User Tutorial:Performing a Mu Rhythm Feedback Session

From BCI2000 Wiki
Revision as of 14:39, 27 February 2020 by Mellinger (talk | contribs) (→‎Instructions to the Subject)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

This step assumes that you created a subject-specific configuration file for the on-line system as described in the previous step of this tutorial.

Preparation

If you quit BCI2000 after the previous step, start it using the appropriate batch file at batch/CursorTask_<YourAmplifier>.bat, or the link to that file which you created on the desktop.

Then, load the configuration file that you saved in the previous step.

Click Set Config to view the EEG signal, and prepare the subject for EEG recording as you did for the initial session.

Instructions to the Subject

When the subject is ready for EEG acquisition, it is time to brief the subject about the experimental task. Suggested instructions to the experimenter and subject are listed below.

CursorTask.PNG

A screen with a black background is initially presented. As soon as the subject is ready and the EEG traces are stabilized, the investigator will start the acquisition. For each trial, four phases will occur:

  1. Target presentation. A target appears on the edge of the screen for about 10 second.
  2. Cursor movement. A cursor appears on the middle of the screen, and begins to move vertically towards the edge of the screen. Its position is controlled by the EEG features that were defined in the previous step. The subject's task is to engage in respective hand or feet movement based on the location of the cursor.
  3. Result. If the subject successfully hits the target, the target changes its color. Otherwise, no change occurs. In either case, this period lasts one second.
  4. Intertrial Interval. The screen will then turn black for one second. This indicates the end of the trial. After this one-second period, the next trial starts.

When a target is presented, the subject should engage the type of movement associated with the channel-frequency features chosen for feedback. In our example, since the largest r-squared value was associated with "right hand vs. rest", when the target is on the bottom edge, move right hand; when the target is on the top edge, stay rest.

In the first feedback Session, the motion of the cursor might not follow the behavior of the subject. This is because the Adaptaion mode is on. In Filtering tab, Adaptation box contains two numbers 0 2. As our example involves only one dimensional control, the horizontal component is 0. By having 2 on the vertical component, the normalizer will "learn" the best coefficient to distinguish the difference of mu power between "right hand" and "rest". When the first session, or the "learning session" ended, one can set the vertical component to 0, and subject can then try to gain control on the cursor.

Explicitly specifying a type of imagination to control cursor movement will help the subject achieve initial cursor control. Once the subject has become more proficient with the task, motor imagery typically becomes less important. In this situation, it is not uncommon that subjects report that they just "imagine to move the cursor."

The second set of instructions to the subject regards the minimization of artifacts from

  • Contraction of the muscles of the face/head, swallowing;
  • Eye blinks and eye movements;
  • Motion during the "rest" phase.

Provided that subjects are asked to minimize artifacts, he/she should be further assisted in these efforts by providing a comfortable chair and a dimly lit room. The experimenter must carefully monitor the EEG and alert the subject in the case he/she has forgotten some of the instructions. When the experimenter is sure that his/her instructions have been well understood, the recording session may start.

Click the Start button to start the feedback experiment. During the experiment, the subject's performance is written into a log window on the experimenter's screen, and recorded into a log file that is saved to disk in the session directory. The experimenter should minimize noise in the room and not disturb the subject.

Monitoring the Recording

After recording has started, the experimenter may feel the temptation to leave the subject alone during the run since most of the experimental activities are automated in BCI2000. However, the experimenter has several important tasks during the experiment:

  • Filling in a run-sheet to report information that is not automatically recorded by BCI2000 and that will later help when data are analyzed (e.g., subject did not seem to understand the instructions in the first run, instructions to the subject for a particular run, etc.).
  • Monitoring the EEG signal to verify the quality of the recording (e.g., no electrode contact failure, muscular, ocular, or motion artifacts, etc.)
  • Reinforce the subject: notify the subject if he/she is producing artifacts, keep the subject alert if getting drowsy, give the subject feedback about his/her performance so that interest, alertness, and attention is kept high.

The run sheet is the most important means of communication between the technician who performs the recordings and the person who analyzes the data (or a comprehensive reminder if somebody does both). It is important that it is compiled carefully and that it is rich in what may seem to be obvious detail: only time will say what is standard and what changes from session to session, and if you will need to analyze data acquired years before, you are likely to miss information if you did not record all information.

Multiple Sessions

Once the initial run has ended, BCI2000 goes into suspended state.

GainControl.PNG

As discussed above, you can now change the Adaptation vertical component from 2 to 0. To control sensitivity of the cursor, you can increase the vertical component of the NormalizerGains.

Click Resume to test out the gain value. After the session has finished, you may want to save auto-adjusted parameters for the next session. Use Save Parameters from the configuration window to do this.

Alternatively, the Load Parameters dialog allows you to choose a data file rather than a parameter file, and thus use the configuration contained in a previous session's data file for the next session. However, parameters contained in a data file reflect the state at the beginning of the recording, so changes during a session's last run cannot be recovered that way.

When starting the next session, don't forget to increment the SessionNumber parameter on the Storage tab. Otherwise, new runs will be added to the previous session's directory. As a safety net, BCI2000 will never overwrite existing data files, and it documents date and time in the StorageTime parameter. This allows to later separate data files into runs even if the SessionNumber parameter has not been increased.

After each session, it is recommended that you analyze the recorded data in the same way as you did for the initial session. This allows you to track and adapt to signal changes in the subject's parameters that may occur in the course of learning.

As mentioned in introduction, ERD and ERS's occurrence don't require a physical presence of movements. Using the same methodology, subject can try to the cursor with pure imaginations. Explicitly specifying a type of imagination to control cursor movement will help the subject achieve initial cursor control. Once the subject has become more proficient with the task, motor imagery typically becomes less important. In this situation, it is not uncommon that subjects report that they just "imagine to move the cursor."

Important Remarks

One critical element of such experiments is that they need to be consistent and rigorous. For example, a typical session will consist of a number (e.g., 4-8) of 3-min experimental runs. Unless there is an obvious technical problem (e.g., the cursor always immediately jumps to the bottom of the screen, which would point to a misconfiguration of BCI2000), do not change any of the parameters (such as locations, frequencies, etc.) across these runs. When doing offline analyses, always strive to collect at least four runs with the exact same configuration. Because there is so much variability in the subject's performance and in the EEG, it is likely that you will otherwise not be able to derive meaningful results or conclusions. You may find that, for example, for three consecutive sessions the subject's best frequency is 12 Hz and not 10 Hz as initially configured. In this case, you could make this small adaptation to the parameters, and have a reasonable chance that it will actually improve the subject's performance.

Finished

Here, the Mu rhythm tutorial is finished.

Congratulations! You are now able to perform mu rhythm feedback experiments.

See also

User Tutorial:Mu Rhythm BCI Tutorial