NSLS-II   |   Brookhaven National Laboratory

An Overview of Multi-dimensional, Multi-modal, Image Registration Methods and their Application with the Insight Toolkit (ITK) and Tomviz

Matthew McCormick, Kitware, Inc.

New scientific insights into materials are possible when information from images obtained with multiple modalities or at multiple scales are combined.  Registration through automated or semi-automated software algorithms enables quantification and visualization of multi-dimensional and multi-modal image information.

In this talk, first we describe challenges in this analysis domain, including variations in voxel intensities, noise, and artifacts across modalities.  Another challenge is divergence in structure features at different scales.

Next, we describe approaches to the multi-modal registration problem.  These include mutual information-based image similarity metrics, registration of segmented structures, and registration of feature points.

Finally, we describe how these approaches can be applied with algorithms implemented in the open-source, Python-wrapped Insight Toolkit (ITK).  We describe how they can be applied in Tomviz, an open source application for materials science and tomography based on the Visualization Toolkit (VTK), ParaView, ITK, and Python.