Neuroimaging is widely used to study neurodegeneration, neurodevelopment, neuropsychiatry, and brain cognitive function. In neuroimaging-based studies, the focus of study often centers on elucidating associations between brain structure or function measurements and clinical measurements. We refer to this as neuroimaging marker detection because its goal focuses on identifying neuroimaging markers characterizing a brain disorder or a cognitive process. Statistical analysis methods play a crucial role in inferring the associations among brain structure or function measurements and the clinical variable. The dominant approach for neuroimaging marker detection is general linear model (GLM) based and mass- univariate. The limitations of this approach are its focus on functional segregation, its inability to model multivariate interactions among brain regions, the assumption of normality, the cut-off threshold problem, and the restriction of detecting linear associations. To address these limitations, we have developed a Bayesian statistical analysis method, called Graphical-model-based Multivariate Analysis (GAMMA), for neuroimaging marker detection. The differences between GAMMA and GLM-based mass-univariate methods are three-fold. First, GAMMA is nonparametric: it does not rely on statistical assumptions, such as normality. Second, GAMMA is multivariate, whereas general-linear-model based methods are mass-univariate. GAMMA has the potential to detect nonlinear, multivariate associations among brain regional structure or function measurements and the status of a clinical variable, which would not be detected by univariate approaches. Third, GAMMA is fully automated: it requires minimum user input, such as region-of-interest (ROI) specification or a predefined significance threshold. Other features of GAMMA include the explicitly modeling of spatial correlations among voxels and model stabilization mechanism. We have applied GAMMA for structural magnetic resonance (MR) image analysis, function MR analysis, and lesion-deficit analysis. GAMMA has been under-utilized due to the lack of resources to improve the usability and interoperability. The goal of the present proposal is to develop an open-source and well documented Bayesian neuroimaging data analysis software with enhanced usability, interoperability, and accessibility.
The specific aims are:
Specific Aim 1 : Enhance the software's reporting, project management, and visualization module.
This specific aim focuses on designing a graphical user interface (GUI). This GUI can present the results generated by GAMMA in a readily understandable and interpretable fashion for other investigators. This GUI also supports project management and visualization.
Specific Aim 2 : Enhance the software's ability to support various image and data file formats.
This specific aim i ncludes providing support for analyze, NIFTI, and DICOM image format;add support module to read clinical variables in CSV (Comma Separated Values) format and XML (Extensible Markup Language);and output the generated statistical model in a format that is widely used by data mining community.
Specific Aim 3 : Ensure interoperability between GAMMA and widely used neuroimaging data-analysis tools. We will develop a SPM5-GAMMA toolbox;write user-manual sections that describe how to use GAMMA to analyze the results generated by AFNI or FSL, and how to visualize GAMMA's output using AFNI or FSL;and develop a GAMMA application programming interface (API) to facilitate users to incorporate GAMMA into their software.
Specific Aim 4 : Completely document the application, providing user and developer's manuals. We will provide source code, toolbox, developer toolkit, sample test data, pre-compiled end user software for different platforms, and documentation in NIH Neuroimaging Informatics Tools and Resources Clearinghouse (NITRC), in order to share it with scientific community.

Public Health Relevance

This project centers on enhancing an open source Bayesian data mining tool to analyze neuroimaging data. This tool can identify neuroimaging markers characterizing a clinical condition in the study of neurodegeneration, neurodevelopment, neuropsychiatry, and brain cognitive function.

Agency
National Institute of Health (NIH)
Institute
National Institute of Biomedical Imaging and Bioengineering (NIBIB)
Type
Small Research Grants (R03)
Project #
1R03EB009310-01A1
Application #
7758684
Study Section
Neurotechnology Study Section (NT)
Program Officer
Luo, James
Project Start
2009-09-30
Project End
2011-09-29
Budget Start
2009-09-30
Budget End
2011-09-29
Support Year
1
Fiscal Year
2009
Total Cost
$157,500
Indirect Cost
Name
University of Pennsylvania
Department
Radiation-Diagnostic/Oncology
Type
Schools of Medicine
DUNS #
042250712
City
Philadelphia
State
PA
Country
United States
Zip Code
19104
Chen, Rong; Krejza, Jaroslaw; Arkuszewski, Michal et al. (2017) Brain morphometric analysis predicts decline of intelligence quotient in children with sickle cell disease: A preliminary study. Adv Med Sci 62:151-157
Chen, Rong; Herskovits, Edward H (2015) Examining the multifactorial nature of a cognitive process using Bayesian brain-behavior modeling. Comput Med Imaging Graph 41:117-25
Chen, Rong; Herskovits, Edward H (2012) Graphical model based multivariate analysis (GAMMA): an open-source, cross-platform neuroimaging data analysis software package. Neuroinformatics 10:119-27
Jiao, Yun; Chen, Rong; Ke, Xiaoyan et al. (2012) Single nucleotide polymorphisms predict symptom severity of autism spectrum disorder. J Autism Dev Disord 42:971-83
Chen, Rong; Resnick, Susan M; Davatzikos, Christos et al. (2012) Dynamic Bayesian network modeling for longitudinal brain morphometry. Neuroimage 59:2330-8
Chen, Rong; Jiao, Yun; Herskovits, Edward H (2011) Structural MRI in autism spectrum disorder. Pediatr Res 69:63R-8R
Jiao, Y; Chen, R; Ke, X et al. (2011) Predictive models for subtypes of autism spectrum disorder based on single-nucleotide polymorphisms and magnetic resonance imaging. Adv Med Sci 56:334-42
Chen, Rong; Herskovits, Edward H (2010) Machine-learning techniques for building a diagnostic model for very mild dementia. Neuroimage 52:234-44
Jiao, Yun; Chen, Rong; Ke, Xiaoyan et al. (2010) Predictive models of autism spectrum disorder based on brain regional cortical thickness. Neuroimage 50:589-99