Shawn Murphy, Presenter: Nothing to Disclose

Keith J. Dreyer DO, Abstract Co-Author: Nothing to Disclose

Thomas Schultz, Abstract Co-Author: Nothing to Disclose

David S. Hirschorn MD, Abstract Co-Author: Nothing to Disclose

The Partners Healthcare Research Patient Data Registry (RPDR) is a data repository of 2.1 million Mass General & Brigham and Women's patients accumulated since the 1980's. It stores 530 million diagnoses, medications, procedures, reports and lab values with demographic & visit information. Authorized clinicians query it to select research cohorts and to elicit trends in disease incidence and progression. The system is powerful, but contains only textual data, not images. Therefore, the Clinical Image Repository Access System (CIRAS) was designed to provide access to PACS archives from the RPDR.

CIRAS weaves together more than images; it also incorporates radiology textual data. The reports are processed through a natural language processing algorithm called LEXIMER to automatically pick out findings and recommendations. This data is then added to the database for each report, much in the same way keywords might be automatically extracted for a book index. The DICOM headers of the images are also fed into the database. This enables the researcher to limit her query to specific imaging parameters such as only FLAIR MRI images or only craniocaudad mammograms. The results of feature extraction algorithms are loaded into the database as well, such as CAD services which analyze pixel values to produce textual and sometimes quantitative results.

The CIRAS database ties together the exam orders and reports with DICOM header information, automatically extracted report findings and recommendations and automatically extracted image features from the pixel data.

CIRAS dramatically increases the research potential of the RPDR. Cohorts can now be selected based on imaging characteristics and disease trends can now be found and traced in radiologic images. It is hoped that the usage will progress to the development of more sophisticated feature extraction algorithms to actually use image features as part of the selection criteria. The research PACS archive is an idea whose time has come and will allow medical research to better capitalize on the digital transformation of radiology.

Murphy, S, Dreyer, K, Schultz, T, Hirschorn, D, CIRAS: A Tool for Integrating PACS Archive Images and Metadata into a Clinical Data Repository.  Radiological Society of North America 2005 Scientific Assembly and Annual Meeting, November 27 - December 2, 2005 ,Chicago IL. http://archive.rsna.org/2005/4418542.html