Subspace-Guided Semantic and Topological Invariant Registration for Annotation-Free Ultrasound Plane Quality Control
Researchers have developed STRIQ, a novel framework for automated ultrasound image quality control that eliminates the need for manual annotations. This system uses a registration-driven approach to measure consistency within image data, employing a Latent Registration Aligner to map features between images and autonomously derived anchors. An Orthogonal Knowledge Subspace module further refines plane-specific representations to prevent interference and improve accuracy, achieving state-of-the-art correlation with clinical quality scores on benchmark datasets. AI
IMPACT This framework could streamline quality control in medical imaging, potentially improving diagnostic accuracy and reducing clinician workload.