Issue with BlinkIndexMetric
cmmrandau opened this issue · comments
Hello Neil Bailey,
I'm getting this error with some datafiles, just a few out of 100's. I'm attaching the RELAC settings at the end of this message.
Unrecognized function or variable 'BlinkIndexMetric'.
Error in RELAX_blinks_IQR_method (line 84)
ix_blinkstart=find(diff(BlinkIndexMetric)==1)+1; % indices where BlinkIndexMetric goes from 0 to 1
Error in RELAX_excluding_extreme_values (line 68)
[continuousEEG_initialblinkmarking, ~] = RELAX_blinks_IQR_method(continuousEEG, epochedEEG, RELAX_cfg);
Error in RELAX_Wrapper_beta (line 194)
[continuousEEG, epochedEEG] = RELAX_excluding_extreme_values(continuousEEG, epochedEEG, RELAX_cfg); % Mark extreme periods for exclusion from MWF cleaning, and deletion before wICA cleaning
Error in RELAX_SET_PARAMETERS_AND_RUN_beta (line 376)
RELAXProcessing_ICA_AllParticipants, RELAXProcessingRoundThreeAllParticipants, RELAX_issues_to_check, RELAXProcessingExtremeRejectionsAllParticipants] = RELAX_Wrapper_beta (RELAX_cfg);
RELAX settings:
%% RELAX EEG CLEANING PIPELINE, Copyright (C) (2022) Neil Bailey
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% any later version.
%
% This program is distributed in the hope that it will be useful,
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with this program. If not, see https://www.gnu.org/licenses/.
%% RELAX_SET_PARAMETERS_AND_RUN:
%% This pipeline firstly filters your data, then deletes bad electrodes and marks extremely bad periods for exclusion from further processing.
%% It then creates a mask of clean and artifact data specific to the length of your data file, which is submitted to a multiple weiner filter (MWF) in order to clean your data.
%% The MWF cleaned data is then submitted to an independent component analysis, and artifactual components detected by ICLabel are cleaned by wavelet enhanced ICA.
%% The resulting saved files are continuous, very well cleaned of artifacts, while still preserving the neural signal.
%% Note that the pipeline will only work on continuous data.
% RELAX is intended to be fully automated, using the most empirical approaches
% we could find in order to identify periods of EEG data containing
% artifacts, as well as removing channels that are either bad, or contain
% above a specified threshold for the proportion of epochs containing
% muscle artifacts.
% In order to use this script, do the following:
% 1) Install the dependencies listed below, and define the folder where
% they are installed (where required)
% 2) Load the "to be processed" files into a single folder, EEGLAB
% formatted (.set format), and specify this folder location in RELAX_cfg.myPath
% (these files should have triggers recoded to reflect participant accuracy
% if desirable, and not have been re-referenced or have any other
% pre-processing completed yet).
% 3) Specify the parameters in RELAX_cfg. These parameters let you determine
% a range of factors that influence the mask creation, eg. "how many
% epochs can show muscle activity in a channel before that channel is
% deleted" (note that a maximum of 20% of channel deletions are allowed
% after the PREP electrode deletion, "what is the maximum proportion of
% epochs that can be masked as an artifact for each type of artifact",
% "should I run the MWF twice or just once", and "what is the threshold
% of artifact severity, above which sections are marked as an artifact,
% below which they are not?".
% Additionally, particular artifact detection and masking functions
% could be excluded if you're not concerned about those artifacts by
% commenting that function out (adding a % symbol at the start of the
% line). For example, you might only want to include muscle, and
% drift artifacts in the MWF cleaning, and not blink and horizontal
% eye movements if you are analysing eyes closed resting data.
% 4)Click "Run" in the menu up the top of this script
% The script will then process all participants, taking ~10 min per participant.
