Rebecca S. Crowley, MD, MS
- Intelligent Medical Training Systems
- Student and User Modeling
- Empirical studies of medical decision making and expertise
- Information Extraction and Medical Natural Language Processing for Translational Informatics
- Image-guided decision support systems
- Human-Computer Interaction
- Multi-institution data and tissue sharing
Appointments and Positions
Associate Professor of Biomedical Informatics
Associate Professor of Intelligent Systems
Associate Professor of Pathology
Director of the University of Pittsburgh Biomedical Informatics Training Program
Dr. Rebecca Crowley has been teaching University of Pittsburgh post-graduate classes since 2001 and has served as Director of the Department of Biomedical Informatics (DBMI) Training Program since 2008. Established in 1987, the DBMI Training Program prepares individuals for research and development careers emphasizing the application of modern information technology to healthcare, basic biological and clinical research, and the education of health professionals. As Director, Dr. Crowley oversees more than 28 core faculty, over 50 affiliated faculty from a total of 25 University departments and centers, and approximately 40 students pursuing masters degrees, doctoral degrees, non-degree postdoctoral fellowships and certificates. Funding for the DBMI Training Program is primarily provided by the National Library of Medicine and Dr. Crowley was recently awarded two American Recovery and Reinvestment Act supplements to support new students, faculty and course offerings.
Current Research Projects and Collaborations
caTIES/TIES (Cancer/Text Information Extraction System) makes available highly annotated and de-identified clinical reports for use in biomedical research. The local caTIES deployment at the University of Pittsburgh, TIES, provides researchers access to over 1.7 million surgical pathology reports across all UPMC hospitals over the last 15 years. The system leverages natural language processing algorithms and query visualization methods, and provides secure, easy to use, and highly accurate access to research data and associated tissue. It allows researchers from different institutions to collaborate on studies from within the system, build cross-institutional case sets and order related tissue. The caTIES system uses role based authorization, an integrated honest broker framework, grid-based communication and security frameworks that are based on studies conducted on institutional IRB policies and federal regulatory requirements to ease acceptance and adoption by any institution. The caTIES system is a mature, HIPAA compliant, collaborative research system already deployed at a number of institutions. Future development of a network of systems could facilitate multi-center research projects across the country. The system is a National Cancer Institute and Clinical and Translational Science Awards funded project and more information can be found at ties.upmc.com or caties.cabig.upmc.edu/.
ODIE (Ontology Development and Information Extraction) system is an open-source, extensible toolkit for ontology annotation and enrichment from clinical text. ODIE is designed for use by biomedical informatics researchers to: 1) Annotate document sets with one or more ontologies; 2) Visualize relationships between documents and ontologies; and 3) Suggest additions to existing ontologies. ODIE enables use and extension of National Center for Biomedical Ontology (NCBO) ontologies using natural language processing methods. The system is built on the open-source Unstructured Information Management Architecture (UIMA) framework and includes multiple analysis engines for both ontology annotation and enrichment. ODIE can be used to provide named entity annotations within clinical texts based on any NCBO ontology. ODIE also incorporates a co-reference resolution analysis engine which has been specifically developed for clinical documents. For concept suggestion, the current version of ODIE utilizes both symbolic and statistical methods including lexico-syntactic patterns, similarity and mutual information and users may include their own UIMA engines as additional methods. The ODIE project is funded by the National Cancer Center and more information can be found at bmir-gforge.stanford.edu/gf/project/odie.
SlideTutor is an intelligent tutoring system supporting pathologists at all stages of development—from 1st year residents to community practitioners. The SlideTutor system focuses on visual classification problem solving and monitors mastery of all skills, since novice and experts users employ very different strategies and make different types of errors based on individual levels of knowledge and experience. Using SlideTutor, students: 1) Examine virtual slides using various magnifications; 2) Point to specific areas on the slide; 3) Identify features and their qualities; and, 4) Make hypotheses and diagnoses based on feature sets. Simultaneously, the SlideTutor system compares student actions against an expert module and provides individualized feedback on accuracy. Students can dialog with the tutor to analyze the effectiveness of their learning strategies. A Natural Language Interface (NLI) allows students to enter diagnostic and prognostic reports, echoing common clinical workflows and closely simulating existing practices. Evaluations of the SlideTutor system have shown dramatic improvements in diagnostic reasoning and reporting tasks. The SlideTutor project is supported by the National Library of Medicine and the National Cancer Institute and more information can be found at slidetutor.upmc.edu.
Crowley RS, Legowski E, Medvedeva O, Reitmeyer K, Tseytlin E, Castine M, Jukic D, Mello-Thoms C. Automated detection of heuristics and biases among pathologists in a computer-based system. Adv in Health Sci Educ, Theory and Practice, Epub ahead of print, May 29, 2012. PMID:22618855
Landis Lewis Z, Mello-Thoms C, Gadabu OJ, Gillespie M, Douglas GP, Crowley RS. The Feasibility of Automating Audit and Feedback for ART Guideline Adherence in Malawi. J Am Med Inform Assoc. 2011 Nov-Dec;18(6):868-74. Epub 2011 May 12. PMCID: PMC3197989
Zheng J, Chapman WW, Miller TA, Chen L, Crowley RS, Savova GK. A system for coreference resolution for the clinical narrative. J Am Med Inform Assoc. 2012 Jul 1;19(4):660-7. Epub 2012 Jan 31. PMCID: PMC3384116 [Available on 2013/7/1]
Chapman WW, Savova GK, Zheng J, Castine M, Crowley RS. Anaphoric Reference in Clinical Reports: Characteristics of an Annotated Corpus. J Biomed Inform. 2011 Jul-Aug;18(4):459-65. Epub 2011 Apr 1.
Crowley RS, Castine M, Mitchell KJ, Chavan G, McSherry T, Feldman M. caTIES - A Grid Based System for Coding and Retrieval of Surgical Pathology Reports and Tissue Specimens In Support Of Translational Research. J Am Med Inform Assoc. 2010 May 1;17(3):253-64 PMID: 20442142 [PubMed—in process]
Bartos CE, Butler BS, Crowley RS, Ranked Levels of Influence Model: Selecting Influence Techniques to Minimize IT Resistance. J Biomed Inform. 2010 Apr. [Epub ahead of print].PMID: 20176135
Landis Lewis Z, Douglas G, Monaco V, Crowley RS, Touchscreen Task Efficiency and Learnability in an Electronic Medical Record at the Point-of-Care. 13th World Congress on Medical and Health Informatics Medinfo 2010, Cape Town, South Africa, 12-15th September 2010 (Accepted February 2010, Nominated for Best Student Paper, June 2010). PMID: 20841658
Crowley RS, Gryzbicki D, Legowski E, Wagner L, Castine M, Medvedeva O, Tseytlin E, Jukic D, Raab S. Use of a Medical ITS Improves Reporting Performance among Community Pathologists. Tenth International Conference on Intelligent Tutoring Systems: Bridges to Learning (ITS 2010), Carnegie Mellon University, Pittsburgh PA, June 14-18, 2010.
Lui K, Hogan WR, Crowley RS. Natural Language Processing Methods and Systems for Biomedical Ontology Learning. J Biomed Inform. 2010 Jul 17. [Epub ahead of print] PMID: 20647054 [PubMed—in process]