[Eeglablist] Problem with EEG dataset loading and visualization in BCILAB

[Tarik S Bel-Bahar]

Hello Nilofer,

1. From what it looks like in the images you showed, you may not be
filtering the data properly as per BCI import functions.
Take a closer look through the bcilab documentation on that.

2. You should not be experiencing data loss. This may be due to the very
long length of your total data, or to "epoching" settings in the bcilab
import function.

3. In the future, and for this email, you probably want to also CC
Dr.Kothe, the developer of BCILAB. You may also want to check that you
original data are okay (via a non-bcilab path), and that you have the most
recent versiosn of bcilab/eeglab.

4. One alternative you could consider is to properly load up the files
within eeglab, save them as .set files, confirm that they are OK, and that
they have all the data, and then from there attempt loading them into
BCILAB.

5. If you lack of events is only due to the data loss, once you resolve
proper import, you should be fine. If you need to add event markers, you
should be able to do that via eeglab or bcilab functions.

​6. When you have a solution to your queries, please share them with the
list so that other users can benefit from your experiences.​













On Thu, Nov 2, 2017 at 8:32 PM, Nilofer Husain <nilofer.husain at gmail.com>
wrote:

> Hi,
>
> I have been trying to load my EEG data (~8 hours data from an overnight
> sleep study) into BCILAB for offline analyses but when I inspect it before
> training, the EEG plot isn't displayed correctly and chunks of my data are
> missing. I collected my EEG data using BrainVision Recorder and added
> channel locations and event labels using EEGLAB. I loaded this dataset
> (.set) to BCILAB. Interestingly, as you can see in the images (
> https://ibb.co/nvYQCb; https://ibb.co/gSrCsb; https://ibb.co/jtgCsb),
> when I inspect the data I see 1-10 seconds data in the first window,
> 299.9-309.9 data in the second window, 599.8-609.8 in the third window (a
> consistent loss of 289.9 seconds data across windows) and so on. Moreover,
> the axes are wrong and it's not 'channel activities EEG plot' that I see
> with EEGLAB and what's shown in the BCILAB tutorial (
> https://www.youtube.com/watch?v=Sup2SaJyFQE&t=33). Does anyone have any
> idea why the data doesn't load properly? I tried different paradigms but I
> cannot train the model because of the lack of event markers in the data
> training because of data loss. Does anyone have any suggestions as I am
> very new to BCILAB? I will really appreciate any recommendations. Thanks!
>
> Software versions used:
> MATLAB version 2015a (Mac OS Sierra 10.12.4 but I also observed this on
> Windows)
> BCILAB DEVEL (with EEGLAB 13_4_4b in dependencies) - With BCILAB master, I
> get supergui problems
>
> Error obtained when I try to train my data (my event markers are present
> in the data when I check with EEGLAB but because of data loss, most of the
> events are absent in the data loaded by BCILAB)
>
> WARNING: the data set did not contain matching target markers.
> WARNING: set_gettarget did not find any target information in this data
> set. See help of set_gettarget and set_targetmarkers for how data sets can
> be annotated with target information.
>
> The loaded EEGLAB set is lacking an online expression; assuming it
> contains unfiltered data.
> If it contains filtered data, however, BCI models derived from it will
> likely not be online-capable.
> fold.
> WARNING: This data set contains no trials for one of your target classes:
> please check whether your target marker naming is correct, and ensure that
> your data has enough target events in each of its required subsets (e.g.,
> in a (nested) cross-validation).
> WARNING: After set_selepos your signal is empty.
> WARNING: This training set contains only one class - the subsequent
> learning phase will likely fail.
>
> Computation failed; error trace:
>  Your training data set has no trials for one of your classes; you need at
> least two classes to train a classifier.
>  The most likely reasons are that one of your target markers does not
> occur in the data, or that all your trials of a particular class are
> concentrated in a single short segment of your data (10 or 20 percent). The
> latter would be a problem with the experiment design.
>  occurred in:
>    ml_trainlda: 103
>    ml_train: 170
>    ParadigmDataflowSimplified.calibrate_prediction_function: 481
>    ParadigmDataflowSimplified.calibrate_simple: 602
>    ParadigmBaseSimplified.calibrate: 184
>    instantiate_paradigm/@(varargin)instance.calibrate(varargin{:}): 0
>    @(trainset,varargin)utl_complete_model(args.calibrate_
> func('collection',{trainset},varargin{:}),args.predict_func): 903
>    evaluate_internal: 62
>    cached_evaluate: 42
>    utl_evaluate_fold: 34
>    par_beginschedule: 166
>    par_schedule: 71
>    utl_crossval: 311
>    utl_nested_crossval: 141
>    run_computation: 938
>    make_func/@(f,a,frame__f1)feval(f,a{:}): 0
>    hlp_scope: 53
>    bci_train: 856
>    pushbutton1_Callback: 156
>    gui_mainfcn: 95
>    gui_calibratemodel: 30
>    @(hObject,eventdata)gui_calibratemodel('pushbutton1_
> Callback',hObject,eventdata,guidata(hObject)): 0
>
> Thank you very much!
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.
> ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20171106/f0162ad0/attachment.html>