EFTA02693109.pdf

lopscience iopscience.iop.org Home Search Collections Journals About Contact us My lOPscience Fast attainment of computer cursor control with noninvasively acquired brain signals This article has been downloaded from lOPscience. Please scroll down to see the full text article. 2011 J. Neural Eng. 8 036010 (http://iopscience.iop.org/1741-2552/8/3/036010) View the table of contents for this issue, or go to the journal homepage for more Download details: IP Address: 66.44.80.91 The article was downloaded on 16/04/2011 at 04:09 Please note that terms and conditions apply. EFTA_R1_02036188 EFTA02693109 KW PUBLISHING J€rl ANAL OF NEURAL ENGINF/ROW J. Neural Eng. 8 (2011103601019pp) I 7 II -2560'S I WOW, Fast attainment of computer cursor control with noninvasively acquired brain signals Trent J Bradberry I • • , Rodolphe J Gent1112.3 and Jose L Contreras-Vidal l '23' 5 Fischell Department of Bioengineering, University of Maryland, College Park, MD 20742. USA 2 Department of Kinesiology, University of Maryland. College Park. MI) 20742. USA 'Graduate Program in Neuroscience and Cognitive Science, University of Maryland. College Park. MD 20742. USA E-mail: tretilhGt unid.cdu and pepcutraaurnd cdu Received 7 January 2011 Accepted for publication 18 February 2011 Published 15 April 2011 Online at stacks.iop.org/JNE/8/036010 Abstract Brain-computer interface (BCI) systems are allowing humans and non-human primates to drive prosthetic devices such as computer cursors and artificial arms with just their thoughts. Invasive BCI systems acquire neural signals with intracranial or subdural electrodes, while noninvasive BO systems typically acquire neural signals with scalp electroencephalography (EEG). Some drawbacks of invasive BCI systems are the inherent risks of surgery and gradual degradation of signal integrity. A limitation of noninvasive BCI systems for two-dimensional control of a cursor, in particular those based on sensorimotor rhythms. is the lengthy training time required by users to achieve satisfactory performance. Here we describe a novel approach to continuously decoding imagined movements from EEG signals in a BCI experiment with reduced training time. We demonstrate that, using our noninvasive BCI system and observational learning, subjects were able to accomplish two-dimensional control of a cursor with performance levels comparable to those of invasive BCI systems. Compared to other studies of noninvasive BC1 systems, training time was substantially reduced, requiring only a single session of decoder calibration (-'20 min) and subject practice (--20 min In addition, we used standardized low-resolution brain electromagnetic tomography to reveal that the neural sources that encoded observed cursor movement may implicate a human mirror neuron system. These findings offer the potential to continuously control complex devices such as robotic arms with one's mind without lengthy training or surgery. El Online supplementary data available from cracks itip.orONK/8/036010/mmedia (Some figures in this article are in colour only in the electronic version) I. Introduction and smart artificial anns. Currently the most promising BCI systems rely on neural signals acquired noninvasively Brain—computer interface (BCI) systems may potentially with electroencephalography (EEG) (Wolpaw and McFarland provide movement-impaired persons with the ability to 2004) or invasively with electroconicography (ECoG) (Schalk interact with their environment using only their thoughts to et al 2(x0K) or microelectrode arrays seated into cortical tissue control assistive devices such as communication programs (Hochberg et al 2000. Current noninvasive EEG-based BCI systems for 2D Prclecnt address: Matron. Inc.. Reston, Virginia 20190. USA. cursor control require subjects to learn how to modulate S Author to whom any correspondence should be addressed. specific frequency bands of neural activity, i.e. sensorimotor 1741-2560/11/036010.09533.00 0 2011 HOP Publishing Lid Printed in the UK EFTA_R1_02036189 EFTA02693110 I. Neural Eng. 8 (2011) 036010 T1Bradbeny et al rhythms, to move a cursor to acquire targets (Wolpaw and McFarland 2004). These types of studies based on sensorimotor rhythms require weeks to months of training before satisfactory levels of performance are attained. Relative to EEG signals, the increased signal-to-noise ratio and bandwidth of invasively acquired neural data are commonly thought to be factors that reduce the training time required by users of invasive BCI systems (Schalk a al 2(08). In addition. studies of tetraplegic humans with implanted microelectrode arrays have exclusively demonstrated 213 control of a cursor