The analysis of electrophysiological signal (M/EEG) has significantly benefited from multi-variate methods, aka decoding. It allows to extract the information from all sensors and use to discriminate between two (or more) conditions for each time-point. However, this approach derives from fMRI (Haxby, 2012) where larger numbers of voxels are available. In electrophysiology, only a few spatial information sources (electrodes/sensors) are available. In contrast, M/EEG excels in temporal resolution, where time-points contains the rich information. Therefore, the use of temporal decoders (using time and no electrodes) becomes crucial to better capture the cognitive information in the brain, leading to topographic maps of decoding rates.
Recently, we developed a new toolbox called Time-GAL toolbox (Santos-Mayo, 2025) where temporal decoders are trained and tested in different electrodes/brain areas. As a result, the cross-decoding matrix (Generalization Across Location, GAL) indicates potential connectivity between brain areas behaving similarly during a cognitive task. It depicts a MVPA-based procedure to estimate relationship between regions during the processing of an stimulus or cognitive processing. Also, linking such matrix with the features (in this case time-points) the methodology reveals the cognitive process in terms of space, time and connectivity.
Time-GAL toolbox and example
The adaptation of this methodology into MNE-python is extremely easy and all functions are already coded, as the temporal decoders can use the scripts/functions used in already-implemented spatial decoders (only a transpose of time*sensor input data is required). No coding work is needed to expand the current MVPA/decoding documentation by adding this new Time-GAL. The inclusion of such methodology only involves changes in documentation, elaboration of one example and, maximum, wrap existing functions into a new name.
The analysis of electrophysiological signal (M/EEG) has significantly benefited from multi-variate methods, aka decoding. It allows to extract the information from all sensors and use to discriminate between two (or more) conditions for each time-point. However, this approach derives from fMRI (Haxby, 2012) where larger numbers of voxels are available. In electrophysiology, only a few spatial information sources (electrodes/sensors) are available. In contrast, M/EEG excels in temporal resolution, where time-points contains the rich information. Therefore, the use of temporal decoders (using time and no electrodes) becomes crucial to better capture the cognitive information in the brain, leading to topographic maps of decoding rates.
Recently, we developed a new toolbox called Time-GAL toolbox (Santos-Mayo, 2025) where temporal decoders are trained and tested in different electrodes/brain areas. As a result, the cross-decoding matrix (Generalization Across Location, GAL) indicates potential connectivity between brain areas behaving similarly during a cognitive task. It depicts a MVPA-based procedure to estimate relationship between regions during the processing of an stimulus or cognitive processing. Also, linking such matrix with the features (in this case time-points) the methodology reveals the cognitive process in terms of space, time and connectivity.
Time-GAL toolbox and example
The adaptation of this methodology into MNE-python is extremely easy and all functions are already coded, as the temporal decoders can use the scripts/functions used in already-implemented spatial decoders (only a transpose of time*sensor input data is required). No coding work is needed to expand the current MVPA/decoding documentation by adding this new Time-GAL. The inclusion of such methodology only involves changes in documentation, elaboration of one example and, maximum, wrap existing functions into a new name.