Curia: A Multi-Modal Foundation Model for Radiology

2025-09-08

AI Summary

Current AI-assisted radiology models are predominantly single-task and narrow-domain, exhibiting poor generalizability. To address this, we introduce the first general-purpose radiology foundation model for the full spectrum of tomographic imaging modalities—including CT, MRI, and PET—trained via large-scale, self-supervised, multimodal pretraining on 130 TB of real-world clinical data (150,000 cases). Our end-to-end architecture enables, for the first time, cross-modal alignment and few-shot transfer across real-world, multi-modality tomographic scans. Evaluated on 19 external clinical tasks, the model achieves or exceeds radiologist-level performance in organ localization, intracranial hemorrhage and myocardial infarction detection, and tumor staging. It further demonstrates clinically meaningful emergent capabilities. The model weights are publicly released, establishing a new paradigm and foundational infrastructure for radiology foundation model research.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

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.

Problem

• Develops a multi-modal foundation model for radiology
• Addresses limitations of narrow single-task AI models
• Enables broad generalization across imaging modalities and diseases

Innovation

• Multi-modal foundation model for radiology
• Trained on largest real-world cross-sectional imaging corpus
• Outperforms radiologists in multi-task clinical validation