Removal of phase artifacts from fMRI data using a Stockwell transform filter improves brain activity detection

摘要

A novel and automated technique is described for removingfMRI image artifacts resulting from motion outside of the imagingfield of view. The technique is based on the Stockwelltransform (ST), a mathematical operation that provides the frequencycontent at each time point within a time-varying signal.Using this technique, 1D Fourier transforms (FTs) are performedon raw image data to obtain phase profiles. The timeseries of phase magnitude for each and every point in the phaseprofile is then subjected to the ST to obtain a time-frequencyspectrum. The temporal location of an artifact is identifiedbased on the magnitude of a frequency component relative tothe median magnitude of that frequency’s occurrence over alltime points. After each artifact frequency is removed by replacingits magnitude with the median magnitude, an inverse ST isapplied to regain the MR signal. Brain activity detection withinfMRI datasets is improved by significantly reducing image artifactsthat overlap anatomical regions of interest. The majoradvantage of ST-filtering is that artifact frequencies may beremoved within a narrow time-window, while preserving thefrequency information at all other time points.