through imagined natural movement (Hochberg et al 2006. Kim et al 2008). This decoding of imagined natural movement is also a likely factor in reduced training time since neural signals directly correlate with intended actions. However, recently several off-line decoding studies have demonstrated the reconstruction of cursor and hand kinematics from noninvasive magnetoencephalography Figure 1. Setup of EEG-based BCI experiment. Subjects EEG (MEG) (Bradbeny et al 2009) and EEG (Bradberry et al 2010). signals were acquired while sitting in a chair facing a monitor that The noise and bandwidth limitations of the noninvasively displayed a cursor and targets (only target acquisition phase). acquired signals did not impede decoding kinematics of During the calibration phase, subjects observed a computer- natural movement. This finding implies that a noninvasive controlled cursor to collect data for subsequent initialization of the decoder. In the target acquisition phase (shown in the photo). BCI system based on the decoding method reported in those subjects moved the brain-controlled cursor to acquire targets that studies may require little training time. appeared at the left, top, right, or bottom of the computer screen. In this study, we sought to investigate the use of the (for a more detailed schematic, see figure sI in the supplementary decoding method reported in those off-line studies in an EEG- data, available at Ntachs.iopmrvLINE/8/0360 Illinunethaj based BCI system during a single session lasting less than 2 h that required only brief training. We hypothesized that the 30 cm and a cursor of diameter 1.5 cm (0.20% of workspace) putative human mirror neuron system (MNS), which predicts (figure O. Subjects were instructed to remain still and relax and interprets one's own actions and the actions of others their muscles to reduce the introduction of artifacts into the (Tkach et al 2008), could be exploited during training by EEG recordings. asking subjects to combine motor imagery with observation of a video of cursor movement. In fact., several of the 2.1.1. Calibration phase. During the calibration phase, aforementioned invasive studies (Hochberg a al 2006, Kim subjects were instructed to imagine moving their right et al 2(08) successfully demonstrated a similar approach arm/finger to track a computer-controlled cursor that moved to training. We further hypothesized that a neural decoder in two dimensions on the computer screen. The movements of could subsequently be built off-line that would predict cursor the computer-controlled cursor were generated by replaying movement from neural activity, and the decoder could then be a 10 min recording of a pilot subject's brain-controlled used on-line for real-time brain-control of cursor movement cursor movements from one of his practice runs (this pilot with little training time. Furthermore, to provide additional subject did not participate as one of the five subjects in validation of our hypotheses, we sought to examine the this study). Histograms of the horizontal and vertical involvement of neural regions in encoding cursor velocity positions and velocities of the computer-controlled movements during observation of the cursor movement and during tasks indicated approximately uniform coverage of the workspace requiring a brain-controlled cursor to acquire targets in 213 and biological motion respectively (figure 2). The decoding space. procedure described in section 2.3 below was subsequently executed (^-10 min of computation time) to calibrate the 2. Materials and methods decoder so that it best mapped the EEG signals to observed horizontal and vertical cursor velocities. During pilot testing, 2.1. Experimental tasks we discovered that asking subjects to visually fixate the center The Institutional Review Board of the University of Maryland of the workspace while simultaneously tracking the cursor at College Park approved the experimental procedure. After added attentional demands that burdened the subjects and giving informed consent, five healthy. right-handed, male likely compromised the decoding; therefore, we told subjects subjects performed a three-phase task: calibration, practice they were free to move their eyes but to always maintain eye and target acquisition. None of the subjects had previously contact and spatial attention with the moving cursor. participated in a BO study. In all phases. their EEG signals were acquired while they sat upright in a chair with hands 2.1.2. Practice phase. During the practice phase. the resting in their laps at arm's length away from a computer subjects used the calibrated decoder to attempt to move monitor that displayed a