Real-time EEG artifact correction during fMRI using ICA

- A real-time method based on ICA (rtICA) is proposed to remove artifacts from EEG data acquired simultaneously with fMRI.
- The rtICA effectively reduces ocular, motion, BCG, muscle and residual MR artifacts and retrieves EEG signals.
- The rtICA method following the rtAAS outperforms the rtAAS for removing artifacts in real time.
- The rtICA revealed reliable artifact suppression results for further applications of real-time multimodal EEG-fMRI.

BackgroundSimultaneous acquisition of EEG and fMRI data results in EEG signal contamination by imaging (MR) and ballistocardiogram (BCG) artifacts. Artifact correction of EEG data for real-time applications, such as neurofeedback studies, is the subject of ongoing research. To date, average artifact subtraction (AAS) is the most widespread real-time method used to partially remove BCG and imaging artifacts without requiring extra hardware equipment; no alternative software-only real time methods for removing EEG artifacts are available.New methodsWe introduce a novel, improved approach for real-time EEG artifact correction during fMRI (rtICA). The rtICA is based on real time independent component analysis (ICA) and it is employed following the AAS method. The rtICA was implemented and validated during EEG and fMRI experiments on healthy subjects.ResultsOur results demonstrate that the rtICA employed after the rtAAS can obtain 98.4% success in detection of eye blinks, 4.4 times larger INPS reductions compared to RecView-corrected data, and effectively reduce motion artifacts, as well as imaging and muscle artifacts, in real time on six healthy subjects.Comparison with existing methodsWe compared our real-time artifact reduction results with the rtAAS and various offline methods using multiple evaluation metrics, including power analysis. Importantly, the rtICA does not affect brain neuronal signals as reflected in EEG bands of interest, including the alpha band.ConclusionsA novel real-time ICA method was proposed for improving the EEG quality signal recorded during fMRI acquisition. The results show substantial reduction of different types of artifacts using real-time ICA method.