Authored by Anthony Boyer

Fix conversion error: "in_fread_edf.m" was the same function as in "in_fopen_edf…

…"; add "struct_fix_events.m" for FRE patients.
function [sFile, ChannelMat, ImportOptions] = in_fopen_edf(DataFile, ImportOptions)
% IN_FOPEN_EDF: Open a BDF/EDF file (continuous recordings)
function F = in_fread_edf(sFile, sfid, SamplesBounds, ChannelsRange)
% IN_FREAD_EDF: Read a block of recordings from a EDF/BDF file
%
% USAGE: [sFile, ChannelMat, ImportOptions] = in_fopen_edf(DataFile, ImportOptions)
% USAGE: F = in_fread_edf(sFile, sfid, SamplesBounds, ChannelsRange)
% F = in_fread_edf(sFile, sfid, SamplesBounds) : Read all channels
% F = in_fread_edf(sFile, sfid) : Read all channels, all the times
% @=============================================================================
% This function is part of the Brainstorm software:
% https://neuroimage.usc.edu/brainstorm
%
% Copyright (c)2000-2020 University of Southern California & McGill University
% Copyright (c)2000-2019 University of Southern California & McGill University
% This software is distributed under the terms of the GNU General Public License
% as published by the Free Software Foundation. Further details on the GPLv3
% license can be found at http://www.gnu.org/copyleft/gpl.html.
... ... @@ -21,477 +23,203 @@ function [sFile, ChannelMat, ImportOptions] = in_fopen_edf(DataFile, ImportOptio
% For more information type "brainstorm license" at command prompt.
% =============================================================================@
%
% Authors: Francois Tadel, 2012-2019
% Parse inputs
if (nargin < 2) || isempty(ImportOptions)
ImportOptions = db_template('ImportOptions');
end
%% ===== READ HEADER =====
% Open file
fid = fopen(DataFile, 'r', 'ieee-le');
if (fid == -1)
error('Could not open file');
end
% Read all fields
hdr.version = fread(fid, [1 8], 'uint8=>char'); % Version of this data format ('0 ' for EDF, [255 'BIOSEMI'] for BDF)
hdr.patient_id = fread(fid, [1 80], '*char'); % Local patient identification
hdr.rec_id = fread(fid, [1 80], '*char'); % Local recording identification
hdr.startdate = fread(fid, [1 8], '*char'); % Startdate of recording (dd.mm.yy)
hdr.starttime = fread(fid, [1 8], '*char'); % Starttime of recording (hh.mm.ss)
hdr.hdrlen = str2double(fread(fid, [1 8], '*char')); % Number of bytes in header record
hdr.unknown1 = fread(fid, [1 44], '*char'); % Reserved ('24BIT' for BDF)
hdr.nrec = str2double(fread(fid, [1 8], '*char')); % Number of data records (-1 if unknown)
hdr.reclen = str2double(fread(fid, [1 8], '*char')); % Duration of a data record, in seconds
hdr.nsignal = str2double(fread(fid, [1 4], '*char')); % Number of signals in data record
% Check file integrity
if isnan(hdr.nsignal) || isempty(hdr.nsignal) || (hdr.nsignal ~= round(hdr.nsignal)) || (hdr.nsignal < 0)
error('File header is corrupted.');
end
% Read values for each nsignal
for i = 1:hdr.nsignal
hdr.signal(i).label = strtrim(fread(fid, [1 16], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).type = strtrim(fread(fid, [1 80], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).unit = strtrim(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).physical_min = str2double(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).physical_max = str2double(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).digital_min = str2double(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).digital_max = str2double(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).filters = strtrim(fread(fid, [1 80], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).nsamples = str2num(fread(fid, [1 8], '*char'));
end
for i = 1:hdr.nsignal
hdr.signal(i).unknown2 = fread(fid, [1 32], '*char');
end
% Unknown record size, determine correct nrec
if (hdr.nrec == -1)
datapos = ftell(fid);
fseek(fid, 0, 'eof');
endpos = ftell(fid);
hdr.nrec = floor((endpos - datapos) / (sum([hdr.signal.nsamples]) * 2));
end
% Close file
fclose(fid);