% You can run just a subset of the participants first by altering the
% following line: "for FileNumber=1:numel(RELAX_cfg.files)"
% 5) The script will produce files that contain cleaning quality values for
% relevant processes in this script. Means, SD's (or medians and MADs)
% and ranges across all participants can be examined to determine
% the success of the script for the study as a whole, and to identify
% potential issues with specific files (eg. the artifact to residue ratio
% is ~1, suggesting artifacts weren't removed successfully, or >75% of the
% epochs contained muscle activity, suggesting bad data throughout). These
% statistics should be reported in publications.
% Note that it's particularly important to check for files that show rank
% deficiency in the MWF processing, as this could prevent the data from
% being cleaned. If this happens, I've found that adjusting the RELAX_cfg
% parameters can fix the problem. Particularly the amount of data that can
% be included in the mask (the mask needs to contain both enough artifact
% and enough clean data to create artifact and clean templates), and the
% minimum length of a marked patch of clean or artifact data (if
% patches are too short, the MWF can't create an adequate mask. The amount
% of data marked as artifact around eye blinks can particularly affect
% this).
% 6) The script will produce output EEG files that have been cleaned into a
% folder labelled 1xMWF for the first run, 2xMWF, and 3xMWF (if
% you select to save these). Cleaned data will be saved into the Cleaned_data
% folder. The cleaned files will have electrodes removed, and are referenced to the
% average reference. From this point, I would interpolate missing electrodes,
% re-code the task related triggers as to whether the participant
% was correct or not, epoch the data, reject epochs that still show artifacts
% (the epoch rejection wrapper could be adapted to achieve this)
% and separate different conditions into different files in preparation
% for analysis.
% Note that if you have very large files (or your computer doesn't have much RAM),
% the MWF cleaning may run out of RAM (and suggest a created array is too large).
% This could be addressed by downsampling the data, or reducing
% RELAX_cfg.MWFDelayPeriod=8; to 4-5. This will reduce the number of samples
% that the MWF cleaning takes into account when characterising artifacts,
% but still cleans very effectively.
% Note that the pipeline does not work well on MATLAB versions <2016,
% we suggest updating to a more recent version of MATLAB if your version is
% pre-2016.
%clear all; close all; clc;
%% DEPENDENCIES (toolboxes you need to install, and cite if you use this script):
% use fileseparators 'filesep' for increased compatability if necessary (replace the \ with a filesep [outside of quotes])
% MATLAB signal processing toolbox (from MATLAB website)
% MATLAB statistics and machine learning toolbox (from MATLAB website)
% EEGLAB:
% https://sccn.ucsd.edu/eeglab/index.php
% Delorme, A., & Makeig, S. (2004). EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of neuroscience methods, 134(1), 9-21.
addpath('/home/martin/Documents/MATLAB/eeglab-develop');
eeglab;
% PREP pipeline to reject bad electrodes (install plugin to EEGLAB, or via the github into the EEGLAB plugin folder):
% https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4471356/
% Bigdely-Shamlo, N., Mullen, T., Kothe, C., Su, K. M., & Robbins, K. A. (2015). The PREP pipeline: standardized preprocessing for large-scale EEG analysis. Frontiers in neuroinformatics, 9, 16.
% http://vislab.github.io/EEG-Clean-Tools/ or install via EEGLAB extensions
% Specify the MWF path:
% https://github.com/exporl/mwf-artifact-removal
% Somers, B., Francart, T., & Bertrand, A. (2018). A generic EEG artifact removal algorithm based on the multi-channel Wiener filter. Journal of neural engineering, 15(3), 036007.
addpath(genpath('/home/martin/Documents/MATLAB/eeglab-develop/plugins/mwf-artifact-removal-master/'));
% Fieldtrip:
% http://www.fieldtriptoolbox.org/
% Robert Oostenveld, Pascal Fries, Eric Maris, and Jan-Mathijs Schoffelen. FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data. Computational Intelligence and Neuroscience, vol. 2011, Article ID 156869, 9 pages, 2011. doi:10.1155/2011/156869.
