channel selection in Mu rhythm

[Yueqing Li]

Hello:

I am training users to use Mu rhythm with BCI2000. I found several questions:

1. I am trying to use left hand movement as a control signal in the curse movement. According to the tutorial, electrode C3 should be used. But, I found a bigger r square value in electrode F3. So, should I use F3 instead of C3?

If I choose F3, should I replace C3_OUT in spatial filter and classifier with F3_OUT?

2. How big should the r square value be? I mean is there a threshold value for the channel to be chosen? Sometimes, I could get 0.15, but sometimes, only 0.05. I noticed that it is 0.40 in the tutorial.

3. Since the EEG features are the only control signals in the cursor task application, should the cursor just move horizontally without any ups and downs when the users relax? I think it should be. But I found that the cursor is always moving up and down even when the users relaxed. Is there any explanation or I did something wrong?



Thanks a lot.

[jawilson]

Hello Yueqing,

1. I am trying to use left hand movement as a control signal in the curse movement. According to the tutorial, electrode C3 should be used. But, I found a bigger r square value in electrode F3. So, should I use F3 instead of C3?

If I choose F3, should I replace C3_OUT in spatial filter and classifier with F3_OUT?

First, if you are moving your left hand, then you would expect a change over contra-lateral sensorimotor cortex, i.e., on C4, not C3. C3 should change when you move your right hand.

Next, you should really only expect to see changes on C3, C4, and Cz for hand and foot movements. If you are not seeing a relatively large r^2 for real movements (e.g., r^2 > 0.1), then you likely need to collect more screening data using the StimulusPresentation task. Another possibility might be that the cap was not put on correctly, and F3 is close to the area that C3 should be located; this would happen if the front of the cap is too far back.

2. How big should the r square value be? I mean is there a threshold value for the channel to be chosen? Sometimes, I could get 0.15, but sometimes, only 0.05. I noticed that it is 0.40 in the tutorial.

The values from the tutorial were chosen because they were a very good data set recorded from a well trained subject, and had many trials. Typically, an r^2 of 0.4 is about as large as you would expect to see. For real movements, you probably want r^2 to be about 0.2, and for imagined, about 0.1. These are only guidelines though, and you probably need to collect many runs data to determine for yourself what a good result looks like.

3. Since the EEG features are the only control signals in the cursor task application, should the cursor just move horizontally without any ups and downs when the users relax? I think it should be. But I found that the cursor is always moving up and down even when the users relaxed. Is there any explanation or I did something wrong?

Are you trying to configure BCI2000 to only move the cursor up and down, e.g., by using hand and foot imagery? If so, then ideally the cursor would not move when the subject relaxes. In practice, though, there is a lot of variability in the brain signals that are recorded that are unrelated to the task being performed, and so the cursor will continue to move even if the subject is relaxing.

Does all of this make sense? Let me know if you have more questions.
Adam

[Yueqing Li]

Hi, Adam:

Thank you very much for your quick answer.

I made a typo there. I mean right hand movement.

I think you are right. I may put the cap too far back. I will try another time.

Thanks again and have a good day.

[jawilson]

No problem. By the way, in case you need a reference for putting the cap on, there is one on the BCI2000 wiki here:
http://www.bci2000.org/wiki/index.php/U ... ment_Setup
Adam