%% ===== RECONSTRUCT INFO =====
% Individual signal gain
for i = 1:hdr.nsignal
% Interpet units
switch (hdr.signal(i).unit)
case 'mV', unit_gain = 1e3;
case {'uV', char([166 204 86])}, unit_gain = 1e6;
otherwise, unit_gain = 1;
end
% Check min/max values
if isempty(hdr.signal(i).digital_min) || isnan(hdr.signal(i).digital_min)
disp(['EDF> Warning: The digitial minimum is not set for channel "' hdr.signal(i).label '".']);
hdr.signal(i).digital_min = -2^15;
end
if isempty(hdr.signal(i).digital_max) || isnan(hdr.signal(i).digital_max)
disp(['EDF> Warning: The digitial maximum is not set for channel "' hdr.signal(i).label '".']);
hdr.signal(i).digital_max = -2^15;
end
if isempty(hdr.signal(i).physical_min) || isnan(hdr.signal(i).physical_min)
disp(['EDF> Warning: The physical minimum is not set for channel "' hdr.signal(i).label '".']);
hdr.signal(i).physical_min = hdr.signal(i).digital_min;
end
if isempty(hdr.signal(i).physical_max) || isnan(hdr.signal(i).physical_max)
disp(['EDF> Warning: The physical maximum is not set for channel "' hdr.signal(i).label '".']);
hdr.signal(i).physical_max = hdr.signal(i).digital_max;
end
if (hdr.signal(i).physical_min >= hdr.signal(i).physical_max)
disp(['EDF> Warning: Physical maximum larger than minimum for channel "' hdr.signal(i).label '".']);
hdr.signal(i).physical_min = hdr.signal(i).digital_min;
hdr.signal(i).physical_max = hdr.signal(i).digital_max;
end
% Calculate and save channel gain
hdr.signal(i).gain = unit_gain ./ (hdr.signal(i).physical_max - hdr.signal(i).physical_min) .* (hdr.signal(i).digital_max - hdr.signal(i).digital_min);
hdr.signal(i).offset = hdr.signal(i).physical_min ./ unit_gain - hdr.signal(i).digital_min ./ hdr.signal(i).gain;
% Error: The number of samples is not specified
if isempty(hdr.signal(i).nsamples)
% If it is not the first electrode: try to use the previous one
if (i > 1)
disp(['EDF> Warning: The number of samples is not specified for channel "' hdr.signal(i).label '".']);
hdr.signal(i).nsamples = hdr.signal(i-1).nsamples;
else
error(['The number of samples is not specified for channel "' hdr.signal(i).label '".']);
end
end
hdr.signal(i).sfreq = hdr.signal(i).nsamples ./ hdr.reclen;
end
% Find annotations channel
iAnnotChans = find(strcmpi({hdr.signal.label}, 'EDF Annotations')); % Mutliple "EDF Annotation" channels allowed in EDF+
iStatusChan = find(strcmpi({hdr.signal.label}, 'Status'), 1); % Only one "Status" channel allowed in BDF
iOtherChan = setdiff(1:hdr.nsignal, [iAnnotChans iStatusChan]);
% % Remove channels with lower sampling rates
% iIgnoreChan = find([hdr.signal(iOtherChan).sfreq] < max([hdr.signal(iOtherChan).sfreq])); % Ignore all the channels with lower sampling rate
% if ~isempty(iIgnoreChan)
% iOtherChan = setdiff(iOtherChan, iIgnoreChan);
% end
% Get all the other channels
if isempty(iOtherChan)
error('This file does not contain any data channel.');
end
% Read events preferencially from the EDF Annotations track
if ~isempty(iAnnotChans)
iEvtChans = iAnnotChans;
elseif ~isempty(iStatusChan)
iEvtChans = iStatusChan;
% Authors: Francois Tadel, 2012-2017
%% ===== PARSE INPUTS =====
nChannels = sFile.header.nsignal;
iChanAnnot = find(strcmpi({sFile.header.signal.label}, 'EDF Annotations'));
iBadChan = find(sFile.channelflag == -1);
iChanSignal = setdiff(1:nChannels, iChanAnnot);
iChanWrongRate = find([sFile.header.signal(iChanSignal).sfreq] ~= max([sFile.header.signal(setdiff(iChanSignal,iBadChan)).sfreq]));
iChanSkip = union(iChanAnnot, iChanWrongRate);
if (nargin < 4) || isempty(ChannelsRange)
ChannelsRange = [1, nChannels];
end
if (nargin < 3) || isempty(SamplesBounds)
SamplesBounds = [0, sFile.header.nrec * sFile.header.signal(ChannelsRange(1)).nsamples - 1];
end
nTimes = sFile.header.reclen * sFile.header.signal(ChannelsRange(1)).sfreq;
iTimes = SamplesBounds(1):SamplesBounds(2);
% Block of times/channels to extract
nReadChannels = double(ChannelsRange(2) - ChannelsRange(1) + 1);
% Read annotations instead of real data ?
