Consortium for Health Care Informatics Research (CHIR): Translational Use Case Project Natural Language Processing of Chest Radiograph Reports

The Consortium for Healthcare Informatics Research (CHIR) is a multisite project funded by Department of Veterans Affairs Health Services Research and Development (HSR&D). The projects develops methods in natural language processing (NLP) to advance the effective use of unstructured text and other types of electronic health record (EHR) clinical data to improve the health and health care of Veterans. 

This CHIR project focuses on chest radiograph (x-ray) reports for the critically ill patients who are cared for in intensive care units (ICUs). This chest x-ray (CXR) project has developed an NLP tool to extract unstructured data around device and line presence in CXR reports. Currently researchers are further developing methods for counting the number of days that lines/devices have been present in a patient and extending the NLP to CT scans of the chest.  

Radiology reports are unstructured free text documents that describe abnormalities in patients that are visible via imaging modalities such as X-rays. The number of imaging examinations performed in clinical care is enormous, and mining large repositories of radiology reports connected with clinical data-- such as patient outcomes-- could enable epidemiological studies, such as correlating the frequency of infections to the presence or length of time medical devices are present in patients. 

To evaluate the NLP algorithms, the project staff annotated CXR reports to develop a reference set.  The human annotations for the first round of testing were done with Knowtator, a Protégé plug-in.  Annotations for current work are being done with the Extensible Human Oracle Suit of Tools (eHOST), an open source annotation tool.  This project is conducted in collaboration with the CHIR group at VA Salt Lake City Health Care System.