The Cancer Imaging Archive

The Cancer Imaging Archive
Name	The Cancer Imaging Archive
Established	2011
Location	National Institutes of Health
Type	Medical imaging repository

The Cancer Imaging Archive is a large-scale public repository of medical imaging data created to support research in oncology, radiology, machine learning, and computational pathology. The archive supports investigators from institutions such as National Institutes of Health, National Cancer Institute, Food and Drug Administration, Massachusetts Institute of Technology, and Stanford University by providing curated collections of radiologic studies, pathology images, and corresponding metadata. The resource has been used by investigators affiliated with Harvard University, University of California, Los Angeles, Johns Hopkins University, Dana–Farber Cancer Institute, and Mayo Clinic to accelerate diagnostic algorithm development, multi-institutional trials, and translational research.

Overview

The repository hosts imaging collections spanning modalities such as X-ray, Computed Tomography, Magnetic Resonance Imaging, Positron Emission Tomography, and digital microscopy, enabling cross-disciplinary studies involving teams from American College of Radiology, Radiological Society of North America, European Society for Medical Oncology, International Atomic Energy Agency, and World Health Organization. The platform integrates standardized formats like DICOM and metadata schemas used by projects at Broad Institute, European Bioinformatics Institute, Wellcome Trust Sanger Institute, Carnegie Mellon University, and University of Oxford to support reproducible analyses. Users from Google Research, Microsoft Research, IBM Research, NVIDIA, and DeepMind have accessed the archive for training machine learning models and benchmarking performance. The archive interfaces with registries and consortia including Cancer Imaging Program (NCI), ClinicalTrials.gov, Cancer Genome Atlas, International Cancer Genome Consortium, and Precision Medicine Initiative to align imaging phenotypes with molecular datasets.

History and Development

The archive was initiated to address needs identified in workshops convened by National Cancer Institute, National Biomedical Imaging Archive, Office of Cancer Clinical Proteomics Research, American Association for Cancer Research, and Society of Nuclear Medicine and Molecular Imaging, with early pilots involving collaborators from University of Pennsylvania, Memorial Sloan Kettering Cancer Center, University of Michigan, University of Toronto, and Yale University. Development phases incorporated software and data-management practices from projects at OpenStreetMap, GitHub, Apache Software Foundation, Kaggle, and PhysioNet to enable community contribution, versioning, and challenge hosting. Major milestones included release of curated datasets timed with scientific competitions run by IEEE, MICCAI, ISBI, RSNA, and NIH to evaluate segmentation, detection, and radiomics algorithms. Governance and technical evolution involved partnerships with Amazon Web Services, Google Cloud Platform, Microsoft Azure, Hadoop, and Docker to scale storage and compute.

Collections and Data Types

Collections include annotated series for tumor types such as breast cancer (partners Susan G. Komen, Beth Israel Deaconess Medical Center), lung cancer (partners Memorial Sloan Kettering Cancer Center, University of California, San Francisco), brain tumors (partners Massachusetts General Hospital, Brigham and Women's Hospital), prostate cancer (MD Anderson Cancer Center), and head and neck cancers (University College London Hospitals). Data types span volumetric CT scans used in trials with Radiation Therapy Oncology Group, dynamic contrast-enhanced MRI linked to studies at Mayo Clinic, FDG-PET studies contributed by Stanford University School of Medicine, and whole-slide images from pathology labs affiliated with Mount Sinai Health System and Cleveland Clinic. Annotations include lesion segmentations, RECIST measurements, pathology labels harmonized with SNOMED CT, structured clinical metadata linked to ICD-10, and outcomes curated in coordination with registries such as SEER and National Surgical Quality Improvement Program.

Access, Curation, and Data Standards

Data submission and access workflows follow standards and protocols adopted by Digital Imaging and Communications in Medicine (DICOM), Health Level Seven International (HL7), Fast Healthcare Interoperability Resources (FHIR), Common Data Elements (CDE), and metadata frameworks used by National Library of Medicine, Clinical Data Interchange Standards Consortium (CDISC), International Organization for Standardization (ISO), and College of American Pathologists. Curation teams include specialists from American Society of Clinical Oncology, European Society of Radiology, Pathology Informatics Society, Biomedical Informatics Research Network, and academic cores at Cornell University to de-identify Protected Health Information in compliance with policies influenced by Health Insurance Portability and Accountability Act and ethical guidance from Belmont Report principles. Access models support open-access collections, controlled-access datasets requiring data use agreements with institutions such as University of Washington, University of Chicago, and Emory University, and challenge-specific embargoes used in competitions hosted with ISIC, Kaggle, and MICCAI.

Research Applications and Impact

The archive has catalyzed research in radiomics and imaging biomarkers used in studies by European Organisation for Research and Treatment of Cancer, National Comprehensive Cancer Network, American Society for Radiation Oncology, Translational Cancer Research, and biotechnology companies like Genentech, Roche, AstraZeneca, Pfizer, and Novartis. Publications leveraging the archive have appeared in journals such as Nature Medicine, The Lancet Oncology, Radiology, Journal of Clinical Oncology, and IEEE Transactions on Medical Imaging and have informed clinical guidelines developed by NCCN and policy discussions at FDA. The resource has enabled algorithmic advances demonstrated in competitions organized by RSNA, ISBI, MICCAI, Kaggle, and NIH that improved tumor segmentation, outcome prediction, and treatment planning.

Governance, Funding, and Collaborations

Governance involves stakeholders from National Cancer Institute, National Institutes of Health, National Institute of Biomedical Imaging and Bioengineering, Office of the Director (NIH), and advisory input from academic partners such as Harvard Medical School, Johns Hopkins Bloomberg School of Public Health, Imperial College London, ETH Zurich, and University of Melbourne. Funding sources have included grants from National Institutes of Health, cooperative agreements with National Cancer Institute, philanthropic support from foundations like Lasker Foundation and industry collaborations with Amazon, Google, Microsoft, and IBM. Collaborations span consortia including Cancer Genome Atlas, International Cancer Imaging Society, Quantitative Imaging Network, Global Alliance for Genomics and Health, and regional networks like European Cancer Organisation and Asia-Pacific Advanced Network to foster interoperable, multi-center research.

Category:Medical imaging repositories