This proposal aims to develop object oriented data analysis (OODA) methods that are highly novel and complementary to existing methods of analysis of human brain scan connectomes, defined as graphs representing brain anatomical or functional connectivity. OODA is an emerging field of statistics where classical statistical approaches (e.g., hypothesis testing, regression, estimation, confidence intervals) are applied to data objects such as graphs or functions. By analyzing data objects directly we can avoid loss of information that necessarily occurs when data objects are transformed into numerical summary statistics. The conceptual leap in this proposal is that OODA statistical methods will be developed and applied directly to connectomes without needing to transform them into summary features which incur loss of information. By providing statistical tools that analyze sets of connectomes without loss of information, new insights into neurology and medicine may be achieved.
The Specific Aims of this proposal are: (1) Develop OODA methods and software for analyzing human connectome data. More specifically, a mathematical framework for hypothesis testing, regression and Principal Components Analysis will be developed to model and analyze set of connectomes. The proposed methodology will allow to compare groups of connectomes (e.g., Is the brain structure/function different in cases versus controls?), to perform regression for modeling connectomes as a function of subject covariates such as age, gender, disease, or longitudinally (e.g., How does the brain structure/function change over time and does it change differently in males and females?), and to measure sources of structural or functional variation across populations of connectomes (e.g., What is the natural variability of brain structure/function within the population?);(2) Validate the tools developed in Specific Aim 1 using existing connectome datasets generated by our co-investigators to answer clinical questions relevant to their research objectives. The validation of the methodology to be developed in connectome data from will contribute to assess the biological significance of the methods proposed to be developed;and (3) Compare the performance of connectome OODA methods developed in Specific Aim 1 as complementary to existing methods of analysis of human brain scan connectomes by analyzing the same data used in Specific Aim 2 using graph-theoretical measurements.
Developing statistical methods for hypothesis testing, regression, and principal component analysis of sets of connectomes without loss of information, new insights into neurology and medicine may be achieved. We propose to develop statistical methods within the framework of object oriented data analysis (OODA) which do not require a reduction of the connectome to features, and therefore avoid loss of information. This proposal will help bridge the translation of connectome to clinical applications.
|La Rosa, Patricio S; Brooks, Terrence L; Deych, Elena et al. (2016) Gibbs distribution for statistical analysis of graphical data with a sample application to fcMRI brain images. Stat Med 35:566-80|