A system for scalable 3D visualization and editing of connectomic data

Abstract

The new field of connectomics is using technological advances in microscopy and neural computation to form a detailed understanding of structure and connectivity of neurons. Using the vast amounts of imagery generated by light and electron microscopes, connectomic analysis segments the image data to define 3D regions, forming neural-networks called connectomes. Yet as the dimensions of these volumes grow from hundreds to thousands of pixels or more, connectomics is pushing the computational limits of what can be interactively displayed and manipulated in a 3D environment. The computational cost of rendering in 3D is compounded by the vast size and number of segmented regions that can be formed from segmentation analysis. As a result, most neural data sets are too large and complex to be handled by conventional hardware using standard rendering techniques. This thesis describes a scalable system for visualizing large connectomic data using multiple resolution meshes for performance while providing focused voxel rendering when editing for precision. After pre-processing a given set of data, users of the system are able to visualize neural data in real-time while having the ability to make detailed adjustments at the single voxel scale. The design and implementation of the system are discussed and evaluated.