UltraWorld: Learning Interactive Ultrasound World Models from Untracked Clinical Videos with Acoustic Sampling Map

UltraWorld is a self-distillation approach that transfers priors from clinical ultrasound videos into interactive world models without real action annotations. It adapts a video foundation model into an ultrasound generator conditioned on reference images and anatomical masks, using programmable trajectories through 3D anatomy to synthesize action-video pairs. The Acoustic Sampling Map (AsMap) represents probe poses and imaging settings as pixel-wise 3D sampling positions, beam directions, and depths. Experiments show improved prediction fidelity and action following, with reductions in mean final distance to the goal and orientation error compared to visual servoing.