Meeting Banner
Abstract #2100

Grading Resting-State FMRI Datasets by Reweighted L1 Regression

Chia-Jung Yeh1, Yu-Sheng Tseng1, Teng-Yi Huang1, Shang-Yueh Tsai2, 3

1Dept. of EE, National Taiwan University of Science and Technology, Taipei, Taiwan; 2Graduate Institute of Applied Physics, National Chengchi University, Taipei, Taiwan; 3Reasearch Center for Mind, Brain and Learning, National Chengchi University, Taipei, Taiwan

The rest-state fMRI (rsfMRI) detects the dynamic neuronal activity using a long series of BOLD imaging and measures the connectivity of brain functional areas using either correlation analysis or independent component analysis. However, in our experience, rsfMRI sometimes shows unstable results even if we preformed study using the same imaging protocols and data analysis methods. In this study, we proposed to use reweighted L1 regression, a form of robust regression, to find the outliers of the rsfMRI time series in the default-mode network (DMN) and developed a grading method for rsfMRI datasets.

Keywords

abstract according activities activity amount analyzed anything applied asleep authors awake body bold brain called caused clearly coefficient coefficients component connectivity consistent control coordinates correction correlation corrupted cortex count counted criterion dataset datasets default denoted dept desirable detects determine determined develop developed differently displays distributed distribution either emerging empirically even exhibit experiences experiment explore find finding foot form formed generated grade gradient grading graduate head highly improper in vivo independent index institute instructed intensities intensity intercept intrinsic known learning listed long mapping maps material matrices matrix measures mind minimization mode motion national network networks neuronal notice outliers pair pairs parallel particular physics polynomial potentially powerful preform preliminary preprocessing processing projection promising proposed protocols quick rapidly reasons reconstructed regression relatively remain respectively rest resting robust scanner seed seems selected sensitive series serve significantly space spontaneous studies subject subjects subsequent substantial table technology think threshold thresholds unstable useful utilized various visually volunteers weakness weightings whole