addpath('/home/martin/Documents/MATLAB/eeglab-develop/plugins/Fieldtrip-lite20210601/');
% fastica:
% http://research.ics.aalto.fi/ica/fastica/code/dlcode.shtml
addpath('/home/martin/Documents/MATLAB/bin/FastICA_25/');
% ICLabel in your eeglab folder as a plugin or via the github:
% https://github.com/sccn/ICLabel
% Specify RELAX folder location (this toolbox):
addpath('/home/martin/Documents/MATLAB/eeglab-develop/plugins/RELAX-1.0.1/');
% Specify your electrode locations with the correct cap file:
RELAX_cfg.caploc=[] %'/home/martin/Documents/standard-10-5-cap385.elp'; % path containing electrode positions. Set to =[] if electrode locations are already in your EEG file.
% Specify the to be processed file locations:
RELAX_cfg.myPath= procDir;
% Or specify the single file to be processed:
RELAX_cfg.filename=[]; % (including folder path)
%% List all files in directory
cd(RELAX_cfg.myPath);
RELAX_cfg.dirList=dir('*.set');
RELAX_cfg.files={RELAX_cfg.dirList.name};
if isempty(RELAX_cfg.files)
disp('No files found..')
end
%% Parameters that can be specified:
% The following selections determine how many epochs of each type of
% artifact to include in the mask. It's not obvious what the best choices are.
% I have selected as defaults the parameters that work best for our data.
% If only performing one MWF run, perhaps including all artifacts in the mask
% is best (as long as that leaves enough clean data for the clean data
% mask). Somers et al. (2018) suggest that it doesn't hurt to
% include clean data in the artifact mask, and so it's helpful to have wide
% boundaries around artifacts.
% However, I wonder if some files might include the majority of trials in the artifact
% mask could mean that most of the task related ERPs might be considered
% artifacts, and cleaned out from the data. Including ERP trials in the clean
% mask reduces this issue, as the clean mask defines what the good
% data should look like, and if it has ERPs, then the artifact periods will
% be cleaned of only the difference between the clean mask and the artifact
% mask, leaving the ERP in the data. This relies upon a consistent
% ERP across both the clean and artifact masks however. An alternative fix
% could be to implement NaNs (which the MWF mask ignores) over ERP periods
% instead of marking them as artifacts, so the ERP is not included in the
% artifact mask, but this would mean any artifacts that are present in the
% ERPs but not outside the ERP periods aren't cleaned.
% Lastly, you may want to delete more unused electrodes than specified by
% this script. If so, modify the section titled: "Delete channels that
% are not relevant if present" to include the electrodes you would like to
% delete.
RELAX_cfg.Do_MWF_Once=1; % 1 = Perform the MWF cleaning a second time (1 for yes, 0 for no).
RELAX_cfg.Do_MWF_Twice=1; % 1 = Perform the MWF cleaning a second time (1 for yes, 0 for no).
RELAX_cfg.Do_MWF_Thrice=1; % 1 = Perform the MWF cleaning a second time (1 for yes, 0 for no). I think cleaning drift in this is a good idea.
RELAX_cfg.Perform_wICA_on_ICLabel=1; % 1 = Perform wICA on artifact components marked by ICLabel (1 for yes, 0 for no).
RELAX_cfg.Perform_ICA_subtract=0; % 1 = Perform ICA subtract on artifact components marked by ICLabel (1 for yes, 0 for no) (non-optimal, intended to be optionally used separately to wICA rather than additionally)
RELAX_cfg.ICA_method='fastica_symm';
RELAX_cfg.Report_all_ICA_info='no'; % set to yes to provide detailed report of ICLabel artifact information. Runs ~20s slower per file.
RELAX_cfg.computerawmetrics=1; % Compute blink and muscle metrics from the raw data?
RELAX_cfg.computecleanedmetrics=1; % Compute SER, ARR, blink and muscle metrics from the cleaned data?
RELAX_cfg.MWFRoundToCleanBlinks=2; % Which round to clean blinks in (1 for the first, 2 for the second...)
