ProjectDescriptions Comment | Edit | Print | Subscribe to this page

(Completed draft, moved to http://path.upmc.edu/cpi/cpi-res.html)

This material from path.upmc and the BOIC site provides a starting point for developing a list and descriptions of current projects.

Telepathology

Personnel

  • Yukako Yagi - Director
  • Ishtiaque Ahmed - Pathology Systems Specialist
  • Anurag Sharma - Systems Analyst
  • Jon Duboy - PC Consultant I
  • Othman Abdul-Karim - Programmer IV

Traditionally, TelePathology has been implemented using custom software and high-speed private data networks. Internet and the World Wide Web offer a flexible, ubiquitous, user-friendly, and intrinsically multimedia environment ideally suited for TelePathology applications. A basic TelePathology system includes microscope with digital camera, computer, telecommunications link between sending and receiving sites, and a workstation at the receiving site with a high-quality monitor to view the images. There also might be hardware to allow the receiving pathologist to control the microscope from a distance and view the entire slide in real-time. The images are viewed on a computer screen, rather than through microscope.

At the UPMC, we are actively investigating seven aspects of TelePathology, Telemedicine and Pathology Imaging:

  • TV Conference
    Ten TV Conference Systems have been installed at UPMC in PA. Two different type of systems (Pathology Station and View Only Station) were prepared for different purposes. We chose hardware and software very carefully for each location.
  • Web Cast Conference
    We provide live/archived three Pathology Weekly Conference on the web. Currently, the following 5 conferences are available as live and archived video.
  • Web Based Telepathology System
    Follow the link above for details.
  • Internet Based Transplant Pathology Consortium
    Follow the link above for details.
  • Robotic Microscopy
    With two robotic microscopes, UPMC's Department of Pathology has a number of ongoing projects evaluating this technology, as well as agreement with several companies to participate in the development of their products. This project is appropriate for residents and students. Our Robotic Microscopy System can support Nikon E1000 and Olympus AX80.
  • Virtual Telepathology (Virtual Microscopy) Faculty at the Center for Pathology Informatics and Pittsburgh Supercomputing Center have developed an robot to image entire slides automatically, rapidly and at high resolution. Built by an outside company, the system will undergo trials at UPMC this year. The system includes an imaging robot, slide handling robot, LIS integration and secure web viewer (virtual microscope)
  • Mobile Telepathology
    Mobile TelePathology by Pocket PC. Mobile TelePathology using VAIO GT3. Follow the link above for details.
  • Imaging Research
    Wireless WAN/LAN. IP Telephony. Multi-Spectral Imaging. Network Camera. Automated Scoring System. Dermatology. Emergency Medicine. Follow the link above for details.

Active research support

  • Integrated Medical Information Technology (US Air Force Telemedicine)
  • Natural Vision Project, Japanese Ministry of Public Management, Home Affairs, Posts and Telecommunications
  • HU-UPCI Cancer Education/Career Development Partnership (NCI P20 CA101592)

Improving Patient Safety by Examining Pathology Errors

Personnel

  • Stephen Raab, MD - Principle Investigator
  • Dana Grzybicki, MD, PhD? - Co-investigator
  • Michael J. Becich, MD., PhD? - Co-investigator

A critical component of improving patient safety is reducing medical errors. A paucity of information exists on anatomic pathology diagnostic errors and their effect on patient outcomes. Many previously published studies are limited to single institutions and report variable diagnostic error percentages from less than 1 percent to 43percent of all patients who undergo a biopsy or excisional procedure, with no correlation between error andoutcome. A major goal of this project is to establish a Web-based, pathologist-driven, national, voluntary anatomic pathology error database. These data will be used for continuous quality improvement targeted at error reduction and clinical outcomes improvement. This project will provide valuable information regarding diagnostic pathology errors; it will set the groundwork for future studies focused on the examination of other types of diagnostic pathology error and the effect of error reduction programs in pathology practice.

Active research support

  • Agency for Healthcare Research and Quality: PA 00-111, 12/01/02-11/30/07, total direct costs $3,251,217 (Pitt: $777,136).

