FT_REJECTCOMPONENT backprojects an ICA (or similar) decomposition to the
 channel level after removing the independent components that contain
 the artifacts. This function does not automatically detect the artifact
 components, you will have to do that yourself.

 Use as
    [data] = ft_rejectcomponent(cfg, comp)
 or as
    [data] = ft_rejectcomponent(cfg, comp, data)

 where the input comp is the result of FT_COMPONENTANALYSIS. The output
 data will have the same format as the output of FT_PREPROCESSING.

 An optional input argument data can be provided. In that case
 componentanalysis will do a subspace projection of the input data
 onto the space which is spanned by the topographies in the unmixing
 matrix in comp, after removal of the artifact components.  Please use
 this option of including data as input, if you wish to use the output
 data.grad in further computation, for example for leadfield computation.

 The configuration structure can contain
   cfg.component  = list of components to remove, e.g. [1 4 7] or see FT_CHANNELSELECTION
   cfg.demean     = 'no' or 'yes', whether to demean the input data (default = 'yes')
   cfg.updatesens = 'no' or 'yes' (default = 'yes')

 To facilitate data-handling and distributed computing you can use
   cfg.inputfile   =  ...
   cfg.outputfile  =  ...
 If you specify one of these (or both) the input data will be read from a *.mat
 file on disk and/or the output data will be written to a *.mat file. These mat
 files should contain only a single variable, corresponding with the
 input/output structure.