PanTS Mini: Pancreatic Tumor CT Segmentation Dataset from 145 Medical Centers

A large-scale, multi-institutional dataset containing 36,390 three-dimensional CT volumes from 145 medical centers. It includes expert-validated, voxel-wise annotations of over 993,000 anatomical structures, covering pancreatic tumors and surrounding organs. The dataset was created by JHU and is hosted by BodyMaps.

Use Cases

• Train segmentation models for pancreatic tumors based on expert-validated voxel-wise annotations.
• Develop multi-organ segmentation algorithms based on annotations for 24 surrounding anatomical structures.
• Benchmark model performance on a large-scale, multi-institutional dataset of 36,390 CT volumes.
• Conduct research on pancreatic anatomy and tumor morphology using annotated pancreas head, body, and tail structures.

Strengths

• Large scale with 36,390 three-dimensional CT volumes.
• Multi-institutional origin from 145 medical centers.
• Extensive annotations covering over 993,000 anatomical structures.
• Expert-validated, voxel-wise annotations for high precision.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic or institutional bias inherent to the contributing medical centers.

Provenance

Source

Johns Hopkins University (JHU) and 145 medical centers.

Collection Method

Collection and expert validation of CT volumes and annotations.

Time Range

null

Freshness

Last updated 2025-12-22 19:35:48; freshness should be verified.

Geography

null