Large-scale medical imaging studies have collected a rich set of ultra-high dimensional imaging data, behavioral data, and clinical data in order to better understand the progress of neuropsychiatric disorders, neurological dis- orders and stroke, normal brain development, diagnosis of colorectal cancer, osteoarthritis, and prostate cancer, among many others. However, the development of statistical and computational methods for the joint analysis of imaging and clinical data has fallen seriously behind the technological advances. Three common and important themes of these image data are (T1) ultra-high dimensional functional data with a multi-dimensional tensor struc- ture, (T2) complex geometric structures of human organs, and (T3) complex spatial correlation structures. To meet this critical and important challenge, we will establish a comprehensive statistical framework by addressing three methodological problems. First, there are few efficient and fast methods on modeling high-dimensional imaging data as piecewise smooth functions, while accounting for themes (T2) and (T3). Second, there are few efficient methods on the use of ultra-high dimensional tensor data to predict cognitive development and high- dimensional imaging data, while accounting for the themes (T1)-(T3). Third, little has been done on the analysis of imaging data from longitudinal twin studies. We will establish a comprehensive statistical framework to address these methodological problems. Specifically, we will develop a class of hierarchical functional process models, a class of functional tensor prediction process models, and a class of functional structural equation process mod- els. Scientifically, these new statistical methods are motivated by the analysis of a longitudinal neuroimaging database on early brain development in high-risk children from the Conte study. Our new methods can dramatically increase scientists' ability to better address important scientific questions associated with many imaging studies, particularly those for the Conte study. As these tools are being developed, they will be evaluated and refined through extensive Monte Carlo simulations and the Conte database. Companion software, which will pro- vide much needed analytic tools for the joint analysis of imaging and clinical data, will be disseminated to imaging researchers through ://www.nitrc.org/ and ://www.bios.unc.edu/research/bias.The proposed methodology will have wide applications in neuropsychiatric and neurodegenerative diseases, neurological disorders and stroke, and osteoarthritis, among others.

Public Health Relevance

The project proposes to analyze imaging, behavioral, and clinical data from a longitudinal neuroimaging database on early brain development in high-risk children from the Conte study. New statistical methods are developed and verified by using extensive simulation studies and the Conte dataset.

Agency
National Institute of Health (NIH)
Institute
National Institute of Mental Health (NIMH)
Type
Research Project (R01)
Project #
5R01MH086633-06
Application #
8879203
Study Section
Biostatistical Methods and Research Design Study Section (BMRD)
Program Officer
Freund, Michelle
Project Start
2009-07-01
Project End
2019-06-30
Budget Start
2015-09-17
Budget End
2016-06-30
Support Year
6
Fiscal Year
2015
Total Cost
$381,594
Indirect Cost
$124,185
Name
University of North Carolina Chapel Hill
Department
Biostatistics & Other Math Sci
Type
Schools of Public Health
DUNS #
608195277
City
Chapel Hill
State
NC
Country
United States
Zip Code
27599
Kim, Janet S; Staicu, Ana-Maria; Maity, Arnab et al. (2018) Additive Function-on-Function Regression. J Comput Graph Stat 27:234-244
Zhang, Zhengwu; Descoteaux, Maxime; Zhang, Jingwen et al. (2018) Mapping population-based structural connectomes. Neuroimage 172:130-145
Miranda, Michelle F; Zhu, Hongtu; Ibrahim, Joseph G et al. (2018) TPRM: TENSOR PARTITION REGRESSION MODELS WITH APPLICATIONS IN IMAGING BIOMARKER DETECTION. Ann Appl Stat 12:1422-1450
Li, Tengfei; Xie, Fengchang; Feng, Xiangnan et al. (2018) Functional Linear Regression Models for Nonignorable Missing Scalar Responses. Stat Sin 28:1867-1886
Tang, Man-Lai; Tang, Niansheng; Zhao, Puying et al. (2018) Efficient Robust Estimation for Linear Models with Missing Response at Random. Scand Stat Theory Appl 45:366-381
Wang, Ching-Wei; Lee, Yu-Ching; Calista, Evelyne et al. (2018) A benchmark for comparing precision medicine methods in thyroid cancer diagnosis using tissue microarrays. Bioinformatics 34:1767-1773
Kang, Kai; Song, Xinyuan; Hu, X Joan et al. (2018) Bayesian adaptive group lasso with semiparametric hidden Markov models. Stat Med :
Yang, Hojin; Zhu, Hongtu; Ibrahim, Joseph G (2018) MILFM: Multiple index latent factor model based on high-dimensional features. Biometrics 74:834-844
Chen, Stephanie T; Xiao, Luo; Staicu, Ana-Maria (2018) A Smoothing-based Goodness-of-Fit Test of Covariance for Functional Data. Biometrics :
Kim, Janet S; Maity, Arnab; Staicu, Ana-Maria (2018) Additive Nonlinear Functional Concurrent Model. Stat Interface 11:669-685

Showing the most recent 10 out of 122 publications