Trial classification: automatic detection of trials containing artefacts in MEG data

meg-trial-classification

This repository is a project in collaboration between dianadima and lorenzomagazzini. The code was written for two purposes:

1. Features extraction: extract informative summary statistics from epoched MEG data (directory trialfeatures);
2. Classification: use the features as input to a Support Vector Machine (SVM) classifier, in order to automatically identify trials containing artefacts (directory trialclassification).

The code can be run using the example .mat files included within this repository (directory data).

Dependencies

This code requires the Fieldtrip toolbox and was last tested with FieldTrip version 20180630 in MATLAB R2015a.

The SVM classifier requires two libraries, liblinear and libsvm, both included within this repository.

The colour maps in some of the figures were generated using cmocean for MATLAB and viridis for MATLAB.