Brief

Researchers have developed a novel neural rendering-based optimization framework to reconstruct 3D vascular structures from sparse-view dynamic digital subtraction angiography (DSA) images. This method, termed vessel probability guided attenuation learning, aims to reduce the significant radiation exposure associated with current commercial DSA systems, which typically require hundreds of scanning views. The approach models DSA imaging as a weighted combination of static and dynamic attenuation fields, using a vessel probability field to guide the decomposition of static backgrounds and dynamic contrast agent flow, thereby improving reconstruction quality. The framework is trained by minimizing discrepancies between synthesized and real DSA images, employing progressive training and temporal consistency loss strategies to enhance geometric accuracy and temporal coherence. AI