EEG gradient artifact removal by compressive sensing and Taylor-Fourier transform

Simultaneous recording of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) represents a powerful tool for brain activity investigation. Unfortunately, EEG data collected during concurrent fMRI are affected by very large artifacts. This paper focuses on the gradient artifact (GRA), related to the sawtooth profiles of magnetic flux inside the MRI scanner. A novel removal algorithm is proposed and validated on both simulation and experimental data. A super-resolution method, based on compressive sensing, determines GRA harmonic frequencies. Amplitudes and phases of GRA components are estimated by means of the Taylor-Fourier transform (TFT), complying with dynamic operating conditions. Unlike averaging techniques, well-known in the literature, this allows computation of a specific template for each artifact occurrence, which is subtracted from the original data. Experimental results show a significant reduction of spurious components in all the considered conditions. No significant distortions are introduced in spectral power distribution, allowing reliable clinical interpretation of the acquired trace.