workspace of dimensions 30 cm x the cursor with their thoughts in two dimensions as desired 2 EFTA_R1_02036190 EFTA02693111 1. Neural Eng. 8 (2011) 036010 T.1 Bradbeny et al x (B) 2000 2000 1500 1500 WOO 1000 500 500 S00 -100 0 100 -100 0 100 Vela* feral) Wooly (cm%) Figure 2. Histograms of observed cursor kinematics during the calibration phase. (A) Histograms of horizontal (left) and vertical (right) positions indicated approximately uniform coverage of the workspace. (13) Histograms of horizontal (left) and vertical (right) positions inferred movement: with bell-shaped velocity profiles (although these are more super-Gaussian than typical point-to-point movements). indicative of biological motion. The velocity histograms actually peak near 5000 but were truncated so the shape of the base could be viewed. (without task constraints). They were instructed to determine activity was measured with a bipolar sensor montage with for themselves how to best control the cursor by exploring the sensors attached superior and inferior to the orbital fossa of workspace. They were also informed as to where the target the right eye for vertical eye movements and to the external locations would be in the target acquisition phase that would canthi for horizontal eye movements. The EEG signals were follow. Again, they were free to move their eyes. During the continuously sent to the BCI2OOO software system (Schalk initial portion of the practice phase. horizontal and vertical et al 20lµ) for online processing and storage. BCI2OOO gains were independently adjusted by the investigators to was responsible for moving the cursor based on our decoder balance cursor speed and to ensure full coverage of the display function. which we integrated into the open source software workspace by the brain-controlled cursor. After the gains were system. BC12OOO was also responsible for storing cursor manually adjusted (^-10 min). subjects practiced moving the movement data as well as collecting markers of workspace cursor without task constraints for 10 min. events such as target acquisition. Electromyographic (EMG) signals were amplified and collected at 2000 Hz from two 2.1.3. Target acquisitionphase. During the target acquisition bipolar surface electrodes over the flexor carpi radialis and phase, subjects were instructed to use their thoughts to move extensor digitorum muscles of the right forearm using an the cursor in two dimensions to reach a peripheral target (1.3% Aurion ZeroWire system (10-1000 Hz bandwidth, constant of workspace) that would appear pseudorandomly at the top, electrode gain of 1000). bottom, left, or right side of the computer screen. They were informed that if they did not acquire the target within 15 s, a new target would appear, and the trial was considered a failure. 2.3. Decoding method Four 10 min runs of target acquisition were performed with a rest interval of 1 min between runs. The decoding method employed in this study has been previously described (Bradberry et a! 2010) so will only 2.2. Data acquisition briefly be described here. First, a fourth-order, low-pass A 64-sensor Electro-Cap was placed on the head according Butterworth filter with a cutoff frequency of I Hz was applied to the extended International 10-20 system with ear-linked to the kinematic and EEG data. Very low frequencies have reference and used to collect 58 channels of EEG activity. previously been shown to possess kinematic information (Jerbi Continuous EEG signals were sampled at 100Hz and amplified eta! 2OO7, Walden et al 2(08, Bradberry eta! 2O09), including 1000 times via a Synamps I acquisition system and Neuroscan those from low-pass filtered electrocorticographic signals (the v43 software. Additionally, the EEG signals were band- local motor potential, LMP) (Schalk et al 2007). Next, the pass filtered from 0.01 to 30 Hz. Electroocular (EGG) first-order temporal difference of the EEG data was computed. 3 EFTA_R1_O2O36191 EFTA02693112 I. Neural Eng. 8 (2011) 036010 Ti Bradbeny et al To continuously decode cursor velocity from the EEG signals. 