Improving Patient Safety by Redesigning Pathology Practice

Personnel

  • Stephen Raab, MD - Principle Investigator
  • Dana Grzybicki, MD, PhD? - Co-investigator

The major goal of this project is to investigate and demonstrate the usefulness of the application of a Toyota Production System (TPS) model for continuous quality improvement to anatomic pathology practice. This project is being performed in collaboration with staff from the Pittsburgh Regional Healthcare Initiative (PRHI). It is focused on immediate problem solving based on laboratory problems and errors detected in real time on an on-going basis. A TPS Learning Line will be created and established in the anatomic pathoogy laboratories at Shadyside Hospital that will target pathology problems and errors in real time in order to improve the quality of anatomic pathology services and diagnostic information.

Active Research Support

  • Jewish Healthcare Foundation: 7/1/03-12/31/04, $140,500

Molecular Reclassification of Prostate Cancer

Personnel

  • Michael Becich, MD, PhD? - Tissue bank and informatics lead

The aims of this project include extending the Western Pennsylvania Prostate Tissue bank to provide a national resource for research, and to analyze gene expression patterns in cancer of the prostate in comparison with non-tumor and BPH tissue to establish correlations with distinct subsets of cancer behavior. The studies will allow us to establish a molecular reclassification of prostate cancer based on coordinated expression of groups of specific genes. Complete prostatectomy specimens available from the Tissue Bank will be processed by microdissection and used to extract RNA. This will in turn be processed for analysis through the Affymetrix gene chip set. Our tissue bank contains complete and well stratified information that will be used by the bioinformatics teams of HLR and Pitt to provide correlation between coordinated expression of specific gene sets and distinct tumor behavior.

Active research support

  • National Cancer Institute: U01 CA86735-01 A Shared Resource for the Molecular Classification of Prostate Cancer; Comprehensive Prostate Cancer Tissue Resource (CPCTR) Consortium National Cancer Institute, 04/01/00-03/31/05, total direct costs $2,992,038.
  • National Cancer Institute: U01 CA88110-02 Molecular ReClassification? of Prostate Cancer; Director's Challenge for the Molecular Classification of Cancer Consortium, 08/29/00-01/31/05, total direct costs $2,975,486.

Temporal Pattern Recognition and Communication of Clinical Laboratory Data

Personnel

  • James Harrison, MD, PhD?, principle investigator
  • Valerie Monaco, PhD?, MS, co-investigator
  • Andrew Post, MD, graduate student

Clinical laboratory data displays organized by physiologic system or concept hierarchies may improve the speed and/or accuracy of clinical decisions as compared with displays organized by laboratory section. Non-medical cognitive analysis studies indicate that graphical representation of data is associated with improved decision making in comparison with tabular numerical data, particularly in time-pressured situations. We hypothesize that identification and explicit graphical representation of important temporal relationships in clinical laboratory data will improve the ability of clinicians to reach decisions rapidly and correctly. We have developed a framework for identifying statistical patterns in timeseries data and temporal relationships between those patterns. We propose to 1) incorporate our framework into a problem-oriented display that prioritizes and visualizes clinical laboratory data adaptively based on types of temporal patterns in the data, 2) conduct patient case simulations in which clinician subjects access clinical laboratory data using traditional, problem-oriented, and temporal data-driven displays, and 3) evaluate the efficiency and accuracy of decision-making by clinicians under these simulated case conditions with mild time pressure, using “think aloud” techniques in which clinicians are video- and audiotaped during decision-making. The influence of these three user interfaces on decision-making will be evaluated by the accuracy of orders in comparison with an expert clinical gold standard, the relative proportions of information acquisition vs. information evaluation cognitive steps during case analysis, the total cognitive steps (cognitive load) required for case disposition, and the total number of screens and data elements viewed. We anticipate that the problem oriented display will support improved clinical decision-making through increasing the efficiency of information gathering; the temporal data-driven display may enhance decision-making further by focusing attention on, and directly visualizing, important relationships that would ordinarily have to be inferred.

