Dataset for the paper "A volumetric orbital computed tomography dataset with expert-annotated radiological findings"

The 500 NIfTI format CT studies were randomly assigned unique scan IDs (001–500) and then partitioned at the scan level into a training set (scans 001–300), a validation set (scans 301–400), and a test set (scans 401–500) in a 60:20:20 ratio, without stratification by radiological label. Accordingly, both orbits from each scan were retained in the same subset. The corresponding metadata and labels are stored in a single CSV file (metadata.csv) under the root of the dataset repository.