[FieldTrip] Importing B-Alert X10 data

Robert Oostenveld r.oostenveld at donders.ru.nl
Thu Jun 9 13:08:25 CEST 2016 

Hi Robin

In terms of segmenting data after it having been read into memory, the following procedure works:

1) ft_preprocessing, without specifying cfg.trl or trialfun -> this reads the data as a single segment (i.e. one very long trial)
2) ft_definetrial, with cfg.trialfun, which could get events from another file -> this results in cfg.trl
3) ft_redefinetrial which takes the trl from step 2 and the continuous data from step 1.

With regards to the data format: the files are plain EDF and as such supported by FieldTrip out the box. E.g.

>> cfg = []
>> cfg.dataset = '000222000.010616.103900.Events.edf'
>> ft_databrowser(cfg)

>> hdr = ft_read_header(cfg.dataset)

hdr = 

             Fs: 1000
         nChans: 1
          label: {'EDF Annotations'}
       nSamples: 13000
    nSamplesPre: 0
        nTrials: 1
           orig: [1x1 struct]
       chantype: {'unknown'}
       chanunit: {'unknown'}

and

>> cfg = []
>> cfg.dataset = '000222000.010616.103900.Signals.Raw.edf'
>> ft_databrowser(cfg)

>> hdr = ft_read_header(cfg.dataset)
Warning: channels with different sampling rate not supported, using a subselection of 10 channels at
256.000000 Hz 
> In read_edf (line 330)
  In ft_read_header (line 678) 

hdr = 

             Fs: 256
         nChans: 10
          label: {10x1 cell}
       nSamples: 1145600
    nSamplesPre: 0
        nTrials: 1
           orig: [1x1 struct]
       chantype: {10x1 cell}
       chanunit: {10x1 cell}


However, non-standard is to have the events in a separate file. Furthermore, the data file has channels with variable sampling rate, which is also not (by default) supported. See http://www.fieldtriptoolbox.org/getting_started/edf and http://www.fieldtriptoolbox.org/reference/edf2fieldtrip. Looking at the Events channel, I do think that you’ll need a separate trialfun anyway.

From the Signals.Raw I see

>> hdr.orig.Label

ans =

ECG             
POz             
Fz              
Cz              
C3              
C4              
F3              
F4              
P3              
P4              
ESUTimestamp    
SystemTimestamp 
Tilt X          
Tilt Y          
Tilt Z          
EDF Annotations 


which means that the annotations also seem to be present in the data file itself and the separate Event file might not be needed.

With 

>> data = edf2fieldtrip('000222000.010616.103900.Signals.Raw.edf')
Warning: channels with different sampling rate not supported, using a subselection of 10 channels at
256.000000 Hz 
> In read_edf (line 330)
  In ft_read_header (line 678)
  In edf2fieldtrip (line 36) 
reading 1 channels with 8 Hz sampling rate
reading 3 channels with 128 Hz sampling rate
reading 10 channels with 256 Hz sampling rate
reading 1 channels with 512 Hz sampling rate
reading 1 channels with 1024 Hz sampling rate
upsampling 1 channels from 8 to 1024 Hz
the input is raw data with 1 channels and 1 trials
the call to "ft_selectdata" took 0 seconds and required the additional allocation of an estimated 3 MB
resampling data
resampling data in trial 1 from 1

original sampling rate = 8 Hz
new sampling rate = 1024 Hz
the call to "ft_resampledata" took 1 seconds and required the additional allocation of an estimated 146 MB
upsampling 3 channels from 128 to 1024 Hz
the input is raw data with 3 channels and 1 trials
the call to "ft_selectdata" took 0 seconds and required the additional allocation of an estimated 1 MB
resampling data
resampling data in trial 1 from 1

original sampling rate = 128 Hz
new sampling rate = 1024 Hz
the call to "ft_resampledata" took 1 seconds and required the additional allocation of an estimated 165 MB
upsampling 10 channels from 256 to 1024 Hz
the input is raw data with 10 channels and 1 trials
the call to "ft_selectdata" took 0 seconds and required the additional allocation of an estimated 1 MB
resampling data
resampling data in trial 1 from 1

original sampling rate = 256 Hz
new sampling rate = 1024 Hz
the call to "ft_resampledata" took 2 seconds and required the additional allocation of an estimated 507 MB
upsampling 1 channels from 512 to 1024 Hz
the input is raw data with 1 channels and 1 trials
the call to "ft_selectdata" took 0 seconds and required the additional allocation of an estimated 2 MB
resampling data
resampling data in trial 1 from 1

original sampling rate = 512 Hz
new sampling rate = 1024 Hz
the call to "ft_resampledata" took 0 seconds and required the additional allocation of an estimated 105 MB
input dataset 1, 1 channels, 1 trials
input dataset 2, 3 channels, 1 trials
input dataset 3, 10 channels, 1 trials
input dataset 4, 1 channels, 1 trials
input dataset 5, 1 channels, 1 trials
concatenating the channels within each trial
the call to "ft_appenddata" took 1 seconds and required the additional allocation of an estimated 560 MB
adding the comment: reordered the channels to the original order in the EDF file
the call to "ft_annotate" took 0 seconds and required the additional allocation of an estimated 463 MB

data = 

      label: {16x1 cell}
      trial: {[16x4582400 double]}
       time: {[1x4582400 double]}
    fsample: 1024
        cfg: [1x1 struct]

But then

>> cfg = [];
>> cfg.viewmode = 'vertical';
>> ft_databrowser(cfg, data)

shows that the EDF Annotations channel (after upsampling) has been slightly messed up (due to filters). Perhaps it can still be of use. Otherwise I also would not directly know how to link the two files to each other, as the sampling rates are very incompatible (1000 in one, and powers of 2 in the other).

Perhaps you can learn something from the EEGLAB code.

Good luck with the data
Robert



> On 09 Jun 2016, at 11:03, Robin Kramer <kramerrobin92 at gmail.com> wrote:
> 
> ​
>  000222000.010616.103900.Events.edf <https://drive.google.com/file/d/0B4UCcTb4Xh26ZU85WUF4aHJFdms/view?usp=drive_web>​​
>  000222000.010616.103900.Signals.Raw.edf <https://drive.google.com/file/d/0B4UCcTb4Xh26N3BPQWFKRTIwSGM/view?usp=drive_web>​Hello,
> 
> Recently we started a research with a new wearable EEG system called the B-Alert X-10 from Advanced Brain Monitoring. ABM does deliver their own analyzing software and it is compatible with EEGLab but not with Fieldtrip. Has anyone attempted to analyze the B-Alert data via FieldTrip already? 
> 
> The way data is stored is rather odd actually. The data is epoched in every second, with 256 samples (Fs = 256) in each epoch for each channel. In other words, there are as much separate trials as there are seconds. After preprocessing, and some manual labour I could rearrange the epochs and make the data continuous, but then I encountered two challenges. 
> 
> First, the event markers are stored in a different EDF file. Therefore, I have to define my own trial function, but the event-marker data is also stored rather oddly. 
> Second, it seems that fieldtrip cannot segment any data after preprocessing with ft_definetrials(). Sadly, before preprocessing the B-Alert data is in the strange format. Is there a workaround? I have send a link with the two EDF files, in case one would like to inspect them themselves.
> 
> Thank you so much.
> 
> Best regards,
> 
> Robin Kramer
> Student MSc Human-Machine Communication
> University of Groningen, the Netherlands
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160609/e06bc5ae/attachment.html>

More information about the fieldtrip mailing list