SMRs accuracies

[BCItrieste]

Dear all,

we are a group of researchers from Trieste, Italy, and we are performing experiments with BCI2000, with both modules (Mu-Cursor task and P300 application). We analyzed data recorded in the calibration sessions with the Offline Analysis tool (i.e. features plots, spectra plots and scalps plots) and with P3 Classifier, and everything works quite well..now we are wondering if it is possible, with BCI2000, to record an objective quantification of the Classifier performance (i.e. accuracy in %) during the calibration and the task phases of the Mu experiments, as it happens with the P3 Classifier tool.
Kind regards,

Marcello

[gschalk]

Hi,

BCI2000 does not come with a general-purpose classification tool for analyzing SMR activity other than OfflineAnalysis. The most important reason for this is that appropriately interpreting mu rhythm activity depends in part on expert interpretation of brain signal patterns (e.g., to differentiate brain signal modulation from muscle artifacts), and it is difficult to completely automate this process.

Gerv

[BCItrieste]

Dear Gerwin,

first of all thank you for your reply: we appreciate it, since we were wondering if we had missed/misunderstood something in the BCI2000 tutorial. Did you ever try (or, at least, do you think it is possible) to obtain such classifier accuracies information analyzing the files of recorded data (.dat) with other software (i.e. eeglab, bcilab, or others), after performing on them the proper artifacts removal and preprocessing? Or BCI2000data do not include such information?
Thank you again,
best wishes

Marcello

[boulay]

Marcello,

It is easy to load BCI2000 .dat files into EEGLAB. There's a plugin to do so.
From there, you have many options for classification. I haven't tried BCILAB but it looks quite capable.

-Chad