Raidium Launches AI-Native Radiology Platform at Moffitt Cancer Center to Revolutionize Tumor Tracking and Clinical Research

The medical imaging landscape has entered a new era of computational precision as Raidium, a Paris and Silicon Valley-based health technology startup, officially deployed its AI-native radiology platform at the Moffitt Cancer Center. This launch represents a significant milestone in the integration of generative AI and foundation models within clinical oncology. The platform, known as Raidium Read, is designed to replace aging, legacy radiomics applications with an integrated system that treats artificial intelligence not as an external plugin, but as the fundamental architecture of the diagnostic viewer itself. Currently utilized for clinical trials and advanced research, the platform aims to solve the chronic issues of inter-reader variability and the labor-intensive nature of longitudinal tumor tracking.

A Paradigm Shift in Medical Imaging Architecture

For decades, the standard in hospital imaging has been the Picture Archiving and Communication System (PACS). While PACS revolutionized the transition from physical film to digital files, the underlying interface has remained largely stagnant for over twenty years. Traditional attempts to integrate artificial intelligence into this workflow have typically involved "bolting on" third-party algorithms to existing viewers. This often results in fragmented workflows, where radiologists must toggle between different windows, manually export data, or reconcile conflicting outputs from various narrow-AI tools.

Raidium’s approach diverges from this industry standard by introducing an AI-native viewer. By building the interface around their proprietary foundation model, Curia, Raidium has eliminated the friction between image visualization and data analysis. Curia was trained on a massive dataset comprising over 200 million CT and MRI slices derived from 150,000 comprehensive exams. This scale allows the model to function as an "organ-agnostic" system, meaning it can identify, segment, and measure lesions across the entire human anatomy without requiring specific, pre-defined modules for every different body part.

Solving the RECIST Bottleneck

At the heart of oncology research and clinical practice is the need for accurate longitudinal tracking. When a patient undergoes cancer treatment, radiologists must compare sequential scans—sometimes spanning months or years—to determine if tumors are shrinking, stable, or growing. The industry standard for this is RECIST (Response Evaluation Criteria in Solid Tumors).

Under current manual workflows, a radiologist must open a prior study, find the specific lesion, measure its longest diameter, and then find the corresponding lesion in the new scan to perform the same measurement. This process is not only tedious but prone to human error. Studies have shown that different radiologists, or even the same radiologist at different times, can produce varying measurements for the same tumor. Raidium Read addresses this by automating the detection and segmentation of lesions. According to the company, the platform reduces inter-reader variability by a factor of three, providing a level of consistency that is vital for the integrity of clinical trials.

Strategic Implementation at Moffitt Cancer Center

The selection of Moffitt Cancer Center as the launch site for Raidium Read is a strategic move that highlights the platform’s potential in high-stakes oncology environments. Based in Tampa, Florida, Moffitt is one of the premier National Cancer Institute (NCI)-designated Comprehensive Cancer Centers in the United States. It handles a massive volume of complex cases and serves as a hub for international clinical trials.

The deployment at Moffitt allows researchers to utilize Raidium Read for retrospective and prospective studies, enhancing the speed at which data can be extracted from medical images. Dr. Cesar Lam, a prominent radiologist at Moffitt, noted that the platform enables research projects that were previously considered impossible due to the sheer volume of manual labor required for precise data extraction. By automating the "heavy lifting" of measurement and mapping, the system allows radiologists to focus on high-level interpretation and patient management.

Furthermore, the implementation process for Raidium Read differs significantly from traditional software. Because the system requires no complex backend integration with existing hospital servers to function in its current research capacity, deployment is significantly faster than traditional PACS installations. This "plug-and-play" capability is a critical advantage in the fast-moving field of oncology research.

Technical Foundation: The Curia Model

The efficacy of Raidium Read is rooted in the Curia foundation model. Unlike narrow AI models that are trained to perform one specific task—such as detecting a lung nodule or measuring a liver lesion—foundation models are trained on vast, diverse datasets to develop a generalized understanding of the underlying data.

