Abstract #4127
Bayesian shrinkage as an alternative to spatial smoothing for multi-echo BOLD fMRI
Feng Xu 1,2 , Joseph S. Gillen 1,2 , Hongjun Liu 3 , Ann Choe 1,2 , Hua Jun 1,2 , Craig K. Jones 1,2 , Suresh E. Joel 1 , Brian S. Caffo 4 , Martin A. Lindquist 4 , Ciprian M. Crainiceanu 4 , Peter C. van Zijl 1,2 , and James J. Pekar 1,2
1
Russell H. Morgan Department of Radiology,
Johns Hopkins University, Baltimore, MD, United States,
2
F.M.
Kirby Research Center, Kennedy Krieger Institute,
Baltimore, MD, United States,
3
Department
of Radiology, Guangdong General Hospital, Guangdong
Academy of Medical Sciences, Guangzhou, Guangdong,
China,
4
Biostatistics,
School of Public Health, Johns Hopkins University,
Baltimore, MD, United States
Spatial smoothing is the most popular way to enhance
sensitivity in fMRI analysis, at a cost of coarsened
spatial specificity. Multi-echo acquisitions can enhance
specificity in fMRI by allowing analysis of effective
transverse relaxation rate (R2*) via least-squares (LS)
fitting to each voxels echo decay. Bayesian shrinkage
improves parallel simultaneous estimation of many
similar parameters by borrowing strength from parallel
measurements. Here, we shrink over grey matter by
applying Bayesian shrinkage to estimation of R2* in grey
matter voxels, and show that shrinkage increases fMRI
sensitivity (with respect to LS fitting) without the
blurring caused by spatial smoothing.
This abstract and the presentation materials are available to members only;
a login is required.
Join Here
