Idea
Curia is a multi-modal radiology foundation model enabling hospitals and AI developers to improve diagnostic accuracy across imaging types and diseases.
Research Paper
Core Innovation
This paper introduces Curia, a foundation model trained on the largest real-world multi-modal radiology dataset to date, covering 150,000 exams across imaging types. Unlike prior narrow single-task models, Curia generalizes across modalities and tasks, achieving radiologist-level performance on 19 diverse external validation tasks. It also demonstrates emergent capabilities in low-data and cross-modality scenarios.
Market Size (TAM)
$20–50B TAM, $2–10B SAM; assumption: global radiology AI market growth driven by demand for multi-modal diagnostic tools and AI integration in hospitals.
Potential Customers & Pain Points
- Hospitals needing faster and more accurate radiological diagnosis
- Radiology AI developers lacking large multi-modal datasets and benchmarks
- Medical imaging centers aiming to reduce diagnostic errors and improve workflow efficiency
Business Model
Licensing the Curia model to hospitals and AI developers via API access and enterprise software solutions; offering customization and support services.
Competitive Landscape
- Zebra Medical Vision
- Aidoc
- Qure.ai
Implementation Challenges
- Regulatory approval and clinical validation
- Integration with existing hospital IT systems
- Data privacy and security concerns
Validation Strategy
- Conduct prospective clinical trials comparing Curia to radiologist performance
- Partner with hospitals for pilot deployments and workflow integration
- Benchmark against existing radiology AI models on diverse datasets
Research Paper Overview
Curia: A Multi-Modal Foundation Model for Radiology
Summary
AI-assisted radiological interpretation is based on predominantly narrow, single-task models. This approach is impractical for covering the vast spectrum of imaging modalities, diseases, and radiological findings. Foundation models (FMs) hold the promise of broad generalization across modalities and in low-data settings. However, this potential has remained largely unrealized in radiology. We introduce Curia, a foundation model trained on the entire cross-sectional imaging output of a major hospital over several years, which to our knowledge is the largest such corpus of real-world data-encompassing 150,000 exams (130 TB). On a newly curated 19-task external validation benchmark, Curia accurately identifies organs, detects conditions like brain hemorrhages and myocardial infarctions, and predicts outcomes in tumor staging. Curia meets or surpasses the performance of radiologists and recent foundation models, and exhibits clinically significant emergent properties in cross-modality, and low-data regimes. To accelerate progress, we release our base model's weights at https://huggingface.co/raidium/curia.