isAnnotOnly = ~isempty(iChanAnnot) && (ChannelsRange(1) == ChannelsRange(2)) && ismember(ChannelsRange(1), iChanAnnot);
isBdfStatus = strcmpi(sFile.format, 'EEG-BDF') && (ChannelsRange(1) == ChannelsRange(2)) && strcmpi(sFile.header.signal(ChannelsRange(1)).label, 'Status');
% Data channels to read
if isAnnotOnly
% Read only on annotation channel
iChanF = 1;
else
iEvtChans = [];
end
% % Detect channels with inconsistent sampling frenquency
% iErrChan = find([hdr.signal(iOtherChan).sfreq] ~= hdr.signal(iOtherChan(1)).sfreq);
% iErrChan = setdiff(iErrChan, iAnnotChans);
% if ~isempty(iErrChan)
% error('Files with mixed sampling rates are not supported yet.');
% Remove all the annotation channels from the list of channels to read
iChanF = setdiff(ChannelsRange(1):ChannelsRange(2), iChanSkip) - ChannelsRange(1) + 1;
% if any(diff(iChanF) ~= 1)
% error('All the data channels to read from the file must be contiguous (EDF Annotation channels must be at the end of the list).');
% end
% Detect interrupted signals (time non-linear)
hdr.interrupted = ischar(hdr.unknown1) && (length(hdr.unknown1) >= 5) && isequal(hdr.unknown1(1:5), 'EDF+D');
if hdr.interrupted
if ImportOptions.DisplayMessages
[res, isCancel] = java_dialog('question', ...
['Interrupted EDF file ("EDF+D") detected. It is recommended to convert it' 10 ...
'to a continuous ("EDF+C") file first. Do you want to continue reading this' 10 ...
'file as continuous and attempt to fix the timing of event markers?' 10 ...
'NOTE: This may not work as intended, use at your own risk!']);
hdr.fixinterrupted = ~isCancel && strcmpi(res, 'yes');
if any(diff(iChanF) ~= 1)
iChanLast = find(diff(iChanF) ~= 1, 1);
iChanF = iChanF(1:iChanLast);
else
hdr.fixinterrupted = 1;
end
if ~hdr.fixinterrupted
warning(['Interrupted EDF file ("EDF+D"): requires conversion to "EDF+C". ' 10 ...
'Brainstorm will read this file as a continuous file ("EDF+C"), the timing of the samples after the first discontinuity will be wrong.' 10 ...
