Deep Reinforcement Learning for Detection of Abnormal Anatomies

Automatic detection of abnormal anatomies or malformations different structures the human body is a challenging task that could provide support for clinicians in their daily practice. Compared to normative anatomies, there low presence anatomical abnormalities patients, and great variation within make it design deep learning frameworks automatic detection. We propose framework abnormality detection, which benefits from using reinforcement model landmark trained data. detect variability between predicted landmarks configurations subspace based on point distribution Procrustes shape alignment principal component analysis projection demonstrate performance this implementation clinical CT scans inner ear, show how synthetically created cochlea anatomy can be detected prediction five around cochlea. Our approach shows Receiver Operating Characteristics (ROC) Area Under The Curve (AUC) 0.97, 96% accuracy synthetic