In the context of Curia, the training on 200 million slices enables the model to understand anatomical context, spatial relationships, and the visual characteristics of various pathologies across different modalities (CT and MRI). This allows for automated "mapping" of historical lesion data against new follow-up scans. When a new scan is uploaded, the system automatically identifies lesions that were previously tracked, segments them in three dimensions, and calculates the change in volume and diameter relative to previous time points.

This transition from "labeling" to "reasoning" reflects a broader trend in the AI industry. Similar to how advanced large language models (LLMs) can reason through complex medical coding or legal documents, Curia treats the task of radiology as a multi-dimensional reasoning problem. It does not just see a dark spot on a lung; it understands that the spot is a continuing manifestation of a specific lesion identified six months prior and provides the mathematical delta between the two states.

Timeline and Regulatory Path

While Raidium Read is already making an impact in the research and clinical trial space, its journey toward full clinical adoption in the United States follows a rigorous regulatory timeline. The company has announced that it expects to receive FDA 510(k) clearance before the end of 2026.

This regulatory pathway is essential for moving the platform from a research tool to a primary diagnostic device used in daily patient care. The 510(k) clearance process will require Raidium to demonstrate that its device is at least as safe and effective as legally marketed devices. Given the platform’s claim of reducing reader variability by 3x, the company is positioning itself to set a new benchmark for "substantial equivalence" in the digital health sector.

Comparative Analysis and Industry Context

The rise of Raidium occurs within a broader context of AI outperforming traditional diagnostic methods. Recent studies have demonstrated that AI algorithms can be twice as effective as physical biopsies in grading certain rare cancers, primarily because AI can analyze the entire volume of a tumor rather than just a small tissue sample. However, the bottleneck for these innovations has always been the "last mile" of clinical integration.

Many AI tools remain research prototypes because they are too cumbersome to use in a high-pressure clinical environment. By building a native viewer, Raidium is betting that the key to AI adoption is not the quality of the algorithm alone, but the quality of the user experience. This mirrors the evolution of other industries; just as the smartphone succeeded by integrating the camera, the phone, and the internet into a single native OS, Raidium seeks to integrate the viewer and the intelligence into a single clinical OS.

Broader Implications for the Healthcare Economy

The implications of the Raidium-Moffitt partnership extend beyond the walls of the radiology suite. For the pharmaceutical industry, the ability to obtain more accurate and consistent RECIST measurements can significantly impact the outcome of clinical trials. High variability in tumor measurement can lead to "noise" in the data, potentially obscuring the efficacy of a new drug or leading to incorrect conclusions about a treatment’s failure.

By providing a 3x reduction in variability, Raidium Read could theoretically allow for smaller trial sizes or more definitive results, potentially saving millions of dollars in drug development costs and bringing life-saving treatments to market faster.

For the healthcare system at large, the automation of tedious tasks addresses the growing crisis of radiologist burnout. As the volume of medical imaging continues to grow—driven by an aging population and more frequent screening—the workload on radiologists has reached unsustainable levels. Tools that can automate the repetitive aspects of the job allow clinicians to operate at the "top of their license," focusing on complex diagnostic dilemmas rather than manual measurements.

Conclusion: The Future of AI-Integrated Oncology

As Raidium moves toward its 2026 goal of FDA clearance, the launch at Moffitt Cancer Center serves as a proof-of-concept for the future of medical imaging. The transition from legacy PACS to AI-native platforms appears inevitable as the volume and complexity of medical data exceed human processing capacity.

Paul Herent, CEO and co-founder of Raidium, has framed this evolution as a long-overdue correction for a field that has resisted technological change for two decades. With the support of institutions like Moffitt and the power of foundation models like Curia, the path is being cleared for a new standard of care—one where artificial intelligence is not an optional add-on, but the very foundation upon which clinical decisions are built. The coming years will determine if this integrated approach can successfully navigate the complexities of the US healthcare regulatory environment and become the new global standard for oncology.