The Cancer Imaging Archive (TCIA) Clinical Trial Data Expansion;

Leidos Biomedical Research, Inc., Frederick MD

Investigators

Abstract

NCI convened the Blue Ribbon Panel (BRP) in 2016 to recommend areas of scientific opportunity for the NCI to pursue aligned with the mission of the Beau Biden Cancer Moonshot initiative. One of their recommendations was the creation of a data science infrastructure to connect repositories, analytical tools, and knowledge bases. This Cancer Data Ecosystem would comprise a dynamic collection of interoperable repositories, analytical services, and interactive portals enabling data to be aggregated, queried, analyzed, and visualized in unique and powerful ways. The Cancer Imaging Archive (TCIA) is one such ecosystem. It manages the collection, de-identification and hosting of large imaging datasets from cancer clinical trials, making the entire archive of cancer medical images (clinical scans and histology) freely accessible for public download. The data are organized as research-focused ?Collections?, typically patients related by a common disease (e.g. lung cancer) or image modality (MRI, CT, etc.). In addition to the images, supporting metadata such as patient outcomes, treatment details, genomics, pathology, and expert analyses are also provided in approximately 2/3 of the collections. These imaging datasets are used to develop computer aided diagnostic tools; to explore how the tumor phenotype is reflective of the genotype; to integrate imaging and clinical data to develop cancer predictive and prognostic models. In the last few years, genomics and proteomics data has exploded along with radiomic and pathomics machine generated output for correlation with the non-depletable data resource of medical and histology images. TCIA was a pioneer in collecting the clinical images associated with the TCGA genomic projects and is currently curating and collecting both histology and clinical images for the Clinical Proteomic Tumor Analysis Consortium (CPTAC) project. This data is from both retrospective and prospective clinical trials. TCIA currently holds 84 collections, with 41,211 patients, 95,450 studies, and 32 million images, of which 80% are completely open to the public. An additional 35 collections are in the processing pipeline, including CPTAC, Division of Cancer Treatment and Diagnosis (DCTD) projects, Cancer Imaging Program Quantitative Imaging Network (QIN) support, and research community donated clinical trials images. TCIA is finishing incorporation of The Cancer Genome Atlas (TCGA) image collections and are rapidly adding CPTAC patient datasets from nine cancer histologies with up to 200 imaging data sets each. DCTD projects include CIP radiologist review of exceptional responder cases, two anti-PD-1 immune checkpoint imaging agent clinical trial collections, PDX animal images, and a pilot project to add images to the National Cancer Trials Network (NCTN) trial semi-public case report data. The QIN activity supports grant mandated data sharing and algorithm challenges that those data enable. The volume for QIN has greatly increased since Eastern Cooperative Oncology Group and American College of Radiology Imaging Network (ECOG-ACRIN) was added as a QIN grantee. At present, almost 2000 imaging datasets from ten completed ECOG-ACRIN clinical trials with included clinical data are in the queue for de-identification and archive curation. The rest of the TCIA acquisition and curation queue includes eleven collections donated by the research community with over 3000 patient images. This task order will support the capture and integration of these, and other, clinical image data collections into TCIA and expand the current underlying infrastructure in anticipation of increased volume and complexity of data from future clinical trials.