[1645] PathRez – A Concept Based Relational Database Model for Archival and Retrieval of Static Images and Virtual Slides

Somak Roy, Matthew Smith, Frank A Fusca, Gary M Burdelski, Denine Maglicco, Liron Pantanowitz, Anil V Parwani. University of Pittsburgh Medical Center, Pittsburgh, PA

Background: Online image libraries are a valuable resource for trainees and practicing pathologists. Whole slide images (WSI) have added a new dimension to them. Several online libraries are available, but many lack indexing to allow easy searches. The result is an unsatisfactory user experience due to inefficient process of image searches. Our study was aimed to build a concept based relational database model to optimizing image and WSI archiving and retrieval.
Design: PathRez, a web interfaced and indexed SQL server 2008 (Microsoft, Richmond, VA, US) database of images and WSI, was built using the .NET framework 4 platform (Microsoft, Richmond, VA, US). HTML, CSS and JavaScript was used for client side scripting and VB.NET was used for server side coding. Based on anatomic location, most recent accepted pathologic classification of disease entities, basic pathologic process and image attributes, a 4-tier disease concept model was developed which consisted of a set of unique numerical codes (concept ID). An image or a WSI was represented in the database by one concept ID. Each image also was accompanied by the following fields; Diagnosis and comment. This classification was used for optimizing database search and to render the web user interface at runtime. The user interface consisted of a header, navigation bar and a content area. Images were displayed using an in-built image viewer or vendor-provided viewers for WSI. The website is currently hosted in the UPMC system and is accessible from work and home (via remote access).
Results: The web interface welcomes the user with a random image from the databank, which currently contains over 500 WSIs of genitourinary, head and neck, pulmonary, infectious disease and cardiac pathology and 500 gross and microscopic images. A brief description with the contributor name of accompanies the latter. Image can be searched using quick search function or advanced search features.
Conclusions: This concept based image library provides a reference database for users at different stages of pathology training and practice. The underlying relational database design makes this a unique tool for fast and efficient searches. Such a tool is applicable as an online reference for beginners, instructional resource for tutors, trainee evaluation tool and a forum for sharing interesting images.
Category: Informatics

Tuesday, March 20, 2012 1:00 PM

Poster Session IV # 226, Tuesday Afternoon


Close Window