Meeting Banner
Abstract #3257

Empirical Bayesian Estimation Improves Analysis of Resting-State Functional Connectivity from Multi-Echo BOLD Data

Feng Xu1, 2, Suresh E. Joel, 12, Jun Hua1, 2, Craig K. Jones, 12, Brian S. Caffo3, Martin A. Lindquist4, Ciprian M. Crainiceanu4, Peter C.M. van Zijl1, 2, James J. Pekar1, 2

1Radiology, Johns Hopkins University, School of Medicine, Baltimore, MD, United States; 2F. M. Kirby Research Center, Kennedy Krieger Institute, Baltimore, MD, United States; 3Biostatistics, Johns Hopkins University, Baltimore, MD, United States; 4Biostatistics, Johns Hopkins University, School of Public Health, Baltimore, MD, United States

Multi-echo acquisitions can improve the sensitivity and specificity of resting-state functional connectivity MRI compared with conventional EPI. Analysis of multi-echo decays to estimate transverse relaxation in each voxel is a case of parallel estimation; statistical theory states that popular maximum likelihood (least-squares) methods are in some ways inferior to empirical Bayesian approaches. In this study we show that the James-Stein estimator yields modest increases in the spatial extent and inter-subject concordance of functional networks estimated from multi-echo resting BOLD data acquired at 7 Tesla.

Abstract PDF Presentation Video

Keywords

acceleration accomplished achieve acquisition acquisitions adjusted adults anatomical application applications applied approaches approved aspect assess asterisk audience band biostatistics bold brain cerebrum classify coefficients computed concordance concordant confidence connectivity consent considering consistency converted correlation covering decay decays decrease decreased default denotes derived empirical error errors estimation estimator even excitation extent facilitating false family filter filtered filtering fisher fluid functional goodness health healthy improved improvements improves inferior informed initial interval johns least likelihood many maps math measures medicine mode modest motor network networks nine noise nuisance original outcome outcomes paired parallel participants participate pass permutation peter physiological popular positive preprocessing primary prior prob radiology rage reduce resolution respectively resting runs sample school score scores seed segmentation sense sensitivity series shrink shrinkage significance slice slices spatial specificity squares stein subject subjects summarize superior table target temporal temporally theoretically though threshold tissue transverse twenty type typically unit variance variation volume volumes white yielded yielding