RELAX_cfg.ProbabilityDataHasNoBlinks=0; % 0 = data almost certainly has blinks, 1 = data might not have blinks, 2 = data definitely doesn't have blinks.
% 0 = eg. task related data where participants are focused with eyes open,
% 1 = eg. eyes closed recordings, but with participants who might still open their eyes at times,
% 2 = eg. eyes closed resting with highly compliant participants and recordings that were strictly made only when participants had their eyes closed.
RELAX_cfg.DriftSeverityThreshold=10; %MAD from the median of all electrodes. This could be set lower and would catch less severe drift
RELAX_cfg.ProportionWorstEpochsForDrift=0.30; % Maximum proportion of epochs to include in the mask from drift artifact type.
RELAX_cfg.ExtremeVoltageShiftThreshold=20; % Threshold MAD from the median all epochs for each electrode against the same electrode in different epochs. This could be set lower and would catch less severe voltage shifts within the epoch
RELAX_cfg.ExtremeAbsoluteVoltageThreshold=500; % microvolts max or min above which will be excluded from cleaning and deleted from data
RELAX_cfg.ExtremeImprobableVoltageDistributionThreshold=8; % Threshold SD from the mean of all epochs for each electrode against the same electrode in different epochs. This could be set lower and would catch less severe improbable data
RELAX_cfg.ExtremeSingleChannelKurtosisThreshold=8; % Threshold kurtosis of each electrode against the same electrode in different epochs. This could be set lower and would catch less severe kurtosis
RELAX_cfg.ExtremeAllChannelKurtosisThreshold=8; % Threshold kurtosis across all electrodes. This could be set lower and would catch less severe kurtosis
RELAX_cfg.ExtremeDriftSlopeThreshold=-4; % slope of log frequency log power below which to reject as drift without neural activity
RELAX_cfg.ExtremeBlinkShiftThreshold=8; % How many MAD from the median across blink affected epochs to exclude as extreme data
% (applies the higher value out of this value and the
% RELAX_cfg.ExtremeVoltageShiftThreshold above as the
% threshold, which caters for the fact that blinks don't affect
% the median, so without this, if data is clean and blinks are
% large, blinks can get excluded as extreme outliers)
% Clean periods that last for a shorter duration than the following value to be marked as artifacts,
% and pad short artifact periods out into artifact periods of at least the following length when
% they are shorter than this value to reduce rank deficiency issues in MWF cleaning).
% Note that it's better to include clean periods in the artifact mask rather than the including artifact in the clean mask.
RELAX_cfg.MinimumArtifactDuration=1200; % in ms. It's better to make this value longer than 1000ms, as doing so will catch diminishing artifacts that aren't detected in a neighbouring 1000ms period, which might still be bad
RELAX_cfg.MinimumBlinkArtifactDuration=800; % blink marking is based on the maximum point of the blink rather than the 1000ms divisions for muscle artifacts, so this can be shorter than the value above (blinks do not typically last >500ms)
RELAX_cfg.BlinkElectrodes={'FP1';'FPZ';'FP2';'AF3';'AF4';'F3';'F1';'FZ';'F2';'F4'}; % sets the electrodes to average for blink detection using the IQR method. These should be frontal electrodes maximally affected by blinks. The order is the order of preference for icablinkmetrics.
% A single HOEG electrode for each side is selected by the script, prioritized in the following order (if the electrode in position 1 isn't present, the script will check for electrode in position 2, and so on...).
RELAX_cfg.HEOGLeftpattern = ["AF7", "F7", "FT7", "F5", "T7", "FC5", "C5", "TP7", "AF3"]; % sets left side electrodes to use for horizontal eye movement detection. These should be lateral electrodes maximally effected by blinks.
RELAX_cfg.HEOGRightpattern = ["AF8", "F8","FT8","F6","T8", "FC6", "C6", "TP8", "AF4"]; % sets right side electrodes to use for horizontal eye movement detection. These should be lateral electrodes maximally effected by blinks.