'This may not cause any major problem unless there are time markers in the file, they might be inaccurate in all the segments >= 2.']);
end
else
hdr.fixinterrupted = 0;
end
%% ===== CREATE BRAINSTORM SFILE STRUCTURE =====
% Initialize returned file structure
sFile = db_template('sfile');
% Add information read from header
sFile.byteorder = 'l';
sFile.filename = DataFile;
if (uint8(hdr.version(1)) == uint8(255))
sFile.format = 'EEG-BDF';
sFile.device = 'BDF';
else
sFile.format = 'EEG-EDF';
sFile.device = 'EDF';
end
sFile.header = hdr;
% Comment: short filename
[tmp__, sFile.comment, tmp__] = bst_fileparts(DataFile);
% No info on bad channels
sFile.channelflag = ones(hdr.nsignal,1);
% Acquisition date
sFile.acq_date = str_date(hdr.startdate);
%% ===== PROCESS CHANNEL NAMES/TYPES =====
% Try to split the channel names in "TYPE NAME"
SplitType = repmat({''}, 1, hdr.nsignal);
SplitName = repmat({''}, 1, hdr.nsignal);
for i = 1:hdr.nsignal
% Removing trailing dots (eg. "Fc5." instead of "FC5", as in: https://www.physionet.org/pn4/eegmmidb/)
if (hdr.signal(i).label(end) == '.') && (length(hdr.signal(i).label) > 1)
hdr.signal(i).label(end) = [];
if (hdr.signal(i).label(end) == '.') && (length(hdr.signal(i).label) > 1)
hdr.signal(i).label(end) = [];
if (hdr.signal(i).label(end) == '.') && (length(hdr.signal(i).label) > 1)
hdr.signal(i).label(end) = [];
end
end
end
% Remove extra spaces
signalLabel = strrep(hdr.signal(i).label, ' - ', '-');
% Find space chars (label format "Type Name")
iSpace = find(signalLabel == ' ');
% Only if there is one space only
if (length(iSpace) == 1) && (iSpace >= 3)
SplitName{i} = signalLabel(iSpace+1:end);
SplitType{i} = signalLabel(1:iSpace-1);
% Accept also 2 spaces
elseif (length(iSpace) == 2) && (iSpace(1) >= 3)
SplitName{i} = strrep(signalLabel(iSpace(1)+1:end), ' ', '_');
SplitType{i} = signalLabel(1:iSpace(1)-1);
iChanLast = length(iChanF);
end
ChannelsRange = [iChanF(1), iChanF(iChanLast)] + ChannelsRange(1) - 1;
end
% Remove the classification if it makes some names non unique
uniqueNames = unique(SplitName);
for i = 1:length(uniqueNames)
if ~isempty(uniqueNames{i})
iName = find(strcmpi(SplitName, uniqueNames{i}));
if (length(iName) > 1)
[SplitName{iName}] = deal('');
[SplitType{iName}] = deal('');
end
% Cannot read channels with different sampling rates at the same time
if (ChannelsRange(1) ~= ChannelsRange(2))
allFreq = [sFile.header.signal(ChannelsRange(1):ChannelsRange(2)).nsamples];
if any(allFreq ~= allFreq(1))
error(['Cannot read channels with mixed sampling rates at the same time.' 10 ...
'Mark as bad channels with different sampling rates than EEG.' 10 ...
'(right-click on data file > Good/bad channels > Edit good/bad channels)']);
end
end
%% ===== CREATE EMPTY CHANNEL FILE =====
ChannelMat = db_template('channelmat');
ChannelMat.Comment = [sFile.device ' channels'];
ChannelMat.Channel = repmat(db_template('channeldesc'), [1, hdr.nsignal]);
chRef = {};
% For each channel
for i = 1:hdr.nsignal
% If is the annotation channel
if ~isempty(iAnnotChans) && ismember(i, iAnnotChans)
ChannelMat.Channel(i).Type = 'EDF';
ChannelMat.Channel(i).Name = 'Annotations';
elseif ~isempty(iStatusChan) && (i == iStatusChan)
ChannelMat.Channel(i).Type = 'BDF';
ChannelMat.Channel(i).Name = 'Status';
% Regular channels
else
% If there is a pair name/type already detected
if ~isempty(SplitName{i}) && ~isempty(SplitType{i})
ChannelMat.Channel(i).Name = SplitName{i};
ChannelMat.Channel(i).Type = SplitType{i};
else
% Channel name
ChannelMat.Channel(i).Name = hdr.signal(i).label(hdr.signal(i).label ~= ' ');
% Channel type
if ~isempty(hdr.signal(i).type)
if (length(hdr.signal(i).type) == 3)
ChannelMat.Channel(i).Type = hdr.signal(i).type(hdr.signal(i).type ~= ' ');
elseif isequal(hdr.signal(i).type, 'Active Electrode') || isequal(hdr.signal(i).type, 'AgAgCl electrode')
ChannelMat.Channel(i).Type = 'EEG';
else
ChannelMat.Channel(i).Type = 'Misc';
end
%% ===== READ ALL NEEDED EPOCHS =====
% Detect which epochs are necessary for the range of data selected
epochRange = floor(SamplesBounds ./ nTimes);
epochsToRead = epochRange(1) : epochRange(2);
% Initialize block of data to read
if isAnnotOnly
F = zeros(nReadChannels, 2 * length(iTimes));
else
F = zeros(nReadChannels, length(iTimes));
end
% Marker that we increment when we add data to F
iF = 1;
% Read all the needed epochs
for i = 1:length(epochsToRead)
% Find the samples to read from this epoch
BoundsEpoch = nTimes * epochsToRead(i) + [0, nTimes-1];
BoundsRead = [max(BoundsEpoch(1), iTimes(1)), ...
