Adaptive neural network classifier for decoding MEG signals

• DOI: 10.1016/j.neuroimage.2019.04.068
• Publication Date: 2019-05-03T20:57:41Z
• Authors: Ivan Zubarev; Rasmus Zetter; Hanna-Leena Halme; Lauri Parkkonen
• Keywords: Magnetoencephalography; Neural decoding

We introduce two Convolutional Neural Network (CNN) classifiers optimized for inferring brain states from magnetoencephalographic (MEG) measurements. design follows a generative model of the electromagnetic (EEG and MEG) signals allowing explorative analysis neural sources informing classification. The proposed networks outperform traditional as well more complex when decoding evoked induced responses to different stimuli across subjects. Importantly, these models can successfully generalize new subjects in real-time classification enabling efficient brain–computer interfaces (BCI).