Tissue Banking Information System

Personnel

The Center for Pathology Informatics is in the process of designing, building and implementing a tissue banking information system to support the growing demand for highly qualified tissue samples by researchers and industry. The system is being developed in two phases. The first is acquiring - from existing systems - clinical (text), morphologic (image) and genomic (microarray) data on tissue stored in the tissue bank. The second is developing tools to allow researchers to query and visualize this very large data set for tissue annotated with specific, user-defined parameters (e.g., prostate tissue with adenocarcinoma showing microinvasion and high STK2 activity). The development of this system will be involve hypothesis-driven empirical research studies to determine that most efficient mechanisms to store, represent and query this very large multimedia data set,. The result should be a commercial product serving the medical, pharmaceutical, bioengineering and genomics industries. This multi-year project is appropriate for residents and graduate students interested in database systems, imaging, or genomic informatics.

Active research support

• Association for the Cure of Cancer of the Prostate (CaP? CURE)
• Comprehensive Prostate Cancer Tissue Resource
• National Medical Technology Test Bed

Ubiquitous, Whole Slide Imaging (Virtual Microscope, Virtual Slides)

Personnel

The goal of this project is to determine the operational requirements, limitations, and, eventually, the diagnostic accuracy of a robotic imaging system desgned to image tissue on all slides generated at UPMC. The project runs parallel to beta tests of the IST imaging product. The image resolution will be high (approximately 0.33 linear µm per pixel) as will file size, on the order of 2 GB/slide uncompressed. Research issues include the reliability of robotic systems, high speed compression and display algorithms, the maintenance and clinical utility of very large image databases, GUI development, bar coding in the anatomic laboratory, and studies of the diagnostic utility of whole slide images vs. glass slides. This project is appropriate for a resident or masters student seeking an immersion in next-generation clinical imaging systems.

Machine Vision in Histology

Personnel

• Art Wetzel, PhD?
• John Gilbertson, MD

The machine vision group is a collaboration between the Center for Pathology Informatics and the Pittsburgh Supercomputing Center. The group is doing active research in four main areas.

Content-based Image Interpretation: This ongoing project is developing software to effectively interpret digital images. Successes include an effective tissue sorter and a Gleason Grader. We expect this technology to become much more important in the near future, as images become more ubiquitous in pathology. This is a high-end computational project appropriate for graduate students and some dedicated residents.

High-end Imaging: This group examines the technical capabilities of lenses, CCDs?, cameras, capture cards and other components to determine ideal pathology imaging system for both human and machine viewing.

Image Libraries: Dedicated to developing systems that can store, index and appropriately display images from large (>100,000) image databases. Work includes mechanisms for text-based indexing as well as automatic indexing of images on the basis of the image content itself.

Automated Diagnosis: Closely associated with the Image Libraries project, the automated diagnosis project is developing working systems and algorithms to assist pathologists in making morphologic diagnoses. Working systems are currently dedicated to prostate diagnosis and Gleason grading.

Text-based Image Indexing

Personnel

This project is looking at which textual annotations best index an image for future queries. Though some annotations may seem obvious (e.g., UMLS or SNOMED terms for diagnosis), others are far less straightforward (e.g., morphology, secondary findings). Though directed at imaging indexing, the project also looks at the uses of images in pathology.

Information Therapy

Personnel

The dramatic growth of health care information available on the Internet often leaves patients with an overwhelming amount of information they need their health care provider to explain. Information Therapy is an "information prescription" that transforms pathology reports into a tool urologists and oncologists can use to educate patients with prostate cancer about their disease.

HUBS Collaborative Telemicroscopy

Personnel

UPMC, the Pittsburgh Supercomputing Center, Johns Hopkins University (JHU), the University of Pennsylvania Health System (UPHS) and the University of Medicine and Dentistry of New Jersey (UMDNJ) are collaborating in the development of a regional telemicroscopy consortium based next-generation Internet technology. The team will eventually integrate a digital image dataset at JHU, a robotic microscope at UMDNJ and content-based image retrieval systems at UMDNJ and UPMC. UPHS will serve as the objective, independent test site for these applications.

Active Research Support

ProstateHealth.org

Personnel

This ongoing project is a website for patients and professionals, focusing on diseases of the prostate gland. The site features case studies of prostate pathology from the UPMC Department of Pathology's genitourinary case database, tailored to both professional and lay audiences.

Clinical Trials Information Management System

Personnel

The Clinical Trials Information Management System is an effort to develop an integrated system for managing various aspects of a clinical trial. This project is a collaboration between the Benedum Oncology Informatics Center and the Clinical Research Services (CRS) Office at the University of Pittsburgh Cancer Institute.

The initial development effort, and implementation, has focused on providing administrative functions to support the following:

• Internet-based submission for protocol review and approval
• Secured web-based access to protocol data
• Automated monitoring of clinical trials events
• Establishment of an efficient and timely communications channel between principal investigator and CRS via event-based e-mail linked directly to the protocol submission.

The overall goal of the project is to create an integrated tool to support all aspects of the clinical trial:

• Protocol development/submission
• Administration
• Patient Management
• Data Collections and Acquisitions.

System Architecture: The current version of the Clinical Trials Information Management System is a Java-based applet running over a secured SSL-based intranet connection, from any web-based browser. All data is stored on a Sun Sparc system running Oracle 8i. All protocol submissions will be done via a Cold Fusion HTML-based form.

Melanoma Information System

Personnel

Development of a shared information system for the optimal clinical management and prospective and retrospective analysis of clinical, laboratory, research of melanoma patients care directed in the Melanoma Program.

A system that will:

• complement clinical operations, daily and long term;
• support the unique needs of the Melanoma Center;
• automate data collection using PDAs?;
• enable collaboration between clinical laboratory from: Dermatology, Surgery, Medicine/Oncology, Pathology and Basic Research

assure access to high quality data; and provide support for patient clinical, pathological film and digital imaging follow-up.

General Features

• Web-based user interface (client - server)
• Industry standard database (Oracle)
• PDAs? - Palm III programmed with Satellite Forms
• Security - Firewall and SSL encryption
• Data dictionary
• Screens (data entry and viewing forms)

Architecture

MCDB is constructed as a three-tier system. The client tier presents clinical research data pages using HTML and Java delivered over a secure connection and accessed via Netscape or Internet Explorer browser. Clinical research data is stored in a relational data model contained in the database tier.

Infrastructure Security

• Protecting MCDB data from unauthorized access and external threats as well as internal unauthorized use is of prime importance.
• Authentication to verify the identity of a system user
• Access control that defines users in groups which limit or permit resources
• Information exchanged between the client and server are encrypted (SSL)
• Data integrity to ensure users cannot damage date
• Firewall the application server; database server and image server.
• Backup and Recovery of data to protect from accidental or deliberate destruction.

Technology/Infrastructure/Architecture

The Melanoma Center database being developed for the UPCI consists of three major components. This style of system is known as a three-tier architecture. The three tiers are data, application and presentation. The first component or tier is the Oracle Database Server will contain the relational data, content and the Web metadata. The second tier is the application server that will handle requests from network users that wish to enter and view stored data. The last or third tier is the client software, consisting of Java applets and dynamically generated HTML pages that represent the finished application. The client side is built using JBuilder? that generates Java code and runs in a standard Web browser and serves the application to the user. The Image server hardware platform is a Dell PowerEdge? 2200 Server. Running under Windows NT with dual 300MHz Pentium processor, 256MB of ram with a 30GB level 5 RAID disk controller.