imaging-data-commons

NCI Imaging Data Commons - Cancer Imaging Query and Download Tool

Overview of Skills

Use the idc-index Python package to query and download publicly available cancer imaging datasets from NCI Imaging Data Commons. Large-scale CT, MR, PET, and pathology slide datasets can be accessed without authentication. It supports AI medical imaging training and cancer research.

Use Cases

1. Building AI Medical Imaging Training Datasets

Obtain high-quality, richly annotated cancer imaging data for deep learning models. Use SQL queries to precisely filter datasets by specific cancer types, imaging modalities, and anatomical sites. Supports bulk downloading of DICOM image files. Suitable for preparing training data for tasks such as medical image classification, detection, and segmentation.

2. Medical Imaging Research and Clinical Analysis

Researchers and clinicians can access large-scale, multi-center cancer imaging datasets for imaging feature analysis, disease pattern studies, and clinical validation. IDC includes extensive clinical metadata and patient demographic information, enabling retrospective research and cross-institution data comparison.

3. Exploring and Visualizing Cancer Imaging Data

Preview and visualize DICOM imaging data directly in the browser without downloading. Supports both single-series viewing and multi-series viewing for an entire examination. Ideal for quickly exploring data quality, selecting appropriate datasets, and generating presentation materials for research.

Key Features

1. SQL Metadata Query and Data Discovery

Use standard SQL queries to search the IDC metadata index, supporting hierarchical filtering by collection, patient, study, series, and more. You can query metadata fields such as imaging modality (CT, MR, PET, SM), anatomic site, manufacturer, data license, and file size. Provides collections_index and analysis_results_index to discover source imaging and derived datasets.

2. Bulk DICOM Data Download

Supports flexible downloads using identifiers such as collection_id, PatientID, StudyInstanceUID, and SeriesInstanceUID. Allows customization of directory structure templates, making it suitable for organizing large datasets. Downloads directly from public storage buckets on AWS S3 and Google Cloud Storage without needing a cloud service account. Offers a command-line tool, idc download, for batch downloading via a manifest file.

3. Data License Checking and Citation Generation

Automatically queries license information (CC BY, CC BY-NC, etc.) to ensure compliance with commercial or academic use requirements. Automatically generates citations in formats such as APA and BibTeX based on source_DOI, meeting data attribution and publication needs. Supports license verification before downloading to avoid compliance risks.

Frequently Asked Questions

How do I get cancer imaging data from NCI Imaging Data Commons?

After installing the idc-index package, use IDCClient() to run SQL queries to locate the target data, then download it via the download_from_selection() method. You can select data by collection, patient, study, or series UID, supporting bulk downloads and custom directory structures. No account registration or authentication is required, but you must comply with the relevant data licenses.

What authentication and permissions does idc-index require?

idc-index itself requires no authentication or permissions. All IDC data is publicly accessible. Simply install it with pip install --upgrade idc-index. Data is stored in publicly accessible AWS S3 and Google Cloud Storage buckets and supports anonymous access. Note that some collections have restrictions on commercial use (e.g., CC BY-NC licenses); please check the license terms before using.

What imaging modalities and formats are supported by IDC data?

IDC data version v23 includes multiple medical imaging modalities: CT (computed tomography), MR (magnetic resonance imaging), PT (positron emission tomography), XA (angiography), RF (radiofrequency), and SM (pathology slide microscopy). Data is stored in standard DICOM format and can be processed with tools such as pydicom, SimpleITK, and 3D Slicer. Pathology slide data supports whole-slide image viewing.