2.5. Source estimation with sLORETA a linear decoding model was employed: To better estimate the sources of cursor velocity encoding. N L X[I)- - 11= ax + EEbakx Sat - kl (I) we used standardized low-resolution brain electromagnetic tomography (sLORETA) (Pascual-Marqui 2(102) software st=l 4.0 version 20081104. Preprocessed (low-pass filtered and N L differenced) EEG signals from all 34 channels for each subject y(t)- yrt — II= ay + EENkys„ — kI. (2) were fed to sLORETA to estimate current sources. First, r n=l 4=0 values were computed between the squared time series of each where x [t]-41- I ) and y — I I are the horizontal and of the 34 sensors with the 6239 time series from the sLORETA vertical velocities of the cursor at time sample r respectively, solution and then averaged across subjects. Second, the mean N is the number of EEG sensors, L (=I I) is the number of of the r values multiplied by the regression weights b (from time lags. S„ [I — k] is the temporal difference in voltage equations ( I ) and (2)) of their associated sensors were assigned measured at EEG sensor n at time lag k. and the a and b to each voxel. The regression weights had been pulled from the variables are the weights obtained through multiple linear regression solution at the time lag with maximum %T. which regression. Only the most important sensors (N = 34) for had the highest percentage of reconstruction contribution. velocity reconstruction found in a previous study (Bradberry Third, for visualization purposes, the upper quartile of voxels et al 2010), which excluded the three most frontal sensors, (r values weighted by b) was set to the value one, and the rest of were used for decoding. the r values were set to zero. Finally these binary-thresholded For the calibration phase, a (10 x 10) -fold r values were plotted onto a surface model of the brain. cross-validation procedure was employed to assess the reconstruction accuracy of observed cursor velocity from EEG 2.6. Eye and muscle activity analysis signals. In this procedure, the entire continuous data were divided into 10 parts: 9 parts were used for training, and the To assess the contribution of eye activity to decoding, the remaining part was used for testing. The cross-validation decoding procedure was executed off-line with channels of procedure was considered complete when each of the ten standardized vertical and horizontal EOG activity included combinations of training and testing data were exhausted. with the 34 channels of standardized EEG activity. The and the mean Pearson correlation coefficient (r) between percent contribution of these cyc channels was then assessed measured and reconstructed kinematics was computed across by dividing the absolute value of their regression weights by folds. Prior to computing r. the kinematic signals were the sum of the absolute value of all the regression weights. smoothed with a fourth-order, low-pass Butterworth filter with To assess whether muscle activity inadvertently aided cursor a cutoff frequency of I Hz. For the ensuing practice and target control, we cross-correlated EMG signals from flexor and acquisition phases. the regression weights (a and b variables) extensor muscles of the right forearm with the .r and y for the cross-validation fold with the highest r were used for components of cursor velocity over 200 positive and negative online decoding. lags (-2 s to 2 s in increments of 10 ms). The start of the EMG and EEG/EOG recordings were not synchronized by 2.4. Scalp maps ofsensor contributions computer, which is why the cross-correlation of the EMG and EOG signals at different lags was examined as opposed to To graphically assess the relative contributions of scalp regions only the zero-lag correlation. Prior to the cross-correlation, to the reconstruction ofcursor velocity, the decoding procedure the EMG signals were decimated 20 times after applying described in the section above was run on standardized EEG a 40 Hz low-pass antialiasing filter; rectified by taking the signals. and the across-subject mean of the magnitude of the absolute value: low-pass filtered with a fourth-order, low-pass best b vectors (from equations (I) and (2)) was projected Butterworth filter at I Hz: and first-order differenced. onto a time series (-110-0 ms in increments of 10 ms) of scalp maps. These spatial renderings of sensor contributions were produced by the topoplot function of EEGLAB (Delorme 3. Results and Makcig 2004), an open-source MATLAB toolbox for 3.1. Calibrating a neural decoderfront observed cursor electrophysiological data processing that performs bihannonic movement spline interpolation (Sandwell 1987) of the sensor values before plotting them. To examine which time lags were BC1 systems are ultimately intended for movement-impaired the most important for decoding. for each scalp map. the persons; therefore, it is imperative that calibration of the neural percentage of reconstruction contribution was defined as decoder does not require overt movement. For this reason, we N calibrated our previously developed decoder (Bradberry et al E 2010) in a manner similar to that described in an invasive BCI sbr, = 100% x a= (3) study (Hochberg et al 2006) that required only motor imagery E E 17bLi 44;:ky during observation of cursor