Tags: example dataformat ctf meg

This is a very old and outdated example page, the referenced code does not exist anymore

Can I create an artificial CTF dataset using MATLAB?

Q: Is there a way to create an artificial CTF dataset using MATLAB or one of the CTF utilities, or alternatively take a real dataset and replace just the channel data with my simulated data?

A: Yes, there is, see below.

Part of function ft_write_data, case ctf_meg4

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% function ft_write_data(filename, dat, varargin)

% FT_WRITE_DATA exports electrophysiological data to a file
%
% Use as ft_write_data(filename, dat, ...)
%
% The specified filename can already contain the filename extention,
% but that is not required since it will be added automatically.
%
% Additional options should be specified in key-value pairs and can be
% 'header' header structure, see READ_FCDC_HEADER
% 'dataformat' string, see below
% 'chanindx' 1xN array

% ...

% get the options
dataformat = keyval('dataformat', varargin); if isempty(dataformat), dataformat = ft_filetype(filename); end
chanindx = keyval('chanindx', varargin);
hdr = keyval('header', varargin);

% determine the data size (input is dataformat as it is returned by preprocessing, should be in ntrials x nchannels x nsamples)
[nchans, nsamples] = size(dat.trial{1});

switch dataformat
case 'ctf_ds'
id = find(filename == '.');
filename = [filename '/' filename(1:id) 'meg4'];
dataformat = ft_filetype(filename);

    case 'ctf_meg4'

        % reshaping the data size
        if isfield(dat, 'trial'),
           for i = 1:size(dat.trial,2)
              data(i,:,:) = dat.trial{i};
           end
        end

        ntrldat   = size(data,1);
        nsmpdat   = size(data,3);
        nchandat  = size(data,2);
        ntrlorig  = hdr.nTrials;
        nchanorig = hdr.nChans;
        nsmporig = hdr.nSamples;

        % In case more trials/channels/samples are present in data compared to
        % header, error message. No fix.
        if ntrldat>ntrlorig
            error('More trials in dataset than header indicates');
        elseif nchandat>nchanorig
            error('More channels in dataset than header indicates');
        elseif nsmpdat>nsmporig
            error('More samples in dataset than header indicates');
        end

        % compare label to orig.label
        [count, index] = vec_match(dat.label,hdr.label);
        if find(index == 0)
           error('Channelnames do not match');
        end

        % If the dat.label contains the same channels (no more), or a
        % subset of existing channels of orig.label, the order of the channels
        % should be checked, and if the nr of channels in the new dataset is
        % smaller than in the header, the rest will be zeropadded.

        NZ_index = nonzeros(index);
        AV = check_asc(NZ_index);
        if length(NZ_index) == length(hdr.label) && AZ == 1
            return
        else
            NewLabel = cell(length(hdr.label),1);
            NewDat = zeros(ntrlorig, nchanorig, nsmpdat);
            for i = 1:length(NZ_index)
               NewLabel(NZ_index(i),:) = dat.label(i);
               NewDat(1:ntrldat,NZ_index(i),:) = data(:,i,:);
            end
            if count < length(hdr.label)
                fprintf('Less channels in dataset than header indicates, data padded with zeros');
            end
        end

        dat.label = NewLabel;
        data = NewDat;

        % Since the size of dat may have changed, new values for trl,smp and
        % chan should be defined
        ntrldat   = size(data,1);
        nsmpdat   = size(data,3);
        nchandat  = size(data,2);

        % Opening the filename and writing the header
        fid = fopen(filename, 'wb', 'ieee-be');
        buf = [77 69 71 52 49 67 80 0]; % 'MEG41CP', null-terminated
        fwrite(fid, buf, 'char');

        if isempty(chanindx) && length(hdr.label)==nchandat
           chanindx = 1:nchandat;
        end

        % Check number of bytes needed to write the data
        nr_bytes = 8 + (ntrldat*nsmpdat*nchandat*4);
        if nr_bytes > 1e9
           error ('this code is not capable of writing datasets larger than 1Gb')
        end

        for i=1:ntrldat
           datorig = zeros(nchanorig,nsmporig); % size 'original' = header data
           if i<=ntrldat
             datorig(chanindx,1:nsmpdat) = data(i,:,:); % Padding with zeros in case nsmpdat<nsmporig
             datorig = transpose(sparse(diag(1./hdr.orig.gainV)) * datorig);
             if any(datorig(:)>intmax('int32')) || any(datorig(:)<intmin('int32'))
                warning('Data reaches limits int32, border values are clipped');
                datorig(datorig>intmax('int32')) = intmax('int32');
                datorig(datorig<intmin('int32')) = intmin('int32');
             end
             datorig = int32(datorig);
           else
              datorig = int32(transpose(datorig));
           end
           fwrite(fid, datorig, 'int32');
        end

        fclose(fid);

    end % switch dataformat

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Function vec_match
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function [count, index] = vec_match(vec1, vec2)

count = 0;
for i = 1:size(vec1,1)
id = strmatch(vec1(i,:), vec2);
if size(id,1) == 0
continue
else
count = count + 1;
index(i) = id;
end
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Function check_asc
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function AZ = check_asc(index)

max = 0;
for i = 1:length(index)
if max < index(i)
max = index(i);
AZ = 1;
else
AZ = 0;
break;
end
end