A Diminutive Suggestion for Real-time Graz Cue-based Brain Computer Interface

Transcription

1 Vol. 1(3), Oct. 2015, PP A Diminutive Suggestion for Real-time Graz Cue-based Brain Computer Interface Sahar Seifzadeh 1, Karim Faez 2 and Mahmood Amiri 3 1 Faculty of Computer and Information Technology Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran 2 Electrical Engineering Department, AmirKabir University of Technology Tehran, Iran 3 Medical Biology Research Center, Kermanshah University of Medical Sciences, Kermanshah, Iran Abstract *Corresponding Author's s.seifzadeh@qiau.ac.ir THese days, Brain Computer interface known as BCI is one of most challenging issue in neuroscience researches. The principal goal of these systems are providing a non-muscular pathway which contribute to direct interaction with environment for physically-challenged people. In this paper, we propose a suggestion for real time implementation of Graz 2a dataset in real world. This dataset has four predefined classes of Motor Imagery task itself, We uphold the notion that we should adopt a particular class as a fifth class for No-movement which lead wheelchair to stop or No-movement when user does not do any predefined Motor Imagery tasks to better performance and adopt a decision more naturally in real world. Keywords: Brain Computer Interface, Motor Imagery, Graz dataset 1. Introduction A brain-computer interface (BCI), also known as a brain-machine interface(bmi) and mind-machine interface (MMI) is a system that allows a physically-challenged person to control electronic devices using only his or her brainwaves, without any use of peripheral nerves or movement of muscles. Advances in cognitive neuroscience technologies have started to provide us with the ability to interface directly with the human brain. One of the most considerable usage of BCI systems is control of devices such as a wheelchair or prosthetic arm to communicate with others to express their needs, feelings, etc. among other applications such as play games virtually or biometrics [1] (Figure 1).Totally, there are two categories of BCI: invasive and non-invasive approaches. Invasive BCI methods such as electrocorticogram (ECoG) and Intracortical Neuron Recording. ECoG is a technique that measures electrical activity in the cerebral cortex by means of electrodes placed directly on the surface of the brain. Compared to EEG, ECoG provides higher temporal and spatial resolution as well as higher amplitudes and a lower vulnerability to artifacts such as blinks and eye movement, this method have shown excellent performance in human [2] and animals [3-6]. Intracortical Neuron Recording is a neuroimaging technique that measures electrical activity inside the gray matter of the brain. This method needs to implant microelectrode arrays inside the cortex to capture spike signals and local field potentials (LFP) from neurons. Nevertheless, non-invasive approaches based on electroencephalogram (EEG), magnetoencephalogram (MEG), positron emission topography (PET), functional magnetic resonance imaging (fmri) and near-infrared spectroscopy (NIRs) are more popular Article History: JKBEI DOI: /11027 Received Date: 16 May Accepted Date: 16 Sep Available Online: 20 Oct

2 as it is safer (for human kind, but absolutely invasive methods are more accurate. Among these noninvasive methods, EEG-based BCI more popular because of its portability and cheapness. We will focus on EEG-based BCI techniques which recorded by Graz 2a dataset in BCI competition 2008 by Clemens Brunner in Institute for Knowledge Discovery (Laboratory of Brain-Computer Interfaces), Graz University of Technology (Austria). This cue-based BCI is designed for the purpose of motor imagery. According to this, in the following there are some description about motor imagery (MI) and cue-based BCI Motor Imagery BCI Fig 1: Brain computer interface applications Motor imagery (MI) is a cognitive process that subject imagines that he or she performs a movement without actually performing the movement and without even tensing the muscles. In other words, motor imagery requires the conscious activation of brain regions that are also involved in movement preparation and execution, accompanied by a voluntary inhibition of the actual movement [7]. MI has been used after a stroke, spinal cord injury, or in diseases such as amyotrophic lateral sclerosis (ALS),disorders of consciousness and multiple sclerosis (MS) to attempt to treat loss of arm, hand and lower extremity movement, to help improve performance in activities of daily living and to minimize the effects of unilateral spatial neglect [8]. As you can see in Figure 2, for instance, the motor imagery of right hand is due to left side of the brain and vice versa. 181

3 Fig 2: Motor Imagery of left and right hand 1.2. Cue-based BCI Cue-based BCI also known as Synchronous BCI analyze brain signals during predefined time windows. Thus, any brain signal outside the predefined window is ignored (Figure 3). So, the user is only allowed to send commands during specific periods determined by the BCI system. The merits of a synchronous BCI system is that the onset of mental activity is known in advance and associated with a specific cue [9]. 2. Graz 2a Dataset Fig 3: A simple schema of a synchronized BCI system For the purpose of illustration, in this section we try to visualize each step by our providing figures. Compared to datasets from past BCI Competitions there were eye movement artifacts in dataset 2a as a new challenging issue. The data set consists of EEG data from nine subjects [10]. Each subject was sitting in a comfortable armchair in front of a computer screen. The cue-based BCI paradigm consisted of four different MI tasks, to with the imagination of movement of the left hand (class 1), right hand (class 2), both feet (class 3), and tongue (class 4). In two sessions on different days of each EEG recorded for each subject that each session consists of 6 runes which is separated from the short break. Each RUN consists of 48 trials (Figure 4) (12 per class, 12 * 4 = 48) thus each subject has 48*6=288 trials (figure 5). Twenty-two Ag/AgCl electrodes (With system electrode montage) were used to record the EEG from each subject, all signals were recorded monopolarly with the left mastoid serving as a reference and the right mastoid as ground.all signals were sampled with 250 Hz and band-pass filtered between 0.5 Hz and 100 Hz. The sensitivity of the amplifier was set to 100 μv. Also 50 Hz notch filter was enabled to repress line noise. Furthure more to Twenty-two EEG channels, three monopolar EOG channels ( ) positioned above the nasion and below the outer canthi of the eyes (figure 6). [11] were recorded and sampled with 250 Hz too, they were band-pass and notch filter as same as EEG signals but the sensitivity of the amplifier was set to 1 mv. The EOG channels are provided for the sake of artifact processing methods and must not be used for classification. Fig 4: The schema of one trial 182

