An automatic identification and removal method for eye-blink artifacts in event-related magnetoencephalographic measurements

While measuring event-related magnetoencephalographic (MEG) signals using visual stimuli, eye blinks were inevitable and generated large magnetic artifacts. Since trials containing eye blinks were excluded from the analyses, the signal-to-noise ratio of the event-related signals was decreased. In this study, we propose a method to identify the eye blink magnetic artifacts and remove them automatically using independent component analysis preprocessed by principal component analysis. The method evaluates the spatiotemporal similarity between independent components and both MEG and electro-oculogram data based on a newly devised cost function. Testing of the method on event-related MEG signals measured by a 306-channel whole-head system in a visual perception task led to the successful identification and removal of eye-blink artifacts in all trials containing eye blinks from all the seven subjects.