Multi Anatomy X-Ray Foundation Model

2025-09-15

AI Summary

Existing X-ray AI models are predominantly confined to thoracic anatomy, exhibiting limited generalizability and insufficient support for diverse clinical tasks. To address this, we propose XR-0—the first large-scale, multi-anatomic X-ray foundation model—pretrained via self-supervised learning on a private dataset of 1.15 million images. XR-0 unifies support for multiple downstream tasks, including classification, retrieval, segmentation, localization, visual grounding, and radiology report generation. Evaluated across 12 datasets and 20 distinct tasks, XR-0 achieves state-of-the-art performance—not only substantially outperforming prior multi-anatomic models but also matching or exceeding specialized thoracic models on chest-specific tasks. Experimental results demonstrate that anatomical diversity is a critical design principle for enhancing robustness, generalizability, and clinical adaptability of medical vision models.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Search and Retrieval-Augmented AI: Web search models and ranking
• 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

X-ray imaging is a ubiquitous in radiology, yet most existing AI foundation models are limited to chest anatomy and fail to generalize across broader clinical tasks. In this work, we introduce XR-0, the multi-anatomy X-ray foundation model using self-supervised learning on a large, private dataset of 1.15 million images spanning diverse anatomical regions and evaluated across 12 datasets and 20 downstream tasks, including classification, retrieval, segmentation, localization, visual grounding, and report generation. XR-0 achieves state-of-the-art performance on most multi-anatomy tasks and remains competitive on chest-specific benchmarks. Our results demonstrate that anatomical diversity and supervision are critical for building robust, general-purpose medical vision models, paving the way for scalable and adaptable AI systems in radiology.

Problem

• Generalizing AI across multiple anatomical regions in X-ray imaging
• Addressing limitations of chest-only foundation models for clinical tasks
• Building robust medical vision models through anatomical diversity and supervision

Innovation

• Multi-anatomy X-ray foundation model
• Self-supervised learning on diverse dataset
• State-of-the-art performance across tasks