medical imaging data suffers from the limited availability of annotation
because annotating 3D medical data is a time-consuming and expensive task.
Moreover, even if the annotation is available, supervised learning-based
approaches suffer highly imbalanced data. Most of the scans durin