RELAX_cfg.BlinkMaskFocus=150; % this value decides how much data before and after the right and left base of the eye blink to mark as part of the blink artifact window.
% I found 100ms on either side of the blink bases works best with a delay of 7 on the MWF. However, it also seemed to create too short artifact masks at times, which may lead to insufficient rank for MWF, so I left the default as 150ms.
RELAX_cfg.HorizontalEyeMovementType=1; % 1 to use the IQR method, 2 to use the MAD method for identifying threshold. IQR method less effective for smaller sample sizes (shorter files).
RELAX_cfg.HorizontalEyeMovementThreshold=2; % MAD deviation from the median that will be marked as horizontal eye movement if both lateral electrodes show activity above this for a certain duration (duration set below).
RELAX_cfg.HorizontalEyeMovementThresholdIQR=1.5; % If IQR method set above, IQR deviation that will be marked as horizontal eye movement if both lateral electrodes show activity above this for a certain duration (duration set below).
RELAX_cfg.HorizontalEyeMovementTimepointsExceedingThreshold=25; % The number of timepoints (ms) that exceed the horizontal eye movement threshold within the test period (set below) before the period is marked as horizontal eye movement.
RELAX_cfg.HorizontalEyeMovementTimepointsTestWindow=(2*RELAX_cfg.HorizontalEyeMovementTimepointsExceedingThreshold)-1; % Window duration to test for horizontal eye movement, set to 2x the value above by default.
RELAX_cfg.HorizontalEyeMovementFocus=200; % Buffer window, masking periods earlier and later than the time where horizontal eye movements exceed the threshold.
RELAX_cfg.LowPassFilterBeforeMWF='no'; % set as no for the updated implementation, avoiding low pass filtering prior to MWF reduces chances of rank deficiencies, increasing potential values for MWF delay period
RELAX_cfg.FilterType='pop_eegfiltnew'; % set as 'pop_eegfiltnew' to use EEGLAB's filter or 'Butterworth' to use Butterworth filter
RELAX_cfg.HighPassFilter=0.25; % Sets the high pass filter. 1Hz is best for ICA decomposition if you're examining just oscillatory data, 0.25Hz seems to be the highest before ERPs are adversely affected by filtering
%(lower than 0.2Hz may be better, but I find a minority of my files show drift at 0.3Hz even).
if RELAX_cfg.HighPassFilter>0.25
Warning='You have high pass filtered above 0.25, which can adversely affect ERP analyses';
end
RELAX_cfg.LowPassFilter=80; % If you filter out data below 75Hz, you can't use the objective muscle detection method
RELAX_cfg.NotchFilterType='ZaplinePlus'; % set as 'Butterworth' to use Butterworth filter or 'ZaplinePlus' to use ZaplinePlus. ZaplinePlus works best on data sampled at 512Hz or below, consider downsampling if above this.
RELAX_cfg.LineNoiseFrequency=50; % Frequencies for bandstop filter in order to address line noise (set to 60 in countries with 60Hz line noise, and 50 in countries with 50Hz line noise).
RELAX_cfg.ElectrodesToDelete={'CB1'; 'CB2'; 'HEOG'; 'IO1'; 'M1'; 'M2'; 'LO1'; 'LO2'; 'E1'; 'E3'; 'ECG'; 'SO1'; 'ECG'; 'SPARE1'; 'SPARE2'; 'SPARE3'; 'BP1'; 'BP2'; 'VEOG'};
% If your EEG recording includes non-scalp electrodes or electrodes that you want to delete before cleaning, you can set them to be deleted here.
% The RELAX cleaning pipeline does not need eye, heart, or mastoid electrodes for effective cleaning.
RELAX_cfg.KeepAllInfo=0; % setting this value to 1 keeps all the details from the MWF pre-processing and MWF computation. Helpful for debugging if necessary but makes for large file sizes.
