Research is part ofRaidium's DNA
Our team explores foundation models and any new machine learning techniques that will help us shape the future of radiology.
We are convinced that making progress in the field requires us to all work together. This is why we're always looking for discussions with bright minds from academia and the industry.
Discover Curia
Curia is Raidium's foundation model, a new cutting-edge foundation model for 3D imaging, opening a new era in AI for precision radiology. Curia was trained on the entire cross-sectional imaging output, including over 200 millions CT and MRI slices, from head to toe. It meets or surpasses the performance of radiologists and recent foundation models on a new 19-task validation benchmark spanning degenerative disorders, infections, oncology and workflow challenges such as feature-based registration.
- 200M+
- CT & MRI slices trained on
- 19
- Task validation benchmark
Core research pillars
Raidium aims to create a fully unified AI for radiology, enabling the achievement of radiology AGI. Today, we're exploring any research topic that will allow us to reach this objective.
Curia, our first foundation model for radiology, equals or surpasses state-of-the-art models on a variety of anatomy or diagnosis tasks. We work continuously to improve our algorithm and keep building state-of-the-art models.
We utilize the capability of foundation models to combine different modalities such as images and text. In medical imaging, text often accompanies and describes the images, which AI learning uses to improve diagnostic accuracy. This multimodal approach, using Vision-Language Models (VLMs), allows for the integration of diverse patient information, supporting healthcare professionals in diagnosis, treatment planning, and automated medical report generation.
Raidium's first deployment is in oncology. With a solution that is organ-agnostic and modality agnostic, Raidium can strongly contribute to moving research forward, largely for oncology, from detection to prognostic improvement.
Commitment to open source
For Raidium, publishing our work in peer-reviewed research papers is a matter of scientific rigor. It is the most precise way to prove the exceptional efficiency and validated performance of our models. Crucially, this practice also reflects our commitment to the clinical community. By making our advancements accessible, we offer direct, supportive evidence to help radiologists trust and integrate the next generation of AI with confidence.
Pierre ManceronMScExplore our Open Source work
