Inside this Issue:

Download the Newsletter

Editors:

Joseph Cummings
Charles Dizard

Informatics Today is a publication of the University of Pittsburgh Department of Biomedical Informatics

The University of Pittsburgh is an affirmative action, equal opportunity institution. Published in cooperation with the Department of University Marketing Communications.

UMC65727-1008

The Biomedical Language Understanding Laboratory

The Biomedical Language Understanding (BLU) Laboratory at the University of Pittsburgh focuses on biomedical language processing. The BLU Lab is directed by Wendy W. Chapman, PhD, an assistant professor of biomedical informatics and intelligent systems in the Department of Biomedical Informatics (DBMI).

Natural language processing (NLP) is a branch of artificial intelligence that classifies, extracts, and encodes human natural language.

Chapman studied linguistics and Chinese language as an undergraduate student. After considering graduate studies in Chinese literature, Chapman serendipitously entered into the world of computational linguistics and medical informatics. Her husband, Brian Chapman, PhD, had interviewed with Peter Haug, MD, professor of biomedical informatics at the University of Utah. At the time, Haug was applying computational linguistics techniques to extract and encode information within chest X-ray reports. "I thought the area was fascinating and convinced Peter to give me a try as a graduate student researcher with no background in science, medicine, or computers but with a degree in linguistics and a lot of energy," recalls Chapman. After several years of classes and experimentation with programming, Chapman soon found her niche in this specialized area, earning her doctoral degree in medical informatics in 2000 at the University of Utah. She came to Pittsburgh for a postdoctoral fellowship with Bruce Buchanan, PhD, and afterward stayed with DBMI as a faculty member.

The main focus of the BLU lab is to develop and evaluate computerized applications for understanding the clinical state of a patient from dictated reports. The initial step is to identify relevant concepts such as "shortness of breath" or "headache" in the text. The BLU lab is particularly interested in the next steps: understanding the context in which the concept is described.

Funded research areas of the BLU lab include the following:

Identifying clinical conditions relevant to surveillance of disease outbreaks, funded by the national Library of Medicine: The BLU Lab is developing a system, Topaz, for identifying acute onset of clinical conditions relevant to public health surveillance. Topaz processes emergency department reports to build a detailed understanding of why a patient presented to the emergency room.

Automatically charting dental findings, funded by the National Institute of Dental and Craniofacial research: The BLU lab is extending the MPLUS system that was developed by Haug and colleagues at the University of Utah. This extension, MEDUSA (Medical Understanding and Semantic Analysis), will automatically chart the results of recorded dental examinations.

The following researchers work in the BLU lab with Chapman: Saeed Amizadeh, a doctoral student in the Intelligent Systems Program, is applying machine learning methods to temporal modeling and clinical classification.

Lee Christensen, codeveloper of MPLUS, is the software architect for MEDUSA.

John Dowling, MD, MS, is a retired physician with a degree in biomedical informatics. He provides domain expertise and supervises annotation efforts.

Henk Harkema, PhD, a postdoctoral research associate at DBMI, collaborates on MEDUSA and Topaz, is developing a part-of- speech tagger for clinical text, and is helping to develop an annotation schema for temporality and discourse structure in reports.

Jeannie Irwin, MS, a doctoral student in the Biomedical Informatics Training Program, is working on developing and evaluating the semantic component in MEDUSA, along with evaluating speech recognition performance on dental charts.

Danielle Mowery, a doctoral student in the Biomedical Informatics Training Program, is developing a richer temporal model for the information described in clinical reports.

Heather Piwowar, MS, a doctoral student in the Biomedical Informatics Training Program, is participating in Topaz development and is applying NLP techniques to the biomedical literature to determine whether a research article shared microarray data.

Tyler Thornblade, a computer science graduate student, is working on an algorithm that examines the contextual features of natural language.

The Biomedical NLP Lab is collaborating with other researchers from both inside and outside the University.

Within the University, the Biomedical NLP Lab is collaborating with Rebecca Crowley, MD, MSIS, assistant professor of biomedical informatics, intelligent systems, and pathology, in developing a system that can be used for information extraction and for automated enrichment of ontologies. The BLU lab also is collaborating on Topaz with Jan Wiebe, PhD, professor and director of the Intelligent Systems Program, and Rebecca Hwa, PhD, an assistant professor of computer science.

Outside Pitt, the Biomedical NLP Lab is collaborating on current projects with Haug, MultiModal Technologies, and Guergana Savova, PhD, from the Mayo Clinic. The lab also is collaborating with researchers from the VA hospitals who are developing NLP applications for clinical text (Matt Samore, MD; Brett South, and Adi Gundlapalli, PhD).

For more information, please visit www.dbmi.pitt.edu.