1. Introduction
Biomedical research relies increasingly on imaging to organize observations of the biophysical world and to quantify these observations in pursuit of better understandings of biological systems, pathology, and treatments. Advances in computer science have led to stunning improvements and great hope for continued growth through the introduction of faster and more sophisticated processing, distributed computing, and high speed networking (Grid) technologies to facilitate the analysis of high resolution, multi-modality images from increasingly large subject populations. However the medical image computing community has historically been fragmented in its approach to software development and distribution. A result of this fragmentation has been a significant amount of re-implementation of common tools (file I/O, filters, core algorithms, user interfaces, etc.) and at the same time a lack of reproducibility of results due to the complexity of algorithms and variations in the implementations of these core components.