OBJECTIVES: The purpose of this study was to assess whether the recently developed method of grating-based x-ray dark-field radiography can improve the diagnosis of pulmonary emphysema in vivo. MATERIALS AND METHODS: Pulmonary emphysema was induced in female C57BL/6N mice using endotracheal instillation of porcine pancreatic elastase and confirmed by in vivo pulmonary function tests, histopathology, and quantitative morphometry. The mice were anesthetized but breathing freely during imaging. Experiments were performed using a prototype small-animal x-ray dark-field scanner that was operated at 35 kilovolt (peak) with an exposure time of 5 seconds for each of the 10 grating steps. Images were compared visually. For quantitative comparison of signal characteristics, regions of interest were placed in the upper, middle, and lower zones of each lung. Receiver-operating-characteristic statistics were performed to compare the effectiveness of transmission and dark-field signal intensities and the combined parameter "normalized scatter" to differentiate between healthy and emphysematous lungs. RESULTS: A clear visual difference between healthy and emphysematous mice was found for the dark-field images. Quantitative measurements of x-ray dark-field signal and normalized scatter were significantly different between the mice with pulmonary emphysema and the control mice and showed good agreement with pulmonary function tests and quantitative histology. The normalized scatter showed a significantly higher discriminatory power (area under the receiver-operating-characteristic curve [AUC], 0.99) than dark-field (AUC, 0.90; P = 0.01) or transmission signal (AUC, 0.69; P < 0.001) alone did, allowing for an excellent discrimination of healthy and emphysematous lung regions. CONCLUSIONS: In a murine model, x-ray dark-field radiography is technically feasible in vivo and represents a substantial improvement over conventional transmission-based x-ray imaging for the diagnosis of pulmonary emphysema.