movement. More specifically, n=l during the calibration phase of our study, subjects imagined for all i from 0 to Is, where %T; is the percentage of using their finger to track biologically plausible movement of reconstruction contribution for a scalp map at time lag i. a computer-controlled cursor for 10 min. and we subsequently 4 EFTA_R1_02038182 EFTA02693113 I. Neural Eng. 8 (2011) 036010 T1Bradbeny et al (A) 0.9 (B) 4, X Velocity 05 02 0$ a +0 10 20 fi r) 40 60 60 • 0.4 Y Velocity 03 0.2 0.I 0 10 20 30 40 50 60 8400•02 44410:13 19.144414 Sublicte Won lime(s) Flgure 3. EEG decoding accuracy of observed cursor velocity during the calibration phase. (A) We computed the mean standard error (SE) of the decoding accuracies (r values) across crass-validation folds (n = 10) for each subject for x (black) and y (white) cursor velocities. (B) Superimposed reconstructed velocity profiles (red) and actual velocity profiles (black) matched well (data from subject 1). Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 ‘^s Figure 4. Mean hrain-contna led cursor paths. Each colored path is the mean of he length-normalized trials for a single direction (left, top. right, or bottom) across all trials of all runs for a subject. Trials in which subjects did not acquire the target within 15 s were excluded from analysis. Table I. Mean (SE) of the hit rate and median MT for each target of each subject across runs = 4). Left Top Right Bottom Mean (SE) Hat% MT Hu% MT MT Hu% MT Hit% MT Subject I 94 (2) 4.24 66 (8) 5.90 98 (2) 4.62 55 (9) 8.88 78(11) 5.91 (1.05) Subject 2 83 (5) 6.52 96 (4) 4.40 85 (2) 3.76 85 (4) 4.40 87 (3) 4.77 (0.60) Subject 3 84 (9) 4.24 45 (4) 9.96 100 (0) 2.32 67 (9) 6.82 74 (12) 5.83 (1.65) Subject 4 71 (7) 3.40 33 (7) 4.88 65 (6) 8.16 21 (4) 6.68 47 (12) 5.78 (1.04) Subject 5 57 (14) 8.56 100(0) 2.72 60(18) 5.48 100(0) 2.00 79 (12) 4.69 (1.49) Mean (SE) 78 (6) 5.39 (1.06) 68 (13) 5.57 (1.35) 81 (8) 4.87 (1.09) 65 (14) 5.76(1.32) 73 (4) 5.40 (0.27) The median MT, instead of the mean (SE) Mt was computed for each direction of each subject because the MT distributions were skewed. computed the parameters of the decoder (--10 min) based on 3.2. Applying the neural decoder to move a computer cursor the cursor velocity and EEG signals. We quantified the accuracy of each subject's calibrated After a subject's neural decoder was calibrated and a —20 min decoder by computing the mean of Pearson's r between actual practice phase with the decoder was performed. the subject and reconstructed cursor velocities across ten cross-validation moved a cursor with his EEG signals to acquire targets that folds (figure 3(A)). The across-subject mean r values for appeared one at a time pseudorandomly at the left, top, horizontal (s) and vertical (y) velocities were 0.68 and 0.50 right, or bottom of a 21) workspace (see movie I. available respectively. indicating high decoding accuracy. In fact, the at .tacks.iop media). Four 10 min runs accuracy was roughly double that of studies that decoded of target acquisition were performed with a rest interval of observed cursor movement from neural activity acquired more 1 min between runs. The length-normalized cursor paths focally with intracranial microelectrode arrays (Kim a al confirmed the subjects' ability to move from the center to 2008. Truccolo a al 2008). Reconstructed velocity profiles the target (figure 4). For each target of each subject, the target also visually matched well with the actual velocity profiles hit rate and movement time (MT) across runs are given in (figure 3(B)). table I. The overall means SE of the hit rate and MT were 5 EFTA_R1_02036193 EFTA02693114 1. Neural Eng. 8 (2011) 036010 T l Bradbeny et al Cursor Movement Observation (A) 62% &I% 52% 77% 101% .122 • • • • • • I I I .110 -100 -90 -80 .70 -80 a 8 12 4%, 715.A 122% 65% • 1% 55% 410 • • • • -50 -40 1 .30 ms 20 -10 0 Brain-Controlled Cursor Movement ID) • • 1••• (G) • 55% 49% 74% 102% 116% 1 '2), •• • • 0 • • S • I -110 400 1 I 1 -80 1 -70 1 -60 .." % ; - _ Xi% ;,/ • zz r;.