Active research support

  • National Library of Medicine R01 LM008192, 10/1/03 - 9/30/06, total direct costs $300,000.

SlideTutor - An Intelligent Tutoring System in Pathology (http://slidetutor.upmc.edu)

Personnel

  • Rebecca Crowley, MD, MS - Principal Investigator
  • Drazen Jukic, MD, PhD? - Dermatopathology Domain Expert
  • Olga Medvedeva, MS, MS - Lead Developer
  • Elizabeth Legowski - Research Associate
  • Ellen Roh, MD - Postgraduate Fellow
  • Eugene Tseytlin, MS - Developer

We are adapting the well-described paradigm of the Model Tracing Intelligent Tutoring System (MT-ITS) to develop multimedia, knowledge-based pathology training systems in several domains. ITS are adaptive, instructional systems that seek to emulate the well known benefits of one-on-one tutoring when compared to other instructional methods. Very few ITS have been developed in medical domains. Our system is designed to provide individualized coaching to students as they search, and interpret virtual pathology slides. The model-tracing aspect of the system guides the student to correctly search a slide, identify relevant evidence, and formulate and test hypotheses before making a diagnosis. As the student works through each case, the system’s model-tracing mechanisms student attention, offer hints based on the student’s progress through the problem, and correct errors in the intermediate steps leading to a diagnosis. Across multiple problems or cases, the system’s knowledge-tracing mechanisms update the system’s assessment of the current state of the student’s skills. Pedagogic decisions about tutorial strategy, case selection or curriculum sequencing can then be based on the current state of the student model. The SlideTutor? project includes both development and evaluation efforts, and therefore requires a cross-disciplinary team including individuals with expertise in artificial intelligence, education, cognitive science, evaluation, and Pathology. SlideTutor? is funded by grants from the National Library of Medicine and National Cancer Institute.

Active research support

  • National Library of Medicine 1R01 LM007891-01, 7/1/03-6/30/07 $269,725 per year (Direct and Indirect).
  • National Cancer Institute 1 R25 CA101959-01, 8/24/03-8/23/06 $232,489 per year (Direct and Indirect).

The Shared Pathology Informatics Network (http://spin.nci.nih.gov/)

Personnel

  • Michael Becich (Site PI)
  • Rebecca Crowley, MD, MS
  • Kevin Mitchell, MS - Lead Developer
  • John Gilbertson, MD
  • James Harrison, MD, PhD?
  • Dilip Gupta, MD

The Shared Pathology Informatics Network (SPIN) is a National Cancer Institute (NCI) sponsored cooperative agreement among four institutions (Harvard University, University of Indiana, UCLA, and Uni-versity of Pittsburgh) to develop a model web-based system for accessing pathology data on archived human tissue speci-mens, across multiple institutions and databases. An important and difficult aspect of this work is the extraction of information (such as the diagnosis, findings, and relationship of tissue blocks to the specimen) from the free-text of Surgical Pathology reports. Information Extraction from pathology reports is complex. For example: (1) reports contain multiple sections (such as Final Diagnosis, Gross Description, Comment, etc) that vary in narrative structure and uniformity (2) there is institutional variation in reporting practices (such as differences in the keywords that delimit important sections of the report), and (3) reports contain negative as well as positive findings and diagnoses. We are working on the development of a pipeline-based system for machine annotation of surgical pathology reports. The system is built with GATE - an open source architecture for Language Engineering, available through the University of Sheffield (http://gate.ac.uk/). De-identified reports and annotations are available for retrieval by all SPIN institutions through a peer-to-peer network available developed by other consortium members. Current development efforts include (1) refinement of our negation tagger, (2) development of general methods for attribute-value extraction, (3) rule based methods for identifying common data elements to be extracted and (4) development of a human annotated corpus for use in performance testing and eventually system training.

Active research support

  • National Cancer Institute: UO1 CA91338-01, 08/13/01-07/31/06, total direct costs $1,376,310.
This page was last edited 4 years ago by harrison. View page history | Edit this page
Subject:


Comment:


    with signature

Powered by Zwiki, Zope, Python, and Mac OSX