Summary
Medical imaging AI usually works either to diagnose conditions from X-rays or to create images from descriptions, but both come from the same underlying information in the X-rays. The authors developed MedDream, a system that understands this underlying radiographic state to both help doctors read images and generate new X-rays based on clinical reports. Trained on millions of chest X-ray and text pairs, MedDream performed better than existing tools in recognizing diseases, handling limited label data, and assessing severity. It also generates realistic synthetic images that improve diagnostic accuracy when added to training data, especially when targeted to specific patient groups. This approach shows promise for AI that can both interpret medical images and generate useful clinical evidence.
Medical imagingArtificial intelligenceChest radiographLatent state representationDiagnostic reasoningSynthetic image generationClinical decision supportData augmentationGeneralizationSeverity assessment
Authors
Suyang Xi, Songtao Hu, Shansong Wang, Mojtaba Safari, Luke del Balzo, Ehsan Ul Karim, Mingzhe Hu, Kuo Zhang, Tonghe Wang, Ralph R. Weichselbaum, Xiaofeng Yang
Abstract
Medical imaging artificial intelligence (AI) is commonly developed as separate mappings from radiographs to diagnostic outputs or from clinical descriptions to generated images, although both arise from the same underlying radiographic state. A world-model formulation instead seeks to learn an internal representation of this state that can support both clinical readout and conditional simulation of radiographic observations. Here we introduce MedDream, a radiographic world model that learns a shared continuous latent state from paired chest radiograph-text observations for diagnostic reasoning and report-conditioned evidence generation. MedDream was pretrained on 2.65 million leakage-controlled chest radiograph-text pairs curated from 4.40 million candidates. Across eight clinical datasets and two independent reader cohorts, MedDream outperformed leading diagnostic and generative comparators. For diagnostic reasoning, MedDream showed strong generalization across disease recognition, label-scarce adaptation, severity assessment, and localization, while MedDream-supported review increased mean resident concordance with independent radiologist consensus from 56.3% to 63.0%. For evidence generation, MedDream produced radiographs that preserved clinically relevant pathology and improved downstream performance on held-out real data, with synthetic augmentation increasing external VinDr-CXR macro-AUROC from 76.4% to 81.4%. More importantly, conditioning generation on prespecified subgroup performance gaps enabled targeted evidence construction, increasing weighted F1 by 3.1 percentage points in Asian patients, whereas matched-volume unguided augmentation decreased it by 2.3 points. These findings establish radiographic world models as a path toward medical AI that learns clinically meaningful internal states for interpreting, simulating, and constructing evidence for clinical use.