4 Fig 5: The schema of all two recording sessions Fig 6: Left: Electrode montage corresponding to the international System. Right: Electrode montage of the three monopolar EOG channels. This dataset consists of 18.mat files which divided into two categories, test and train ones (table 1). Table mat files which divided into two categories ID Training A01T.mat A02T.mat A03T.mat A04T.mat A05T.mat A06T.mat A07T.mat A08T.mat A09T.mat Test A01E.mat A02E.mat A03E.mat A04E.mat A05E.mat A06E.mat A07E.mat A08E.mat A09E.mat 183

5 3. Our Suggestion As we mentioned above, this dataset has four class of MI task (the imagination of movement of the left hand (class 1), right hand (class 2), both feet (class 3), and tongue (class 4)). By way of illustration, imagine that we suppose to implement this MI tasks to real world such as Wheelchair or etc. For instance, we determine that if the user do the MI task of right hand, wheelchair goes right, if the user do the MI task of left hand, wheelchair goes left, if the user do the MI task of feet, wheelchair goes forward and if the user do the MI task of tongue, wheelchair goes back. We uphold the notion that we should adopt a particular class as a fifth class for No-movement which lead wheelchair to stop or Nomovement when user does not do any predefined MI tasks. We demonstrate our suggestion in figure 7 in details. Conclusion Fig 7. The schema of our suggestion Motor imagery is a mental process by which an individual rehearses or simulates a given action. In this paper, we propose new suggestion for implementation of Graz dataset 2a (which has four predefined MI classes) in the real world. We do believe that one specific class should add on four exist classes in this dataset to better performance and to adopt a decision more naturally for physicallychallenged people. We provide fifth class as No-movement class due to No-movement when user does not do any predefined MI tasks. 11. ACKNOWLEDGEMENTS The authors would like to thank the Institute for Knowledge Discovery at Graz University of Technology (Graz BCI Lab group) for sharing their dataset. 184

6 References [1] Palaniappan R (2008) Two-stage biometric authentication method using thought activity brain waves. Int J Neural Syst 18(1):59 66 [2] Ball T., Kern M., Mutschler I., Aertsen A., Schulze-Bonhage A. Signal quality of simultaneously recorded invasive and non-invasive EEG. Neuroimage. 2009; 46: [3] Loeb, G.E.; Walker, A.E.; Uematsu, S.; Konigsmark, B.W. Histological reaction to variousconductive and dielectric films chronically implanted in the subdural space. J. Biomed. Mater.Res. 1977, 11, [4] [Bullara, L.A.; Agnew, W.F.; Yuen, T.G.H.; Jacques, S.; Pudenz, R.H. Evaluation ofelectrodearray material for neural prostheses. Neurosurgery 1979, 5, [5] Yuen, T.G.H.; Agnew, W.F.; Bullara, L.A. Tissue response to potential neuroprosthetic materials implanted subdurally. Biomaterials 1987, 8, [6] Margalit, E.; Weiland, J.D.; Clatterbuck, R.E.; Fujii, G.Y.; Maia, M.; Tameesh, M.; Torres, G.;D'Anna, S.A.; Desai, S.; Piyathaisere, D.V.; et al. Visual and electrical evoked response recorded from subdural electrodes implanted above the visual cortex in normal dogs under two methods of anesthesia. J. Neurosis. Methods 2003, 123, [7] Lotze M, Cohen LG. Volition and imagery in neurorehabilitation. Cogn Behav Neurol. 2006;19: [8] Mulder, Th. "Motor imagery and action observation: cognitive tools for rehabilitation." Journal of neural transmission (2007): [9] Tsui C., Gan J. Asynchronous BCI Control of a Robot Simulator with Supervised Online Training. In: Yin H., Tino P., Corchado E., Byrne W., Yao X., editors. Intelligent Data Engineering and Automated Learning IDEAL Vol Springer; Berlin, Germany: pp [10] Brunner, C., et al. "BCI Competition 2008 Graz data set A." Institute for Knowledge Discovery (Laboratory of Brain- Computer Interfaces), Graz University of Technology (2008): [11] Wang, Deng, Duoqian Miao, and Gunnar Blohm. "Multi-class motor imagery EEG decoding for brain-computer interfaces." Frontiers in neuroscience 6 (2012). 185