Meeting Banner
Abstract #2248

A Case for Correcting Attenuation of Correlation in Resting State FMRI

Jacco A. De Zwart1, Peter van Gelderen1, Jeff H. Duyn1

1Advanced MRI, LFMI, NINDS, National Institutes of Health, Bethesda, MD, United States

In low-noise situations the correlation between two signals is underestimated, an effect known as attenuation of correlation. This work shows that this well-established effect can lead to significant underestimation of cross-correlation in resting-state fMRI, where often the ratio between the neurogenic fluctuation of interest and noise level is less than 10. Correction of the cross-correlation value for this effect is well established and straightforward if the noise level is known.

Abstract PDF Poster PDF

Keywords

accounted accuracy actual additional advanced affects already amplitude attenuate attenuation available background bias bilateral brain cardiac closed coefficient coexist coil combining commonly comprised condition connectivity correct corrected correcting correction correctly correlation correlations course courses cross decrease decreases dependence dependent described determining displays documented drop encountered essential established estimating evaluated even experiment experimental eyes fluctuation form function functional functionally generally gradient health illustrate important improve improves institutes interpretation introduced known lateral longer magnitude measure merely modeling motion national neural noise note nuclei nuisance often optimal overcome pairs peter physiologic plot practice procedure properly propose proposed psychology random readily related relevant remains removal report requires respiratory rest resting samples sampling scanning seed series significance simple simulation simulations sources specifically straightforward substantial substantially suggest temporal thermal typical uncorrected underestimated underestimation unpublished variance volunteers whereas yielding zero