SAROS - A large, heterogeneous, and sparsely annotated segmentation dataset on CT imaging data (SAROS)

The Sparsely Annotated Region and Organ Segmentation (SAROS) collection contributes a large heterogeneous semantic segmentation annotation dataset for building body composition analysis tools with existing CT imaging cases on TCIA.

Existing in-house segmentation models were employed to generate annotation candidates on 900 randomly selected cases (450 male, 450 female) from 29 TCIA collections. All generated annotations were manually reviewed and corrected by medical residents and students on every fifth axial slice while other slices were set to an ignore label. In total, 13 semantic body regions (Subcutaneous Tissue, Muscle, Abdominal Cavity, Thoracic Cavity, Bones, Parotid Glands, Pericardium, Breast Implant, Mediastinum, Brain, Spinal Cord, Thyroid Glands, Submandibular Glands) and 6 body part labels (Torso, Head & Neck, Right Leg, Left Leg, Arm Right, Arm Left) were annotated.