Orientation-Robust Latent Motion Trajectory Learning for Annotation-free Cardiac Phase Detection in Fetal Echocardiography

On September 7, 2026, arXiv cs.CV published a study on using direction-robust potential motion trajectory learning techniques for label-free fetal echocardiographic ventricular phase detection. This method aims to overcome the technical limitations of traditional techniques that rely on manually labeled data. By extracting potential trajectory features of cardiac motion, it enables automatic identification of different ventricular phases without sample labels. The study proposes a robust mechanism that can resist image rotation and deformation interference, enhancing detection stability and accuracy in complex fetal echocardiographic images, and providing a new technical approach for automated fetal cardiac imaging analysis.