DynamicBUS: restoring temporal dynamics from static ultrasound for improved breast cancer diagnosis.

The clinical practice of archiving static 2D images from dynamic breast ultrasound (BUS) examinations loses vital temporal information, limiting computer-aided diagnosis systems. We challenge this constraint by proposing a novel framework that computationally recovers lost temporal dynamics to enhance diagnostic accuracy. Our framework first synthesizes a BUS video from a static key frame using a purpose-built generative model. We introduce a key-frame conditioning strategy to ensure the anatomical fidelity of the lesion is preserved while generating useful dynamic cues. Subsequently, the high-fidelity static image and the synthesized video are fed into Image Video network(IV-Net), a dual-branch fusion network that integrates pristine spatial details with recovered temporal context for robust classification. Evaluated on internal and external datasets, our framework outperforms methods relying solely on static images, achieving AUC of 94.13% and 82.55%, respectively. Furthermore, a reader study indicates the generated videos are indistinguishable to experts and improve diagnostic performance. Overall, our method demonstrates generative model potential to restore lost information, paving the way for reliable BUS diagnostics.