min(BoundsEpoch(2), iTimes(end))];
iTimeRead = BoundsRead(1):BoundsRead(2);
% Convert this samples into indices in this very epoch
iTimeRead = iTimeRead - nTimes * epochsToRead(i);
% New indices to read
if isAnnotOnly
iNewF = iF:(iF + 2*length(iTimeRead) - 1);
else
ChannelMat.Channel(i).Type = 'EEG';
end
end
% Extract reference name (at the end of the channel name, separated with a "-", eg. "-REF")
iDash = find(ChannelMat.Channel(i).Name == '-');
if ~isempty(iDash) && (iDash(end) < length(ChannelMat.Channel(i).Name))
chRef{end+1} = ChannelMat.Channel(i).Name(iDash(end):end);
end
end
ChannelMat.Channel(i).Loc = [0; 0; 0];
ChannelMat.Channel(i).Orient = [];
ChannelMat.Channel(i).Weight = 1;
% ChannelMat.Channel(i).Comment = hdr.signal(i).type;
end
% If the same reference is indicated for all the channels: remove it
if (length(chRef) >= 2)
% Get the shortest reference tag
lenRef = cellfun(@length, chRef);
minLen = min(lenRef);
% Check if all the ref names are equal (up to the max length - some might be cut because the channel name is too long)
if all(cellfun(@(c)strcmpi(c(1:minLen), chRef{1}(1:minLen)), chRef))
% Remove the reference tag from all the channel names
for i = 1:length(ChannelMat.Channel)
ChannelMat.Channel(i).Name = strrep(ChannelMat.Channel(i).Name, chRef{1}, '');
ChannelMat.Channel(i).Name = strrep(ChannelMat.Channel(i).Name, chRef{1}(1:minLen), '');
end
iNewF = iF:(iF + length(iTimeRead) - 1);
end
% Read epoch (full or partial)
F(iChanF,iNewF) = edf_read_epoch(sFile, sfid, epochsToRead(i), iTimeRead, ChannelsRange, isAnnotOnly, isBdfStatus);
% Increment marker
iF = iF + length(iTimeRead);
end
% If there are only "Misc" and no "EEG" channels: rename to "EEG"
iMisc = find(strcmpi({ChannelMat.Channel.Type}, 'Misc'));
iEeg = find(strcmpi({ChannelMat.Channel.Type}, 'EEG'));
if ~isempty(iMisc) && isempty(iEeg)
[ChannelMat.Channel(iMisc).Type] = deal('EEG');
iEeg = iMisc;
end
%% ===== DETECT MULTIPLE SAMPLING RATES =====
% Use the first "EEG" channel as the reference sampling rate (or the first channel if no "EEG" channels available)
if ~isempty(iEeg) && ismember(iEeg(1), iOtherChan)
iChanFreqRef = iEeg(1);
else
iChanFreqRef = iOtherChan(1);
end
% Mark as bad channels with sampling rates different from EEG
iChanWrongRate = find([sFile.header.signal.sfreq] ~= sFile.header.signal(iChanFreqRef).sfreq);
iChanWrongRate = intersect(iChanWrongRate, iOtherChan);
if ~isempty(iChanWrongRate)
sFile.channelflag(iChanWrongRate) = -1;
end
% Consider that the sampling rate of the file is the sampling rate of the first signal
sFile.prop.sfreq = hdr.signal(iChanFreqRef).sfreq;
sFile.prop.times = [0, hdr.signal(iChanFreqRef).nsamples * hdr.nrec - 1] ./ sFile.prop.sfreq;
sFile.prop.nAvg = 1;
%% ===== READ EDF ANNOTATION CHANNEL =====
if ~isempty(iEvtChans) % && ~isequal(ImportOptions.EventsMode, 'ignore')
% Set reading options
ImportOptions.ImportMode = 'Time';
ImportOptions.UseSsp = 0;
ImportOptions.UseCtfComp = 0;
