David W. Shattuck, Ph.D. - US grants

DESCRIPTION (provided by applicant): Over the past decade, the PI and colleagues have developed algorithms and software for the segmentation, registration, and analysis of structural MRI and related image data. These software tools have been integrated into a publicly released package, named Brain Suite that has been used in numerous neuroimaging studies. Our approach has emphasized the development of separate, validated modules addressing different aspects of the image analysis problem, which are then integrated through an interactive interface to allow fast automated or semi-automated processing and image visualization. This application proposes the continued development and support of these tools under the "Continued Development and Maintenance of Software (R01)" program. We will improve our software and its utility through enhancements to its functionality, better user and developer documentation, better user training and support, improved interoperability, specific enhancements to support data from subjects with epilepsy, and routine evaluation of the performance of the tools. The Brain Suite software will be distributed under an open-source software license. Aim 1: We will develop the Brain Suite software package to provide advanced automated and interactive software for the segmentation and registration of brain MRI. Specifically, we focus on: (a) software for segmenting individual subject MRI to identify structures, e.g., skull and scalp models, inner and outer cortical surface mesh models, within human brain MRI;(b) software for performing spatial alignment of brain structures across subjects using surface-based and joint surface/volume methods;(c) analysis software for performing comparisons of brain data extracted and mapped into common spaces using the tools in (a) and (b);(d) a cross-platform graphical user interface providing an easy-to-use, interactive framework in which to apply the segmentation, registration, and analysis tools, as well as to perform interactive delineation and visualization of data. All software development will be performed adhering to sound software engineering practices (coding style, documentation, version control). Aim 2: We will develop the resources and capabilities to enhance the utility of our software for a broad range of users from the neuroimaging and clinical communities. This will be achieve through: online documentation, support forums, and training videos and courses;quality assurance testing to detect potential failures during automated processing;and enhanced interoperability with other software packages used by the neuroimaging community. Aim 3: We will enhance the Brain Suite tools to provide improved functionality for processing data with clinical abnormalities, in particular, data from people affected by epilepsy. Aim 4: We will adopt several procedures for routinely evaluating the quality of the results our software produces and the software's impact in the clinical setting and neuroimaging community. PUBLIC HEALTH RELEVANCE: This project proposes to develop a powerful suite of open source integrated software tools that will allow a high degree of automated analysis of 3D images of the human brain, with optional user interaction where necessary, to automatically identify neuroanatomical structures including the cerebral cortex and subcortical structures. The software package will also provide capabilities to align images from multiple subjects into a common space in which intersubject comparison studies may be performed. The performance of the software will be evaluated using standard databases of hand labeled brains as well as clinical data from patients with epilepsy.

Project Summary Over the past 16 years, we have developed a collection of algorithms and software for the segmentation, registration, labeling, and analysis of structural and diffusion MRI, integrated into the open source package BrainSuite (http://brainsuite.org). Our approach emphasizes the development of separate, validated modules addressing each aspect of the image analysis problem, which are then integrated through an interactive interface, to provide fast automated or semi-automated processing and image visualization. Command line tools using the same functions are also provided for large scale processing. The software runs on and is consistent across Mac, Windows, and Linux platforms. This renewal application builds on these tools with a continued emphasis on the ability to process large (now multimodal) data sets while simultaneously retaining the ability to rapidly visualize, review, and where necessary modify intermediate results to optimize the fidelity of each stage of processing. The renewal emphasizes development of new tools for coregistration of multimodal data, modeling and analysis of diffusion data, and quantitative analysis of functional and structural connectivity. The project has five specific aims. Aim 1 will develop advanced methods for intersubject anatomical, diffusion, and functional MRI analysis that account for individual structural and functional differences. This will improve upon existing methods that rely solely on structural (T1-weighed) images to define homologies between subjects. Aim 2 will develop tools for intrasubject coregistration of multimodal imaging data that explicitly account for and estimate resolution differences between modalities. In combination with the intersubject methods in Aim 1, this will facilitate group pointwise and regional statistical multimodal analysis. Aim 3 will develop tools to analyze diffusion data characterized by flexible sampling schemes and multiple b-values, addressing the limited ability of current tools to model data produced by increasingly widely used modern acquisition schemes such as those required by the Human Connectome Project and related NIH projects. Aim 4 will expand the BrainSuite Statistics toolbox, which uses Python and R to provide an extensible statistical framework for analyzing data; this aim will also facilitate the use of BrainSuite as part of larger image analysis pipelines by continuing to support standard formats and developing our new tools as modular command line programs. Distributions will be compatible with Nipype and NITRC-CE. Under Aim 5, we will continue software development employing standard best practices. We will develop web-based interfaces for rapidly visualizing and evaluating results from large, multisubject studies. User support will be provided through online forums, tutorials, videos, documentation, and hands-on training. New analysis methods developed in the above aims will be validated through simulation and evaluation on existing in vivo imaging data.