<p>X-ray dark-field imaging (XDFI) has been explored as a superior alternative to conventional X-ray imaging for diagnosing many pathologic conditions. However, a simulation tool capable of reliably predicting clinical XDFI images at a human scale, is currently lacking. In this paper, we demonstrate, to the best of our knowledge, the first human-scale XDFI simulation. Using the developed simulation tool, we illustrate the strengths and limitations of XDFI for the diagnosis of emphysema, fibrosis, atelectasis, edema, and pneumonia. We augment the XCAT phantom with Voronoi grids to simulate the alveolar substructure responsible for the lung’s dark-field signal, assign material properties to each tissue type, and simulate X-ray wave propagation through the augmented XCAT phantom using a multi-layer wave-optics propagation. By altering the density and thickness of the Voronoi grids, as well as the material properties, we simulate XDFI images of normal and diseased lungs. Our simulation framework can generate realistic XDFI images of a human chest with normal or diseased lungs. The simulation confirms that the normal, emphysematous, and fibrotic lungs produce clearly distinct dark-field signals. It also shows that alveolar fluid accumulation in pneumonia, wall thickening in interstitial edema, and deflation in atelectasis all result in similar reductions in dark-field signal. It is feasible to augment XCAT with pulmonary substructure and generate realistic XDFI images using multi-layer wave optics. By providing the most realistic XDFI images of lung pathologies to date, the developed simulation framework enables in-silico clinical trials and the optimization of both hardware and software for XDFI.</p>