MEG Cross-Temporal Meta decoding improves performance by using State Information

Nakai Fumiya 1751076


While Magnetoencephalography (MEG) signal has rich temporal and spatial information, it is difficult to interpret by existing decoding analysis techniques.

In this study, we proposed new MEG single trial decoding method called Cross-Temporal Meta decoding (CTM-decoding), which is the expansion of the Cross-Temporal decoding to the multi-layer structure.

We tested the effectivity of the proposed method to the 1 simulation data and 2 empirical MEG data and it showed that proposed method recorded significantly higher performance than previous result.

The proposed method also showed clearer dynamic-state representation than the previous method and it revealed some meaningful dynamic-state feature in the empirical data.

We conclude that these advantages come from the combination of non-diagonal entrainment correction mechanism and non-diagonal updating correction mechanism.