•;% O C> 121% 11.1% 9.0% 5 2% 49% 56% ig FP) . -50 • -40 • -30 me • -20 . -10 • 0 e Figure 5. Neural regions that encoded cursor movement. (A) Scalp sensor contributions to the reconstruction of observed cursor velocity during the calibration phase. Mean (ti = 5) scalp maps of the sensors revealed a network of frontal, central and parietal involvement. In particular, sensors Fl. rcz and CPI-CP4 of the International 10/10 system made the largest contribution. Light and dark colors represent high and low contributors, respectively. Each scalp map with its percentage contribution is displayed above its associated 10 ms time lag. revealing the 12.4% maximal contribution of EEG data at 50 ms in the past. (B) Sources that maximally encoded observed cursor velocity during the calibration phase. We overlaid localized sources (yellow) from 50 ms in the past onto a model of the brain in different orientations to reveal the involvement of the PrG (I), PoG (2), LPM (3). STS (4), and dorsal and ventral LPC (5). (C) Scalp sensor contributions to the brain-controlled cursor velocity during the target acquisition phase. Mean (n = 5) scalp maps of the sensors weights from the subjects' best runs revealed a network that had shifted to involve more central regions than the network of the calibration phase. The scalp maps revealed a 12.1% maximal contribution of EEG data at 50 ms in the past. (D) Sources that maximally encoded brain-controlled cursor velocity during the target acquisition phase. Localized sources (yellow) from 50 ms in the past revealed a substantial involvement of PrG (I) and PoG (2) and some involvement of LPM (3). As in the calibration phase. the STS (4) was involved. In contrast to the calibration phase. the LPC (5) played a minor role, and the IPL (6) played a major role. 73 4% and 5.4O 0.27 s. The change in hit rate across runs (figure 5(A)). Within this network, sensors over the is presented for each subject in figure s2 in the supplementary frontocentral and primary sensorimotor cortices made the data, available at slacks.ior.org/JNE/8/0360I Wnimed i greatest contribution. Concerning time lags. EEG data from 50 ms in the past supplied the most information. In source 3.3. Neural regions that encoded cursor movement space at 50 ms in the past, the precentral gyms (PrG), postcentral gyms (PoG), lateral premotor (LPM) cortex, To visualize the contributions of scalp regions and current superior temporal sulcus (STS). and dorsal and ventral portions sources to the reconstruction of cursor velocity, the weights of of lateral prefrontal cortex (LPC) played a large role in the the decoder were projected onto scalp maps. and sLORETA encoding of observed cursor velocity (figure 5(B)). (Pascual-Marqui 2002) was employed. Scalp maps of Scalp maps of sensor contributions to the brain-controlled sensor contributions to the reconstruction of observed cursor cursor velocity were generated from the mean ofeach subject's movements in the calibration phase depicted the contributions best run in the target acquisition phase. They depicted the as a network of frontal, central and parietal regions contributions as having shifted to be more focused within 6 EFTA_R1_02036194 EFTA02693115 1. Neural Eng. 8 (2011) 036010 Ti Bra bony et al Table 2. Percent contribution of EOG activity to cursor velocity 4. Discussion reconstruction. Target We report the first EEG-based BCI system that acquisition employs continuous decoding of imagined continuous hand Calibration (best run) movements. Furthermore. we emphasize that the system X Y X Y requires only a single session of decoder calibration (--20 min) and subject practice (-20 min) before subjects Subject 1 0.30 1.58 0.00 0.01 can operate it. The off-line decoding results of the calibration Subject 2 0.00 0.01 0.20 0.18 Subject 3 1.99 9.60 1.54 047 phase that used observation of biologically plausible cursor Subject 4 0.00 0.01 94.9 0.04 movement were higher than those of invasive BCI studies Subject 5 0.34 0.65 0.06 0.03 and may imply, as discussed below, the involvement of a widespread MNS in humans. In the on-line target acquisition phase, subjects controlled a cursor with their EEG signals Table 3. Mean (SD) across subjects of maximum absolute r values from cross-correlation of forearm flexor and extensor EMG activity alone with accuracies comparable to other noninvasive and with x and y components of cursor velocity. invasive BCI studies aimed at 2D cursor control. Target acquisition Calibration (best run) 4.1. Comparison to other BCI studies X Our study is the first noninvasive EEG-based BCI study to employ continuous decoding of imagined natural movement. Flexor 0.05 (0.04) 0.05 (0.04) 0.04 (0.02) 0.07 (0.03) Extensor 0.03 (0.02) 0.04 (0.01) 0.07 (0.08) 0.05 (0.04) Previous work in EEG-based BCI systems for cursor control required subjects to overcome an initial disconnect between intended movement and neural activity in order to learn how to modulate their sensorimotor rhythms to control the central regions (figure 5(C)). As in the calibration phase, EEG cursor. These studies based on sensorimotor rhythms required data from 50 ms in the past supplied the most information. In weeks to months of training before levels of performance source space at 50 ms in the past, compared to the calibration were deemed sufficient for reporting (Wolpaw