Organ Annotations for 25,494 CT Scans from Stanford

25,494 CT scans from Stanford's Merlin dataset have received per-voxel annotations for organs and anatomical structures. The annotations were generated by AI models trained on over 14,000 CT scans at Johns Hopkins University. This dataset results from a collaboration between the Merlin Project at Stanford and the R-Super Project at Johns Hopkins.

Use Cases

• Train segmentation models using the per-voxel organ annotations derived from the 25,494 CT scans.
• Validate AI-generated anatomical structure labels against the original CT scan data.
• Analyze the performance of the R-Super AI training methodology on the annotated voxel data.

Strengths

• Annotations cover 25,494 CT scans, providing a substantial volume of labeled medical imaging data.
• AI models were trained on a separate, large set of over 14,000 CT scans to create the labels.
• The dataset is a product of a formal collaboration between two major academic institutions.

Limitations

• Specific column names, data structure, and file formats are unknown, complicating direct usage.
• The sample data is unavailable, preventing assessment of annotation quality or format prior to download.
• The origin and demographic/geographic coverage of the 25,494 CT scans are not detailed, which may limit generalizability.

Provenance

Source

Merlin dataset (Stanford) and AI models from Johns Hopkins University.

Collection Method

Per-voxel annotations generated by AI models trained on more than 14,000 CT scans.

Time Range

null

Freshness

null

Geography

null