Optimized moving-average filtering for gradient artefact correction during simultaneous EEG-fMRI

The strong capability of the combined EEG-fMRI for investigating and revealing new insights on mapping of the brain activity as well as on several other neuroscientific studies has attracted the interest of researchers and clinicians over the past years. However, its consolidation as a powerful and independent technique still depends on enhancing the quality of the EEG signal, mainly due to the occurrence of artefacts. This paper presents a simple and effective approach for removal of the gradient artefact, which is induced in the EEG by the rapidly varying gradient magnetic fields of the fMRI scanner. According to our method, a moving-average filter is used to perform the removal of the gradient artefact. Nevertheless, rather than estimation of an artefact waveform template to be subtracted and achieve the EEG restoration, we have proposed to optimize the moving-average filtering process along the entire EEG excerpt. Thereby, the restored EEG can be estimated either from a sum of partial waveform components resulting from the recursive application of the optimized moving-average filter; or from an estimative of the artefact along the entire excerpt. Our methodology shows to achieve a quite satisfactory restoration of the EEG signal, even for low signal amplitudes. Moreover, in addition to predict the variability of the artefact waveform over the time, synchronization between EEG and fMRI clocks and extensive data segmentation are not required as well.