View Abstract

A7304 - Integrating Radiological Imaging Informatics into a Multi-Site Investigation of Lung Cancer Overdiagnosis
Author Block: S. Antic1, H. Mahany2, T. Osterman3, F. Maldonado1, V. S. Nair4, M. B. Schabath5, R. Gillies5, P. P. Massion1, B. Landman6; 1Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University Medical Center, Nashville, TN, United States, 2VU Institute for Imaging Science, Vanderbilt University Medical Center, Nashville, TN, United States, 3Biomedical Informatics, Vanderbilt University Medical Center, Nashville, TN, United States, 4Medicine, Stanford University, Stanford, CA, United States, 5Moffitt Cancer Center, Tampa, FL, United States, 6Center for Computational Imaging, Vanderbilt University, Nashville, TN, United States.
Rationale: We are establishing a molecular and cellular characterization laboratory (MCL) for benign and malignant lung nodules detected early in their natural history at four collaborative institutions. Given the scale of this project, it is essential to efficiently capture subject demographics and assign consistent identifiers across clinical, bio-specimens, and imaging data. We present a multi-site informatics system integrating multi-modal data and providing workflows consistent with a large clinical trial.
Methods: In the Lung MCL, REDCap serves as the central biorepository without aggregating protected health information across sites. Each site coordinator generates unique identifiers using the id generator, a custom javascript program which runs in a secure web browser without exchanging any information with the coordinating site. Site’s biorepository enters descriptive information regarding samples into the Lung MCL coordinating biorepository REDCap. Physical samples are shipped via couriers. Imaging data is transmitted to the coordinating center via DICOM C-MOVE (to a DICOM receive node), Box transfer, or disk via courier. Once received, the imaging data are parsed by an instance of the RSNA Clinical Trial Processor (CTP) server which performs the following functions: (1) coded identifiers in the image headers are matched the identifiers in Lung MCL REDCap, (2) duplicates are detected and removed, (3) images are decompressed to enhance system compatibility, (4) image meta-data is coded into project/subject/session hierarchy, and (5) images are transmitted to an eXtensible Neuroimaging Repository (XNAT) server. To perform step 1, the CTP server reads a coding table from the Lung MCL REDCap because some sites programmatically generate both imaging and non-imaging ID’s for a single subject. If received images do not match the Lung MCL REDCap, they are held in “quarantine” for future quality assurance. Imaging data can be directly downloaded or transmitted to secondary analysis platforms through the user-interface, or subsets of the data can be pushed in bulk via a backend server application program.
Results: To date, 99 subjects have completed the process, while data from 577 subjects from 5 institutions (325,812 images) are in the process of quality assurance.
Conclusions: Capturing clinical trial data that involves imaging is a non-trivial task from both the perspective of subject privacy (e.g., controlling flow protected health information) and technical systems integration. REDCap, CTP, and XNAT provide the MCL consortium a solid framework on which to perform analyses within and across data aggregation modalities. Sites can maintain control over sensitive data while fully participating in a consortium.