation and workflow. By automating tasks like precise 3D tumor segmentation and longitudinal tracking, Raidium could free up significant time for radiologists, allowing them to focus on more complex cases, patient consultations, and research. This could alleviate some of the pressure on an already strained healthcare system facing radiologist shortages in certain areas.

Moreover, the platform's ability to reduce variability in tumor measurements is critical. In oncology, subtle changes in tumor size or characteristics can dictate treatment paths and patient prognoses. Consistent, high-precision measurements, driven by AI, could lead to more accurate treatment decisions and better patient outcomes. The 'conversational' aspect and 'visual prompting' suggest a more intuitive and interactive experience, potentially lowering the learning curve for new users and enhancing overall user satisfaction.

For the broader medical imaging industry, Raidium's 'AI-native' approach sets a new bar. It challenges existing PACS vendors to rethink their own architectures and integrate AI more deeply, rather than superficially. The success or failure of Raidium's FDA clearance and subsequent market penetration will offer valuable insights into the regulatory and commercial viability of truly AI-centric medical devices. This move could accelerate the development of similar integrated AI solutions across various medical specialties, driving a new wave of innovation in diagnostic imaging.

Scenarios

Analysis

One immediate outcome, assuming successful FDA 510(k) clearance by the end of 2026, is that Raidium Read could move beyond research and clinical trials at Moffitt Cancer Center and become available for broader clinical use. This would allow oncologists and radiologists to integrate its advanced AI capabilities directly into routine patient care, potentially improving diagnostic accuracy and efficiency across a wider patient population.

Another possible outcome is that Raidium's success could spur other medical imaging software developers to accelerate their own 'AI-native' initiatives. The approach of building a viewer from scratch with AI at its core, rather than layering it onto legacy systems, could become a new industry standard. This might lead to a competitive wave of innovation, where existing PACS vendors are compelled to acquire or develop similar integrated solutions to remain competitive.

Conversely, if Raidium faces unexpected delays or difficulties in securing FDA clearance, its momentum could slow. Regulatory hurdles can be complex, and even minor issues can prolong the approval process, impacting market entry and adoption rates. This could give competitors more time to develop their own integrated AI solutions, or allow existing, less 'native' AI tools to solidify their market position further before Raidium can fully compete.

Furthermore, the long-term impact on the role of radiologists remains a key question. While the platform aims to augment their capabilities and reduce manual work, widespread adoption of highly automated AI viewers could reshape the demand for certain radiological tasks. Radiologists may shift their focus towards more interpretive, complex, and patient-facing roles, relying on AI for the initial, high-volume analysis.

Timeline

2026-07-16

Raidium Read U.S. Launch

Raidium officially announced the U.S. launch of its AI-native radiology platform, Raidium Read, for oncology research centers.

2026-07-16

Moffitt Cancer Center Deployment

Raidium Read was deployed at Moffitt Cancer Center, where it is currently used for clinical trials and research, replacing legacy radiomics applications.

2026-12-31

Anticipated FDA 510(k) Clearance

Raidium expects to receive FDA 510(k) clearance for its Raidium Read platform before the end of 2026, which is crucial for broader clinical use.

Frequently Asked Questions

An 'AI-native' radiology viewer means the artificial intelligence is not just an add-on feature but is built into the fundamental design and core functionality of the software. It uses large AI models, often called foundation models, to power every aspect of image interpretation, analysis, and workflow from the ground up, rather than integrating separate AI tools into an older, non-AI-centric system.

Discussion

Be the first to share your thoughts.