RELAX_cfg.saveextremesrejected=0; % setting this value to 1 tells the script to save the data after only filtering, extreme channels have been rejected and extreme periods have been noted
RELAX_cfg.saveround1=0; % setting this value to 1 tells the script to save the first round of MWF pre-processing
RELAX_cfg.saveround2=0; % setting this value to 1 tells the script to save the second round of MWF pre-processing
RELAX_cfg.saveround3=0; % setting this value to 1 tells the script to save the third round of MWF pre-processing
RELAX_cfg.OnlyIncludeTaskRelatedEpochs=0; % If this =1, the MWF clean and artifact templates will only include data within 5 seconds of a task trigger (other periods will be marked as NaN, which the MWF script ignores).
RELAX_cfg.MuscleSlopeThreshold=-0.59; %log-frequency log-power slope threshold for muscle artifact. Less stringent = -0.31, Middle Stringency = -0.59 or more stringent = -0.72, more negative thresholds remove more muscle.
RELAX_cfg.MaxProportionOfDataCanBeMarkedAsMuscle=0.50; % Maximum amount of data periods to be marked as muscle artifact for cleaning by the MWF. You want at least a reasonable amount of both clean and artifact templates for effective cleaning.
% I set this reasonably high, because otherwise muscle artifacts could considerably influence the clean mask and add noise into the data
RELAX_cfg.ProportionOfMuscleContaminatedEpochsAboveWhichToRejectChannel=1; % If the proportion of epochs showing muscle activity from an electrode is higher than this, the electrode is deleted.
% Set muscle proportion before deletion to 1 to not delete electrodes based on muscle activity
RELAX_cfg.ProportionOfExtremeNoiseAboveWhichToRejectChannel=0.05; % If the proportion of all epochs from a single electrode that are marked as containing extreme artifacts is higher than this, the electrode is deleted
RELAX_cfg.MaxProportionOfElectrodesThatCanBeDeleted=1; % Sets the maximum proportion of electrodes that are allowed to be deleted after PREP's bad electrode deletion step
RELAX_cfg.MWFDelayPeriod=30; % The MWF includes both spatial and temporal information when filtering out artifacts. Longer delays apparently improve performance.
%% Check for dependencies:
eeglabPath = fileparts(which('eeglab'));
MWFPluginPath=strcat(eeglabPath,'\plugins\mwf-artifact-removal-master');
addpath(genpath(MWFPluginPath));
if (exist('mwf_process','file')==0)
warndlg('MWF toolbox may not be installed in EEGLAB plugins folder. Toolbox can be installed from: "https://github.com/exporl/mwf-artifact-removal"','MWF Cleaning Not Available');
end
toolboxlist=ver;
if isempty(find(strcmp({toolboxlist.Name}, 'Signal Processing Toolbox')==1, 1))
warndlg('Signal Processing Toolbox may not be installed. Toolbox can be installed through MATLAB "Add-Ons" button','Signal Processing Toolbox not installed');
end
PluginPath=strcat(eeglabPath,'\plugins');
PluginList=dir(PluginPath);
if (exist('iclabel','file')==0)
warndlg('ICLabel may not be installed. Plugin can be installed via EEGLAB: "File" > "Manage EEGLAB Extensions"','ICLabel not installed');
end
if (exist('findNoisyChannels','file')==0)
warndlg('PrepPipeline may not be installed or the folder path for Prep has not been set in MATLAB. Plugin can be installed via EEGLAB: "File" > "Manage EEGLAB Extensions"','PrepPipeline might not be installed');
end
if (exist('ft_freqanalysis','file')==0)
warndlg('fieldtrip may not be installed, or the folder path for fieldtrip has not been set in MATLAB. Plugin can be installed via EEGLAB: "File" > "Manage EEGLAB Extensions"','fieldtrip might not be installed');
end
if (strcmp(RELAX_cfg.ICA_method,'fastica_symm')||strcmp(RELAX_cfg.ICA_method,'fastica_defl')) && (exist('fastica','file')==0)
warndlg('FastICA may not be installed, or the folder path for FastICA has not been set in MATLAB. Plugin can be installed from: "http://research.ics.aalto.fi/ica/fastica/code/dlcode.shtml"','FastICA might not be installed');
