Building Geometric Databases for Anatomy-Based Spatial Queries

Washington University, Saint Louis MO

Investigators

Abstract

This grant aims to develop an atlas-based database architecture suited for organizing 3D anatomical images for efficient spatial queries. Utilizing a new form of atlas, the project will develop tools for accurate and efficient alignment of 3D images collected from a large number of subjects with varying anatomical shapes. The resulting database will facilitate convenient, web-based spatial queries of functional, metabolic, mechanical, and structural data within customizable regions of interest among stored images. As a test-bed application, the project will construct two web-based databases of 3D gene expression patterns for the mouse brain and heart from stacks of 2D tissue sections. At the heart of the architecture is a new type of geometric atlas, represented as a subdivision mesh, which models the partitioning and interior of an anatomical region of interest, such as the brain. The geometric and multi-resolution structure of the atlas allows for flexible registration onto individual subjects with varying anatomical shapes, and enables fast, multi-level comparison of registered data. The tools for constructing, populating, and querying atlas-based databases will be made freely available online to be easily integrated with existing databases for augmented spatial querying capabilities. The use of biological imaging in science continues to grow, and it has thus become critical to many research efforts to be able to efficiently utilize the enormous resources of spatial information being generated. The gene expression databases resulted from the project will enable querying and clustering of genes based on similarity of 3D expression patterns, which may lead to significant breakthroughs in understanding biological processes. The long term goal of this project is to facilitate collaborative knowledge discovery across subjects, image modalities, and geographical locations.