and McFarland phase. a large shift occurred from anterior (frontocentral) to 2004). We believe the fact that we used a decoder based posterior (centroposterior) neural regions. More specifically. on imagined/observed natural movement, as opposed to there was much less involvement of the LPC, the PrG and neurofeedback training of sensorimotor rhythms, reduced the PoG exhibited an even more widespread involvement, and subject training requirements of our target acquisition phase the inferior parietal lobule (IPL) made a large contribution to only a single brief practice session (-20 mint (figure 5(D)). An ECoG study based on sensorimotor rhythms that had objectives similar to ours also observed that several subjects learned to control a 21) cursor over a short period of time 3.4. Eye and muscle contributions (Schalk et al 2008). Although this ECoG study reduced training time compared to previous EEG studies (Wolpaw A concern in BCI studies is that eye or muscle movements and McFarland 2004), some drawbacks included that pre- may contaminate EEG signals thereby inadvertently aiding training time was still taken for the initial selection of control the control of a device/environment that should be controlled features and for training subjects to first move the cursor in by thought-generated neural signals alone. To address this one dimension at a time. We were able to bypass these two concern, we executed the off-line decoding procedure with pre-training steps. Another drawback of the ECoG study was channels of vertical and horizontal EOG activity included, that all five subjects used overt movement for initial selection and assessed the percent contribution of these eye channels of features, and two subjects used overt movement throughout (table 2). The percent contributions were low for the the study. calibration and target acquisition phases except for a very Additionally, the results of our target acquisition phase high percent contribution (94.9%) to .r velocity reconstruction compare favorably to those in tetraplegic humans that were for subject 4 during target acquisition. Interestingly. implanted with intraconical arrays in the arm area of MI this subject had the lowest decoding accuracy of all (Hochberg et al 2006, Kim et al 2008) even though the participants, suggesting that eye movements disrupted performance results of those studies were only computed on data collected weeks to months after training began. Table 4 decoding. Furthermore. the fact that hardly any extreme compares our study to the aforementioned studies. frontal contribution is observed in the scalp maps and sLORETA plots (figure 5) is a testament to the non- contribution of EOG activity to decoding To access whether 4.2. Differential encoding of observed and brain-controlled cursor velocity muscle activity aide) cursor control, we cross-correlated EMG signals from flexor and extensor muscles of the right forearm The most notable differences between the regions that encoded with the x and y components of cursor velocity to find that all for observed cursor velocity and brain-controlled cursor correlations were low (table 3). velocity were with the PIG, PoG, IPL and LPC. There was 7 EFTA_R1_02038195 EFTA02693116 I. Neural Eng. 8(2011) 036010 T1Bradbeny eta! 7Lbµ 4. Comparison to most relevant human BCI studies of 2D cursor control. Number of Target size as% Timeout Movement Target subjects Neural data of workspace (s) time (s) hit% Wolpaw and McFarland 2004 4 EEG 4.9 I0 1.9 92 Hochberg et al 21)06 Single units NA 7 2.5 85 Kim et at 2(Xnc 2 Single units 1.7 3.1 75 Schalk Hal 20)8 5 ECoG 7 16.8 2.4 63 Present study EEG 1.3 Is 5.4 73 a more widespread contribution from the PrG, PoG and 5. Conclusion IPL during brain control, which could reflect the increased involvement of imagined motor execution (Miller et al 20 I()) In the near future, it will be important for whole-arm amputees especially since these regions have previously been shown and persons with impaired upper limb movement (e.g., spinal to be engaged in encoding cursor kinematics (Bradberry cord injury or stroke) to test our noninvasive BCI system since et al 2009. Jerbi et al 2(x7). The contribution from the they are the target population for this assistive technology. LPC was largely attenuated during brain-controlled cursor Since our findings indicate that calibration of our decoder movements, suggesting a transition out of the imitative and initial practice by subjects require a short amount of time learning environment of cursor observation (Vogt et al 2(X07). in a single session, we expect to avoid burdening patients with lengthy training. Employing our method will also permit future i