end
if (strcmp(RELAX_cfg.ICA_method,'cudaica')) && (exist('cudaica','file')==0)
warndlg('CudaICA may not be installed. Instructions for installation can be found at: "https://sccn.ucsd.edu/wiki/Makoto%27s_useful_EEGLAB_code". Warning - installation can be difficult','CudaICA not installed');
end
if (strcmp(RELAX_cfg.ICA_method,'amica')) && (exist('runamica15','file')==0)
warndlg('AMICA may not be installed. Plugin can be installed via EEGLAB: "File" > "Manage EEGLAB Extensions"','AMICA not installed');
end
if (strcmp(RELAX_cfg.NotchFilterType,'ZaplinePlus')) && (exist('clean_data_with_zapline_plus_eeglab_wrapper','file')==0)
warndlg('ZaplinePlus may not be installed. Plugin can be installed via EEGLAB: "File" > "Manage EEGLAB Extensions"','ZaplinePlus not installed');
end
%% Notes on the above parameter settings:
% OnlyIncludeTaskRelatedEpochs: this means the MWF cleaning templates
% won't be distracted by potential large artifacts outside of task related
% periods. However, it also may mean that event related periods are
% included in the artifact template (to be removed). If artifact masks land
% disproportionately around triggers (or certain types of triggers), my
% sense is that all the EEG activity time locked to that trigger will be
% considered an artifact and filtered out by the MWF (including ERPs),
% potentially reducing your ERP for example. An alternative solution to
% this would be to record a sufficient amount of activity before / after
% the task related activity, which includes all the same artifacts as
% within the task, but also a sufficient amount of clean data. Then use
% this data to create the artifact masks (then perhaps marking the task
% related data as only either clean, or NaNs if they contain artifacts
% (which will be cleaned but aren't thought of as artifacts for the
% artifact template). This would prevent the ERP activity of interest from
% being included in the artifact tempalte (and potentially cleaned)
% Delay periods >5 can lead to generalised eigenvector rank deficiency
% in some files, and if this occurs cleaning is ineffective. Delay
% period = 5 was used by Somers et al (2018). The rank deficiency is
% likely to be because data filtering creates a temporal
% dependency between consecutive datapoints, reducing their
% independence when including the temporal aspect in the MWF
% computation. To address this, the MWF function attempts MWF cleaning
% at the delay period set above, but if rank deficiency occurs,
% it reduces the delay period by 1 and try again (for 3
% iterations).
% Using robust detrending (which does not create any temporal
% dependence,unlike filtering) may be an alternative which avoids rank
% deficiency (but our initial test suggested this led to worse cleaning
% than filtering)
% MuscleSlopeThreshold: (-0.31 = data from paralysed participants showed no
% independent components with a slope value more positive than this (so
% excluding slopes above this threshold means only excluding data that we
% know must be influenced by EMG). Using -0.31 as the threshold means
% possibly leaving low level EMG data in, and only eliminating the data we
% know is definitely EMG)
% (-0.59 is where the histograms between paralysed ICs and EMG ICs cross,
% so above this value contains a very small amount of the brain data,
% and over 50% of the EMG data. Above this point, data is more likely to be
% EMG than brain)
% (-0.72 is the maximum of the histogram of the paralysed IC data, so
% excluding more positive values than this will exclude most of the EMG
% data, but also some brain data).
% Fitzgibbon, S. P., DeLosAngeles, D., Lewis, T. W., Powers, D. M. W., Grummett, T. S., Whitham, E. M., ... & Pope, K. J. (2016). Automatic determination of EMG-contaminated components and validation of independent component analysis using EEG during pharmacologic paralysis. Clinical Neurophysiology, 127(3), 1781-1793.
