Continued Development and Maintenance of ITK-SNAP 3D Image Segmentation Software

点击翻译按钮获取中文摘要

This project seeks to continue developing and maintaining a software application ITK-SNAP, which provides functionality for user-guided automatic segmentation and manual annotation of 3D volumes generated by biomedical imaging. ITK-SNAP is a free, open-source software tool that has a large number of users in the biomedical community (estimated in the thousands) and has contributed to over 200 publications since 2006, spanning a wide range of biomedical applications and imaging modalities. Furthermore, ITK-SNAP occupies a unique place in the spectrum of open-source tools available to today's imaging researcher, with a mature user interface and functionality specifically focused on the problem of image segmentation. The broad goals of this project are to ensure the long-term availability and viability of ITK-SNAP in the face of ever increasing complexity of imaging datasets and rapidly changing software environment; and to significantly expand the class of biomedical image segmentation problems that can benefit from the automatic features of ITK-SNAP. Five specific aims are proposed to achieve these goals. Aim 1 will develop a novel software framework for semi-automatic segmentation of multimodality and multichannel imaging data. This aim will extend the existing active contour segmentation framework with a flexible toolbox for user-guided generation of object/background probability maps from image volumes. The toolbox will support texture analysis and pattern classification, as well as user-generated spatial segmentation priors. Aim 2 will boost the performance of ITK-SNAP by employing graphics card acceleration and will change the internal data structures to allow the tool to work with very large image volumes, like those produced by high-resolution multi-slice CT or confocal microscopy. Aim 3 will remove ITK-SNAP dependencies on an aging and poorly supported FLTK user interface software library, transitioning instead to the QT library, which has strong community and industry support. This aim will also make critical improvements to ITK-SNAP usability, including support for the project paradigm. In Aim 4, segmentation protocols based on new ITK-SNAP functionality will be developed to address a diverse set of biomedical image segmentation problems. These protocols will then be validated against manual segmentation using public datasets. The criterion for success is to achieve a two-fold or better reduction in segmentation time with no penalty in inter-observer or intra-observer reliability. Aim 5 is to continue supporting the ITK-SNAP user community by implementing user-requested features, correcting defects in the software, and providing thorough documentation, including video tutorials, and training and outreach efforts.