Using the caBIG Workspace Tools to Re-engineer the
Radiology Structured Report for the Era of Personalized Medicine
 
Authors:
Eliot L. Siegel, MD, FSIIM, University of Maryland School of Medicine, VA Maryland Health Care System; Paul Mulhern; Adam E. Flanders, MD; Carl Jaffe, MD
 
Background:
The Cancer Genome Atlas (TCGA) project has been undertaken to enhance our ability to diagnose, treat, and prevent cancer in a “personalized” way by conducting large scale cataloging of genetic mutations linked to cancer. It is a joint project between the Human Genome Project and the National Cancer Institute. The TCGA Data Portal, which currently includes data from subjects with a form of brain neoplasm, glioblastoma multiforme, provides a platform to mine, retrieve, and perform analysis on the available genetic, laboratory, and clinical data. The TCGA project is embarking on the collection of a large amount of radiology images and related data. There is strong consensus that phenotypic imaging data is likely to add substantially to meaningful discovery in the research setting and for the diagnosis, staging, and prognosis in the clinical setting when cross correlated with laboratory, clinical and proteomic and clinical data.

The first step in realizing the potential of the TCGA radiology data is to enable radiologists to interpret the images and to create markup and annotation data in a standardized and organized way. The development associated with this use case will yield the components necessary for this transaction to occur.

Specifically, this includes the addition of software to three existing radiology workstations: one created from XIP (extensible imaging platform) and AVT (algorithm validation tool) components; another, the freely available Osirix workstation for the Apple Macintosh community; and the third is the Clear Canvas diagnostic imaging workstation, which is free for Windows users. This software for the three workstations allows each workstation to retrieve images from the NCI’s National Biomedical Imaging Archive (NBIA) and to annotate and mark-up the images, then save them using the AIM (annotation and image mark-up) schema funded by the caBIG imaging workspace to an AIM grid data service. An initial prototype template has been created which can be downloaded to a workstation containing the information required for a highly structured report on a disease such as glioblastoma (as in this first use case), or cervical or breast cancer. This can similarly be extended to other neoplasms or other disease processes. These annotations and mark-ups will then be available for retrieval and display by other users, regardless of which workstation they use (one of the three or any other workstation that supports the AIM schema). Users will then be able to utilize the NCI’s caB2B (cancer bench-to-bedside) query tool that permits translational research scientists to search and combine data from multiple caGrid data services. This will permit cross correlation between clinical data (stored on a clinical data service on the grid) and a genomic/proteomic database (also stored on the grid) allowing query and responses to questions that involve the relationship of the phenotypic data from the MRI studies with patient survival and clinical and laboratory data.
 
Discussion:
This effort represents an example of the new care/research paradigm in which phenotypic imaging information can be included in fully-automated and semi-automated research and clinical data searches that will result in new insights into disease and new means of stratifying diseases into subgroups to enhance homogeneity of patient groups to increase the likelihood of treatment response. Research can then be more directly applied to a specific patient’s disease process. It also demonstrates the combined use of the imaging informatics infrastructure projects developed by the NCI’s caBIG imaging workspace. In addition to research, we believe that this has tremendous potential as a tool to facilitate decision support, both for radiologists and for oncologists and surgeons involved in patient care. For example, a patient with a glioblastoma could have his imaging findings, using the structured template developed for the TCGA project, combined with his laboratory and clinical data, as well as the genomic analysis of his tumor from a biopsy in order to predict optimal treatment modalities and information about relative quality of life and survival.
 
Conclusion:
The caBIG imaging workspace has utilized its imaging informatics initiatives to address the challenge to create a highly structured template that saves information in a searchable, standardized and machine readable format which can be used for clinical trials and other research and well as for clinical decision support. The ability to save highly structured information in a machine readable format for automated analysis and correlation with other data seems to be a major pre-requisite for the next generation of diagnostic imaging tools in support of the coming era of personalized medicine.