Real-time filtering of gradient artifacts from simultaneous EEG-fMRI data

Abstract

EEG and fMRI are extremely popular tools to study patterns of functional brain activity. Their utility can be further enhanced when used together in simultaneous EEG-fMRI recordings. However, such recordings are ridden with artifacts due to the gradients switching within an MRI machine. These artifacts need to be filtered before the data can be further processed. Numerous tools exist for filtering such data. However, if one needed to use the data for real-time feedback (such as neurofeedback), the current methods would be too slow. This paper discusses parallel versions of the current methods and a novel FFT based method that reduces the computation time of current methods by a factor of 3 and 23 respectively. This facilitates the use of an EEG-fMRI dataset in real-time neurofeedback studies.