PanTS: The Pancreatic Tumor Segmentation Dataset

2025-07-01

AI Summary

Limited by small-scale data and insufficient anatomical context, pancreatic tumor detection, localization, and segmentation remain challenging. To address this, we introduce PanTS—the largest and most comprehensive multicenter pancreatic CT dataset to date—comprising 36,390 contrast-enhanced CT scans and over 993,000 voxel-level annotations covering tumors and 24 surrounding anatomical structures. We propose a novel multi-institutional collaborative annotation framework, incorporating expert-validated fine-grained labeling and rich metadata (e.g., age, sex, diagnosis, contrast phase). PanTS contains 16× more tumor annotations than all existing public datasets combined and is the first to systematically support tumor–anatomy co-modeling. Models trained on PanTS achieve statistically significant improvements over state-of-the-art baselines on detection, localization, and segmentation tasks, establishing new performance benchmarks. This work demonstrates that large-scale, high-fidelity anatomical context data is critical for advancing AI-driven pancreatic tumor analysis.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

PanTS is a large-scale, multi-institutional dataset curated to advance research in pancreatic CT analysis. It contains 36,390 CT scans from 145 medical centers, with expert-validated, voxel-wise annotations of over 993,000 anatomical structures, covering pancreatic tumors, pancreas head, body, and tail, and 24 surrounding anatomical structures such as vascular/skeletal structures and abdominal/thoracic organs. Each scan includes metadata such as patient age, sex, diagnosis, contrast phase, in-plane spacing, slice thickness, etc. AI models trained on PanTS achieve significantly better performance in pancreatic tumor detection, localization, and segmentation compared to those trained on existing public datasets. Our analysis indicates that these gains are directly attributable to the 16x larger-scale tumor annotations and indirectly supported by the 24 additional surrounding anatomical structures. As the largest and most comprehensive resource of its kind, PanTS offers a new benchmark for developing and evaluating AI models in pancreatic CT analysis.

Problem

• Advancing pancreatic tumor segmentation research with large-scale CT dataset
• Improving AI model performance in pancreatic tumor detection and localization
• Providing comprehensive benchmark for pancreatic CT analysis AI development

Innovation

• Large-scale multi-institutional pancreatic CT dataset
• Expert-validated voxel-wise annotations for 993k structures
• AI models show 16x better tumor detection