The AI-native PACS uses a multimodal foundation model trained on more than a billion images to assist with exam interpretation and workflow tasks.
Raidium introduced at RSNA 2025 its new artificial intelligence (AI)-native PACS viewer powered by Curia, a multimodal and multitask foundation model for radiology trained on more than 1 billion images.
The system can help interpret imaging exams across modalities, contextualize findings, and execute workflow steps. “With this release, Raidium marks the beginning of a new era where radiology consoles are no longer a simple visualization tool. They become a clinical interface augmented by an interactive, promptable foundation model capable of executing full action sequences from a single request—while keeping the radiologist firmly in the loop,” the company says in its announcement.
The system was developed in alignment with US clinical expectations and is designed to support future regulatory pathways, according to Raidium. The launch also marks the company’s entry into the United States, following the relocation of CEO and radiologist Dr Paul Hérent.
“The goal is not to replace the radiologist, but to provide a reliable, interactive copilot capable of executing an entire workflow from a single prompt,” Hérent says in a release. “This is a major step toward general intelligence in radiology—and toward faster, more precise, and more reproducible practice.”
Radiologists can trigger tasks through visual or text prompting, including 3D segmentations, exam comparisons, measurements, and structured reporting.
In benchmarks spanning 19 radiology tasks—including anatomy, oncology, infectious diseases, degenerative disorders, and emergency imaging—the model outperformed published baselines, with average prediction accuracy comparable to or exceeding senior radiologists, the company noted.
Raidium says in a release that it is already collaborating with US academic and cancer research centers and has published related work in npj Precision Oncology and at major scientific meetings, including the American Society of Clinical Oncology and the Medical Image Computing and Computer Assisted Intervention conference.
Photo caption: Raidium view lesions
Photo credit: Raidium
