Abstract
Magnetoencephalography (MEG) detects very weak magnetic fields originating from the neurons so as to study human brain functions. The original detected MEG data always include interference generated by blinks, which can be called blink artifacts. Blink artifacts could cover the MEG signal we are interested in, and therefore need to be removed. Commonly used artifact cleaning algorithms are signal space projection (SSP) and independent component analysis (ICA). These algorithms need to locate the blink artifacts, which is typically done with the identification of the blink signals in the electrooculogram (EOG). The EOG needs to be measured by electrodes placed near the eye. In this work, a new algorithm is proposed for automatic and on-the-fly identification of the blink artifacts from the original detected MEG data based on machine learning; specifically, the artificial neural network (ANN). Seven hundred and one blink artifacts contained in eight MEG signal data sets are harnessed to verify the effect of the proposed blink artifacts identification algorithm. The results show that the method can recognize the blink artifacts from the original detected MEG data, providing a feasible MEG data-processing approach that can potentially be implemented automatically and simultaneously with MEG data measurement.
Highlights
Magnetoencephalography (MEG) is a technique employing sensitive sensors to detect the weak magnetic field signal generated by the neurons of the brain without invasion or damage [1,2,3]
Compared with the 100% accuracy for the first data set, it can be thought that the false alarms are caused by insufficient training data
A method based on the ANN for the identification of the blink artifacts is proposed, and 66,780 2D images generated from eight MEG signal data sets are used to train and test the convolutional artificial neural network, in which GoogLeNet is used as a pre-trained input
Summary
Magnetoencephalography (MEG) is a technique employing sensitive sensors to detect the weak magnetic field signal generated by the neurons of the brain without invasion or damage [1,2,3]. In order to remove blink artifacts and observe the weak signal triggered by the brain activities more precisely [23,24], the original data need to be processed. Signal space projection (SSP) [25,26,27,28] and independent component analysis (ICA) [29,30] are widely used to remove the blinks and other artifacts in the MEG. These two methods may need extra electrodes to measure the blink signals in the electrooculogram (EOG) and decide the occurrence time of the blink artifacts in the MEG for practical use. Machine learning algorithms are introduced into the MEG data processing [36,37]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
