Related Experiment Video
Updated: May 20, 2026
13:01
Industrialized, Artificial Intelligence-guided Laser Microdissection for Microscaled Proteomic Analysis of the Tumor Microenvironment
Published on: June 3, 2022
A longitudinal whole-body CT dataset with manually annotated tumor lesions
Sergios Gatidis1, Felix Peisen2, Andreas Wagner2
1Department of Radiology, Stanford University, Stanford, CA, USA. sgatidis@stanford.edu.
Scientific Data
|May 18, 2026
Summary
Longitudinal-CT is a new dataset of whole-body computed tomography (CT) scans for 300 metastatic melanoma patients. It enables AI development for automated tumor lesion detection and tracking over time.
Area of Science:
- Oncology
- Radiology
- Medical Imaging
Background:
- Accurate tracking of tumor lesion evolution is crucial for evaluating treatment response in metastatic cancers.
- Existing datasets often lack longitudinal data and detailed annotations necessary for advanced AI development.
Purpose of the Study:
- To introduce Longitudinal-CT, a novel, publicly available dataset for whole-body CT studies.
- To provide a standardized resource for developing and validating AI methods for automated oncologic lesion analysis.
Main Methods:
- The dataset includes 600 whole-body CT studies from 300 metastatic melanoma patients, with baseline and follow-up scans.
- Expert manual annotations detail 7,182 tumor lesions, including volume, location, and longitudinal changes.
- Data is provided in anonymized NIfTI format with segmentation masks and metadata.
Main Results:
- The dataset contains 4,079 lesions at baseline and 3,103 at follow-up, capturing lesion persistence, regression, merging, and new appearances.
- A baseline deep learning model (nnU-Net v2) demonstrated the dataset's utility for automated segmentation.
Conclusions:
- Longitudinal-CT offers a robust foundation for advancing AI in oncologic imaging.
- This resource will facilitate the development of AI tools for automated lesion detection, segmentation, and temporal tracking in cancer patients.