Making analysis protocols

The general idea is that you make a Matlab script, in which you call the different FieldTrip functions. Such a script (or set of scripts) can be considered as an analysis protocol, since in them you are defining all the steps that you are taking during the analysis.

For most of the analysis steps, you will be able to use a function from the FieldTrip toolbox, but sometimes you also will want to include your own Matlab code in the script.

Some standard analysis protocols are given below. The figures indicate which functions you (probably) will use in the analysis, and give you a guideline in finding the documentation that you need.

ERP/ERF analysis consists of preprocessing, averaging the data timelocked to the stimulus or response, optionally averaging over subjects and/or testing for significant effects and finally plotting the result.

Frequency analysis consists of preprocessing, performing a Fourier or wavelet decomposition of the data, optionally averaging over subjects and/or testing for significant effects and finally plotting the result.

A schematic display of the analysis steps for source reconstruction using a beamformer approach is given below. Prior to any source reconstruction, you should have performed a complete timelock or frequency analysis of the data at the channel level.

If you do a source reconstruction of the data for multiple subjects, and if you use individual volume conduction models for your subjects (common for MEG), then the individual source reconstructions cannot immediately be averaged or used for statistical comparison. The anatomy of each subject is different (i.e. the shape and size of the brain), and the locations of the voxels (i.e. the dipoles) will also be different for each subject. Using the individual subject's anatomical MRI, you can align all subjects to a template brain using a non-linear deformation. The same deformation can be applied to the functional data. Subsequently the subjects functional data can be compared.

The spatial normalisation towards a template brain is done in FieldTrip with the NORMALISEVOLUME function, which internally calls some low-level functions from the SPM2 toolbox. In general, it is also advisable to normalise your individual subject's data to the SPM/MNI template brain.

Below you can see the protocol that you would use for averaging the source reconstruction over subjects and for group statistics on the source level. If the statistical test involves two conditions, then you should do the source normalisation in both conditions for all subjects and feed the two grandaveraged source reconstructions into SOURCESTATISTICS.