APIII - Advancing Practice, Instruction & Innovation Through Informatics

GridCAD Microscopy: A caBIG Based System for Image Processing and Quantitative Analysis

Tony Pan1, MS (tpan@bmi.osu.edu); Ashish Sharma1, PhD; Metin Gurcan1, PhD; Kun Huang1, PhD; Gustavo Leone2, PhD; Joel Saltz1, MD, PhD.  1Department of Biomedical Informatics, The Ohio State University, Columbus, Ohio; 2Department of Molecular Genetics, The Ohio State University, Columbus, Ohio.

Context: Increasing utilization of digital microscopy in pathology and basic biological research creates large volumes of image data.  Inter-institutional access and analysis of these image datasets pose significant computational and communications challenges.  Here we present our effort in the design and implementation of a caBIG™ caGrid-based system for distributed image access and algorithm execution.

Technology: Histological sections of mouse liver are prepared with macrophage specific stain and digitized.  The images and results are managed using Mobius MakoDB XML database.  The image database and existing analysis algorithms are integrated into the GridCAD Microscopy system using the caBIG™ caGrid toolkit.  caBIG™ grid services use XML for communication of image and analysis results, and the XML documents conform to published schemas in caGrid’s Global Model Exchange.  The schemas are also used to create MakoDB databases and validate XML documents.

Design:  The gridCAD Microscopy system consists of Image Data Services (IDS), analytical services (CS), Results Management Service (RMS), and User Interface (GUI). IDS exposes image databases as caGrid services with query, retrieve, and store functionalities. CS provides grid interface to analytical functions including segmentation and quantification, and interact with IDS and RMS for data retrieval and storage.  Image and results are stored in MakoDB for both IDS and RMS.  The GUI allows query and preview of available image datasets, invocation of one or more CS(s) with selected datasets as input, and display of results.

Results: We have applied the GridCAD Microscopy system to liver macrophage segmentation and quantification.  XML schemas have been developed to represent microscopy images, segmentation output, and quantification results  An automated algorithm identifies the macrophages from background, and the density of macrophages is then computed.  Algorithms are exposed as CSs, while segmentation and density results are stored in IDS and RMS respectively.

Conclusions: The GridCAD Microscopy is a caGrid-based software system that enables distributed execution of multiple image analysis and quantification algorithms on geographically distributed image datasets.  This system demonstrates a framework in which multi-center research projects, involving analyses of histology images with a range of image processing and quantification algorithms, can be carried out in an efficient way.