Shared data for exploring training effect in 42 human subjects using a noninvasive sensorimotor rhythm-based online BCI

<b>Data description for ‘Exploring training effect in 42 human subjects using a noninvasive sensorimotor rhythm based online BCI’</b> Jianjun Meng and Bin He* Department of Biomedical Engineering, Carnegie Mellon University * e-mail: [e-mail] <b>Summary</b> The shared data are organized into three different folders, corresponding to the three different experiments included in the published paper. The folder ‘Exp1’ contains data for 16 of the subjects who participated in the first experiment. The folder ‘Exp2’ contains the data for 12 of the subjects who participated in the second experiment. Finally, the folder ‘Exp3’ contains the data for 14 of the subjects who participated in the third experiment. All of the data (42 subjects in total) is for left and right one-dimensional cursor control. Please refer to the published paper for more details about the experiment. <b>Specification</b> The three experimental datasets contain, in total, the data of 42 subjects. The data for each experimental session are saved as independent Matlab data files (*.mat) and organized into different folders as described above. Each file for the first experiment contains the first 125 trials, while files from experiments 2 and 3 contain the first 120 trials, keeping the relative number of trial types as uniform as possible. Each .mat file contains only the data from a single session, and is titled using a subject’s ID number, session number, and experiment number (e.g. Subj1_S01_Exp1). Each file includes the online results from the BCI experimentation for each run (saved in a cell variable ‘BCI_UseResults’), key parameters for the experiment (saved in a structure ‘Experiment_Parm’), key parameters for the state of the raw EEG signal (saved in a structure ‘Experimental_states’), and the raw EEG signal (saved in a variable ‘output_data’). The raw EEG signals for experiments one and two are composed of 62 channels of EEG data with a sampling frequency of 100Hz, while data for experiment three contains 64 channels of EEG sampled at 128Hz. By importing a file, ‘Subj1_S01_Exp1.mat’ for example, four variables loaded: ‘BCI_UseResults’, ‘Experiment_Parm’, ‘Experimental_states’ and ‘output_data.’. In ‘BCI_UseResults’, the number of trials for each of the five runs, the number of trials hitting or aborting a target in each run, the percent valid correct, and the information transfer rate for each of the five runs are included. In ‘Experiment_Parm’, the sampling frequency of 100Hz, the name of the channels included using the 10-20 system (see original paper for more information), prefeedbackduration, postfeedback duration and inter-trial interval duration are included. ‘Experimental_states’ contains ‘TargetCode’, ‘ResultCode’,’Feedback’,’CursorPosX’,’CursorPosY’. Each of these variables indicates the current status of a state variable throughout the session. Please refer to BCI2000 for the meaning of the variables. Finally, ‘output_data’ contains the raw EEG signals of those select five runs. <b>Additional Information</b> If you choose to use the data described and released here, we request that you cite the adjoining research article: Meng, J., He, B.<i> </i>Exploring training effect in 42 human subjects using a noninvasive sensorimotor rhythm based online BCI. <i>Front. Hum. Neurosci.13:128 </i>doi: 10.3389/fnhum.2019.00128 (2019).<br>