PanTS: Pancreatic Tumor Segmentation with Multi-Structure Context

PanTS is a pancreatic tumor segmentation dataset released by MrGiovanni in 2025 for the NeurIPS conference. It provides medical imaging data with multi-structure context and metadata to facilitate AI-driven tumor detection, localization, and quantitative assessment.

Use Cases

• Developing segmentation algorithms using multi-structure context labels
• Evaluating tumor localization accuracy in pancreatic scans
• Performing quantitative tumor assessment using provided metadata

Strengths

• Peer-reviewed for NeurIPS 2025
• Includes multi-structure context labels for anatomy-aware modeling
• Provides metadata for quantitative assessment

Limitations

• Total record count and specific imaging modalities are not specified in the provided metadata
• Geographic diversity of the patient source is unknown

Provenance

Source

MrGiovanni (NeurIPS 2025)

Freshness

Last updated January 2026; associated with NeurIPS 2025 publication.