% Read EDF annotations
if strcmpi(sFile.format, 'EEG-EDF')
evtList = {};
% In EDF+, the first annotation channel has epoch time stamps (EDF
% calls epochs records). So read all annotation channels per epoch.
for irec = 1:hdr.nrec
for ichan = 1:length(iEvtChans)
bst_progress('text', sprintf('Reading annotations... [%d%%]', round((ichan + (irec-1)*length(iEvtChans))/length(iEvtChans)/hdr.nrec*100)));
% Sample indices for the current epoch (=record)
SampleBounds = [irec-1,irec] * sFile.header.signal(iEvtChans(ichan)).nsamples - [0,1];
% Read record
F = in_fread(sFile, ChannelMat, 1, SampleBounds, iEvtChans(ichan), ImportOptions);
% Find all the TALs (Time-stamped Annotations Lists): separated with [20][0]
% Possible configurations:
% Onset[21]Duration[20]Annot1[20]Annot2[20]..AnnotN[20][0]
% Onset[20]Annot1[20]Annot2[20]..AnnotN[20][0]
% Onset[20]Annot1[20][0]
iSeparator = [-1, find((F(1:end-1) == 20) & (F(2:end) == 0))];
% Loop on all the TALs
for iAnnot = 1:length(iSeparator)-1
% Get annotation
strAnnot = char(F(iSeparator(iAnnot)+2:iSeparator(iAnnot+1)-1));
% Split in blocks with [20]
splitAnnot = str_split(strAnnot, 20);
% The first TAL in a record should be indicating the timing of the first sample of the record
if (iAnnot == 1) && (length(splitAnnot) == 1) && ~any(splitAnnot{1} == 21)
% Ignore if this is not the first channel
if (ichan > 1)
continue;
end
% Get record time stamp
t0_rec = str2double(splitAnnot{1});
if isempty(t0_rec) || isnan(t0_rec)
continue;
%% ===== READ ONE EPOCH =====
function F = edf_read_epoch(sFile, sfid, iEpoch, iTimes, ChannelsRange, isAnnotOnly, isBdfStatus)
% ===== COMPUTE OFFSETS =====
nTimes = sFile.header.reclen * [sFile.header.signal.sfreq];
nReadTimes = length(iTimes);
nReadChannels = double(ChannelsRange(2) - ChannelsRange(1) + 1);
iChannels = ChannelsRange(1):ChannelsRange(2);
% Check that all the channels selected have the same freq rate
if any(nTimes(iChannels) ~= nTimes(iChannels(1)))
error('Cannot read at the same signals with different sampling frequency.');
end
if (irec == 1)
t0_file = t0_rec;
% Find discontinuities larger than 1 sample
elseif abs(t0_rec - prev_rec - (irec - prev_irec) * hdr.reclen) > (1 / sFile.prop.sfreq)
% Brainstorm fills partial/interrupted records with zeros
bstTime = prev_rec + hdr.reclen;
timeDiff = bstTime - t0_rec;
% If we want to fix timing, apply skip to initial timestamp
if hdr.fixinterrupted
t0_file = t0_file - timeDiff;
end
% Warn user of discontinuity
if timeDiff > 0
expectMsg = 'blank data';
else
expectMsg = 'skipped data';
end
startTime = min(t0_rec - t0_file - [0, timeDiff]); % before and after t0_file adjustment
endTime = max(t0_rec - t0_file - [0, timeDiff]);
fprintf('WARNING: Found discontinuity between %.3fs and %.3fs, expect %s in between.\n', startTime, endTime, expectMsg);
% Create event for users information
if timeDiff < 0
endTime = startTime; % no extent in this case, there is skipped time.
