Examination of Non-Physiological Sources of 1/f-like Noise in EPI Power Spectra
Soltysik D, Hyde J, Roopchansingh V
Medical College of Wisconsin
Understanding the noise structure of EPI time-series is important for fMRI data analysis. Previous studies have found a 1/f-like noise structure in data collected from human brains and phantoms. In this study, the 1/f-like shape has been quantized on a voxelwise basis by fitting EPI power spectra to 1/f&[alpha] curves. &[alpha] values are seen to vary across tissue type, phantom type, TR, and voxel size. Based on the results revealed here, it is believed small voxel acquisition is desired to avoid 1/f-like noise from non-physiological sources when performing fMRI.