okayode/NLST_UPMC

Overview

NLST_UPMC is a research dataset containing CT volumes and derived feature representations for 1,986 subjects. It includes Sybil global ResNet embeddings and Sybil local attention embeddings, along with predefined training and test subject lists.

The dataset is intended to support research involving medical imaging, representation learning, and machine learning workflows using CT-derived data.

Dataset contents

The repository is organized as follows:

NLST_UPMC/
├── README.md
├── NLST-T0-pids-train.csv
├── NLST-T0-pids-test.csv
├── volumes/
│   └── NLST13/
├── global_features/
│   └── NLST13/
└── local_features/
    └── NLST13/

1. CT volumes

Directory: volumes/NLST13/

Contains CT volume files for 1,986 subjects.

2. Global feature embeddings

Directory: global_features/NLST13/

Contains Sybil global ResNet feature embeddings for 1,986 subjects.

3. Local attention embeddings

Directory: local_features/NLST13/

Contains Sybil local attention feature embeddings for 1,986 subjects.

4. Training and test splits

• NLST-T0-pids-train.csv: training subject identifiers, containing 1,690 PIDs.
• NLST-T0-pids-test.csv: test subject identifiers, containing 296 PIDs.

The two subject lists contain 1,986 PIDs in total. Users should verify subject-level correspondence between the split files and the available data before analysis.

Intended use

This dataset is intended for research and methodological development involving CT imaging, machine learning, and derived image representations, subject to the applicable data-use terms and approvals.

Data access and privacy

Although this repository is publicly accessible, users must independently verify the applicable data-use agreements, institutional approvals, and legal and ethical requirements before accessing, downloading, redistributing, or using the data.

Do not attempt to identify individual subjects or link records to external identifying information.

Limitations

The exact file formats, preprocessing details, embedding dimensions, model checkpoint and version, and any applicable restrictions should be documented by the dataset maintainers as these details are verified.

Citation

If you use this dataset in a publication, cite the dataset using the citation information supplied by its maintainers. Add a formal dataset citation or DOI when available.

Contact

For questions about dataset contents, provenance, or permitted use, contact the dataset maintainers through the repository owner.