Parameter settings for neuroscan

[emily]

If anyone has used Neuroscan could you please let me know if there is a problem with the following parameter settings as the participants don't seem to be learning anything.
thanks.

We're using Neuroscan acquire 4.3.1 and 16 channels for the 1D 2target task.

Source tab

Neuroscan ADC
SourceCh 16
SampleBlockSize 20
Sampling rate 500
ServerAddress localhost: 3999

Signal Properties
ChannelNames F7 T7 P7 F3 C3 P3 Fpz Fz Cz Pz F4 C4 P4 F8 T8 P8
SourceChOffset 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
SourceChGain 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847 0.167847

Alignment
AlignChannels 0
SourceChTimeOffset

Online Processing
TransmitChList 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16


Filtering Tab

SpatialFilter
SpatialFilter common_average_16.prm loaded

ARFilter
WindowLength 0.5s
Detrend No
ModelOrder 16
OutputType Spectral Amplitude
FirstBinCenter 0Hz
LastBinCenter 30Hz
BinWidth 3Hz
EvaluationsPerBin 15

LinearClassifier
Classifier C3 9Hz 2 0.5
C4 9Hz 2 0.5

LPTFilter
LPTimeConstant 1s

Normalizer
NormalizerOffsets -1 32 (default)
NormalizerGains 1 0.5 (default)
Adaptation 0 2
BufferConditions
0 (Feedback)&&(TargetCode==1)
0 (Feedback)&&(TargetCode==2)

BufferLength 9s
Update Trigger (Feedback==0)

thanks very much
Emily

[gschalk]

Emily,

It is somewhat difficult to remotely verify your parameter settings, because there are so many variables, either in other parameters or in your device, montage, etc.

Your Neuroscan parameters seem to be OK. You can verify this best yourself by making sure you get good signals at all channels. Also, you may want to look at the source display, poke each channel in the cap, and see if the appropriate channel in the display responds.

Regarding the BCI, I would run the experiment for a number of runs, and then use OfflineAnalysis to analyze your results using the same conditions (e.g., (Feedback == 1) && (TargetCode == 1)). If you do not see good responses that somewhat resemble what you see in the tutorials, and even more importantly, at the channel and frequency specified in the classifier, it obviously will not work. In this case, you may have a problem with the signals, the subject, or instructions to the subject. If you do get good results in the offline analysis, but no online control, then the BCI2000 parameterization is at fault.

Why don't you go through these steps and let us know what happened?

Gerv

[emily]

Hi Gerv,

I've checked the neuroscan parameters and everything is responding appropriately.

Can I just clarify that you mean in the offline analysis for condition 1 put (states.TargetCode == 1) and for condition 2 enter (states.TargetCode == 2) and then enter (states.Feedback == 1) in the trial change condition?

If I do this and analyse a feedback session, made up of 15 runs I don’t get anything in the 8-13Hz range at all. Most of the activity appears to be in the 26-28Hz range for one participant and all over the place for the other participant. Which could explain why no one is improving.

However when I analyse the initial session info for these participants there is strong activity in low mu range 6-10 HZ although this seems to vary widely in electrode position.

My supervisor wants to control the electrode position and Hz centering for feedback for all participants so that everyone is being trained with the same protocol. Psychologists and experimental control … Is this an unviable method for BCI? I know you’ve explained before that it requires individual centering and electrode positioning for each participant and each session. If I don’t do this am I just running sham feedback sessions with no hope of my participants improving unless their mu happens to fall in 7.5-11.5 Hz over C3 and C4.

BTW have also replied to Josh regarding offline analysis of initial session.
Thanks
Emily