end
evtList(end+1,:) = {'EDF+D Discontinuity', [startTime; endTime]};
end
prev_rec = t0_rec;
prev_irec = irec;
% Regular TAL: indicating a marker
% Size of one value
if strcmpi(sFile.format, 'EEG-BDF')
% BDF: int24 => 3 bytes
bytesPerVal = 3;
% Reading status or regular channel
if isBdfStatus
dataClass = 'ubit24';
else
% Split time in onset/duration
splitTime = str_split(splitAnnot{1}, 21);
% Get time
t = str2double(splitTime{1});
if isempty(t) || isnan(t)
continue;
end
% Get duration
if (length(splitTime) > 1)
duration = str2double(splitTime{2});
% Exclude 1-sample long events
if isempty(duration) || isnan(duration) || (round(duration .* sFile.prop.sfreq) <= 1)
duration = 0;
dataClass = 'bit24';
end
else
duration = 0;
end
% Unnamed event
if (length(splitAnnot) == 1) || isempty(splitAnnot{2})
splitAnnot{2} = 'Unnamed';
end
% Create one event for each label in the TAL
for iLabel = 2:length(splitAnnot)
evtList(end+1,:) = {splitAnnot{iLabel}, (t-t0_file) + [0;duration]};
end
end
end
end
end
% If there are events: create a create an events structure
if ~isempty(evtList)
% Initialize events list
sFile.events = repmat(db_template('event'), 0);
% Events list
[uniqueEvt, iUnique] = unique(evtList(:,1));
uniqueEvt = evtList(sort(iUnique),1);
% Build events list
for iEvt = 1:length(uniqueEvt)
% Find all the occurrences of this event
iOcc = find(strcmpi(uniqueEvt{iEvt}, evtList(:,1)));
% Concatenate all times
t = [evtList{iOcc,2}];
% If second row is equal to the first one (no extended events): delete it
if all(t(1,:) == t(2,:))
t = t(1,:);
end
% Set event
sFile.events(iEvt).label = strtrim(uniqueEvt{iEvt});
sFile.events(iEvt).times = round(t .* sFile.prop.sfreq) ./ sFile.prop.sfreq;
sFile.events(iEvt).epochs = 1 + 0*t(1,:);
sFile.events(iEvt).select = 1;
sFile.events(iEvt).channels = cell(1, size(sFile.events(iEvt).times, 2));
sFile.events(iEvt).notes = cell(1, size(sFile.events(iEvt).times, 2));
end
end
% BDF Status line
elseif strcmpi(sFile.format, 'EEG-BDF') && ~strcmpi(ImportOptions.EventsTrackMode, 'ignore')
% Ask how to read the events
[events, ImportOptions.EventsTrackMode] = process_evt_read('Compute', sFile, ChannelMat, ChannelMat.Channel(iEvtChans).Name, ImportOptions.EventsTrackMode);
if isequal(events, -1)
sFile = [];
ChannelMat = [];
return;
end
% Report the events in the file structure
sFile.events = events;
% Remove the 'Status: ' string in front of the events
for i = 1:length(sFile.events)
sFile.events(i).label = strrep(sFile.events(i).label, 'Status: ', '');
end
% Group events by time
% sFile.events = process_evt_grouptime('Compute', sFile.events);
% EDF: int16 => 2 bytes
bytesPerVal = 2;
dataClass = 'int16';
isBdfStatus = 0;
end
% Offset of the beginning of the recordings in the file
offsetHeader = round(sFile.header.hdrlen);
% Offset of epoch
offsetEpoch = round(iEpoch * sum(nTimes) * bytesPerVal);
% Channel offset
offsetChannel = round(sum(nTimes(1:ChannelsRange(1)-1)) * bytesPerVal);
% Time offset at the beginning and end of each channel block
offsetTimeStart = round(iTimes(1) * bytesPerVal);
offsetTimeEnd = round((nTimes(ChannelsRange(1)) - iTimes(end) - 1) * bytesPerVal);
% ALL THE "ROUND" CALLS WERE ADDED AFTER DISCOVERING THAT THERE WERE SOMETIMES ROUNDING ERRORS IN THE MULTIPLICATIONS
% Where to start reading in the file ?