%% RUN SCRIPT BELOW:
RELAX_cfg.FilesToProcess=1:numel(RELAX_cfg.files); % Set which files to process
[RELAX_cfg, FileNumber, CleanedMetrics, RawMetrics, RELAXProcessingRoundOneAllParticipants, RELAXProcessingRoundTwoAllParticipants, RELAXProcessing_wICA_AllParticipants,...
RELAXProcessing_ICA_AllParticipants, RELAXProcessingRoundThreeAllParticipants, RELAX_issues_to_check, RELAXProcessingExtremeRejectionsAllParticipants] = RELAX_Wrapper_beta (RELAX_cfg);
Hi Martin
I suspect there are no blinks in those files (or at least no blinks that exceed the threshold). Could you please visually inspect the files where the error occurs to confirm this? If it is the case that there are no blinks in those files, the easiest way to address the issue is to set RELAX_cfg.ProbabilityDataHasNoBlinks=1; for those files.
Kind regards,
Neil
Thanks, I will look into it. Would lowering the threshold be a sensible solution? An automatic function which reverts to NoBlinks=1 would be nice in a future edition of RELAX.
BW
Martin
Hello Neil
Setting RELAX_cfg.ProbabilityDataHasNoBlinks=1 results in the same error:
Unrecognized function or variable 'BlinkIndexMetric'.
Error in RELAX_blinks_IQR_method (line 84)
ix_blinkstart=find(diff(BlinkIndexMetric)==1)+1; % indices where BlinkIndexMetric goes from 0 to 1
Error in RELAX_excluding_extreme_values (line 68)
[continuousEEG_initialblinkmarking, ~] = RELAX_blinks_IQR_method(continuousEEG, epochedEEG, RELAX_cfg);
Error in RELAX_Wrapper_beta (line 194)
[continuousEEG, epochedEEG] = RELAX_excluding_extreme_values(continuousEEG, epochedEEG, RELAX_cfg); % Mark extreme periods for exclusion from MWF cleaning, and deletion before wICA cleaning
Error in RELAX_SET_PARAMETERS_AND_RUN_beta_noblinks (line 376)
RELAXProcessing_ICA_AllParticipants, RELAXProcessingRoundThreeAllParticipants, RELAX_issues_to_check, RELAXProcessingExtremeRejectionsAllParticipants] = RELAX_Wrapper_beta (RELAX_cfg);
Setting it to 2 results in another:
Unrecognized field name "NumberOfMuscleContaminatedChannelsRecomendedToDelete".
Error in RELAX_excluding_channels_and_epoching (line 438)
epochedEEG.RELAXProcessingExtremeRejections.NumberOfMuscleContaminatedChannelsRecomendedToDelete=epochedEEG.RELAXProcessing.Details.NumberOfMuscleContaminatedChannelsRecomendedToDelete;
Error in RELAX_Wrapper_beta (line 193)
[continuousEEG, epochedEEG] = RELAX_excluding_channels_and_epoching(continuousEEG, RELAX_cfg); % Epoch data, detect extremely bad data, delete channels if over the set threshold for proportion of data affected by extreme outlier for each electrode
Error in RELAX_SET_PARAMETERS_AND_RUN_beta_noblinks (line 376)
RELAXProcessing_ICA_AllParticipants, RELAXProcessingRoundThreeAllParticipants, RELAX_issues_to_check, RELAXProcessingExtremeRejectionsAllParticipants] = RELAX_Wrapper_beta (RELAX_cfg);
It actually says:
the call to "ft_freqanalysis" took 5 seconds and required the additional allocation of an estimated 95 MB
Removing 64 channel(s)...
So I guess setting it to 2 simply removes every channel.
This particular dataset is very noisy and I could actually not find a clear case of eye blinks when scrolling through it. It would still be nice if RELAX continues the cleaning process even though it doesn't find any eye blinks. I think the solution in this case is to exclude this particular dataset from the analysis though.
BW
Martin