% => After the header, the number of skipped epochs, channels and time samples
offsetStart = offsetHeader + offsetEpoch + offsetChannel + offsetTimeStart;
% Number of time samples to skip after each channel
offsetSkip = offsetTimeStart + offsetTimeEnd;
% ===== READ DATA BLOCK =====
% Position file at the beginning of the trial
fseek(sfid, offsetStart, 'bof');
% Read annotation data (char)
if isAnnotOnly
dataClass = 'char';
nReadTimes = bytesPerVal * nReadTimes; % 1 byte instead of 2
end
% Read trial data
% => WARNING: CALL TO FREAD WITH SKIP=0 DOES NOT WORK PROPERLY
if (offsetSkip == 0)
F = fread(sfid, [nReadTimes, nReadChannels], dataClass)';
elseif (bytesPerVal == 2)
precision = sprintf('%d*%s', nReadTimes, dataClass);
F = fread(sfid, [nReadTimes, nReadChannels], precision, offsetSkip)';
% => WARNING: READING USING ubit24 SOMETIMES DOESNT WORK => DOING IT MANUALLY
elseif (bytesPerVal == 3)
% Reading each bit independently
precision = sprintf('%d*%s', 3*nReadTimes, 'uint8');
F = fread(sfid, [3*nReadTimes, nReadChannels], precision, offsetSkip)';
% Grouping the 3 bits together
F = F(:,1:3:end) + F(:,2:3:end)*256 + F(:,3:3:end)*256*256;
% 2-Complement (negative value indicated by most significant bit)
if strcmpi(dataClass, 'bit24')
iNeg = (F >= 256*256*128);
F(iNeg) = F(iNeg) - 256*256*256;
end
end
% Check that data block was fully read
if (numel(F) < nReadTimes * nReadChannels)
% Number of full time samples that were read
nTimeTrunc = max(0, floor(numel(F) / nReadChannels) - 1);
% Error message
disp(sprintf('EDF> ERROR: File is truncated (%d time samples were read instead of %d). Padding with zeros...', nTimeTrunc, nReadTimes));
% Pad data with zeros
Ftmp = zeros(nReadTimes, nReadChannels);
F = F';
Ftmp(1:numel(F)) = F(:);
F = Ftmp';
end
% Processing for BDF status file
if isBdfStatus
% Mask to keep only the first 15 bits (Triggers bits)
% Bit 16 : High when new Epoch is started
% Bit 17-19 : Speed bits 0 1 2
% Bit 20 : High when CMS is within range
% Bit 21 : Speed bit 3
% Bit 22 : High when battery is low
% Bit 23 : High if ActiveTwo MK2
F = bitand(F, bin2dec('000000000111111111111111'));
% Processing for real data
elseif ~isAnnotOnly
% Convert to double
F = double(F);
% Apply gains
F = bst_bsxfun(@rdivide, F, [sFile.header.signal(iChannels).gain]');
% IN THEORY: THIS OFFSET SECTION IS USEFUL, BUT IN PRACTICE IT LOOKS LIKE THE VALUES IN ALL THE FILES ARE CENTERED ON ZERO
% % Add offset
% if isfield(sFile.header.signal, 'offset') && ~isempty(sFile.header.signal(1).offset)
% % ...
% end
end
end
... ...