Simultaneous Neural Control of Simple Reaching and Grasping With the Modular Prosthetic Limb Using Intracranial EEG

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 22, NO. 3, MAY 2014 695

Simultaneous Neural Control of Simple Reachingand Grasping With the Modular Prosthetic Limb

Using Intracranial EEGMatthew S. Fifer, Student Member, IEEE, Guy Hotson, Brock A. Wester, Member, IEEE,

David P. McMullen, Student Member, IEEE, Yujing Wang, Matthew S. Johannes, Member, IEEE,Kapil D. Katyal, Member, IEEE, John B. Helder, Member, IEEE, Matthew P. Para, Member, IEEE,R. Jacob Vogelstein, Member, IEEE, William S. Anderson, Nitish V. Thakor, Fellow, IEEE, and

Nathan E. Crone, Senior Member, IEEE

Abstract—Intracranial electroencephalographic (iEEG) signalsfrom two human subjects were used to achieve simultaneousneural control of reaching and grasping movements with theJohns Hopkins University Applied Physics Lab (JHU/APL) Mod-ular Prosthetic Limb (MPL), a dexterous robotic prosthetic arm.We performed functional mapping of high gamma activity whilethe subject made reaching and grasping movements to identifytask-selective electrodes. Independent, online control of reachingand grasping was then achieved using high gamma activity from asmall subset of electrodes with a model trained on short blocks ofreaching and grasping with no further adaptation. Classificationaccuracy did not decline (p < 0.05, one-way ANOVA) over threeblocks of testing in either subject. Mean classification accuracyduring independently executed overt reach and grasp movementsfor (Subject 1, Subject 2) were (0.85, 0.81) and (0.80, 0.96), respec-tively, and during simultaneous execution they were (0.83, 0.88)and (0.58, 0.88), respectively. Our models leveraged knowledgeof the subject’s individual functional neuroanatomy for reachingand grasping movements, allowing rapid acquisition of control ina time-sensitive clinical setting. We demonstrate the potential fea-sibility of verifying functionally meaningful iEEG-based control ofthe MPL prior to chronic implantation, during which additionalcapabilities of the MPL might be exploited with further training.

Manuscript received February 06, 2013; revised August 05, 2013; acceptedOctober 13, 2013. Date of publication October 24, 2013; date of currentversion April 28, 2014. Data analysis and online control were supported in partby the National Institute of Neurological Disorders and Stroke under Grant3R01NS0405956-09S1, in part by DARPA under Contract 19GM-1088724,and in part by the National Institute of Biomedical Imaging and Bioengineeringunder Grant 5T32EB003383-08. The MPL was developed with funds fromDARPA under Contract N66001-10-C-4056. M. S. Fifer and G. Hotson areco-first authors of this publication.This paper has supplementary downloadable material available at http://iee-

explore.ieee.org, provided by the authors.M. S. Fifer, Y. Wang, and N. V. Thakor are with the Department of

Biomedical Engineering, Johns Hopkins University, Baltimore, MD 21205USA (e-mail: [email protected]).G. Hotson is with the Department of Electrical and Computer Engineering,

Johns Hopkins University, Baltimore, MD 21218 USA.B. A. Wester, M. S. Johannes, K. D. Katyal, J. B. Helder, M. P. Para, and R. J.

Vogelstein are with the Research and Exploratory Development Division, JohnsHopkins University Applied Physics Laboratory, Laurel, MD 20723 USA.D. P. McMullen and W. S. Anderson are with the Department of Neuro-

surgery, Johns Hopkins University, Baltimore, MD 21287 USA.N. V. Thakor is also affiliated with the SINAPSE Institute, National Univer-

sity of Singapore, Singapore.N. E. Crone is with the Department of Neurology, Johns Hopkins University,

Baltimore, MD 21205 USA.Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/TNSRE.2013.2286955

Index Terms—Brain–machine interface (BMI), electrocorticog-raphy, functional mapping, high gamma, upper limb prosthesis.

I. INTRODUCTION

R EACHING to and grasping objects is an important skillthat forms the basis for many activities of daily living

(ADLs). It is thus an important target for brain-machine inter-faces (BMIs) being developed for patients with impaired limbfunction due to neurological lesions of motor pathways (e.g.,spinal cord injury, amyotrophic lateral sclerosis, stroke, etc.).Recent work has demonstrated that grasp types [1], grasp timing[2], hand postures [3], and reach parameters [4] can be decodedfrom spectral changes in human intracranial electroencephalo-graphic (iEEG) signals, and that movement-related spectralmodulation of iEEG can be used for online control of BMIs, forexample during dexterous grasping [5], when selecting betweengrasp types and elbow movement [6], or for three dimensionalcursor control [7]. We therefore sought to determine whetherhuman iEEG could be used to provide simultaneous and inde-pendent online control of reaching and grasping movements,thus demonstrating segregation of these two movement types atthe spatial scale of iEEG macroelectrodes. iEEG is an attractiveplatform for the development of BMIs because of the potentialfor better long-term signal stability than multi-unit recordings,as well as the availability of subjects who have accepted therisks of electrode implantation for the mapping of their seizureonset zones prior to epileptic resection surgery [8].Relative to scalp EEG, iEEG provides better spatial res-

olution and better signal quality for high frequency activity[9], [10]. There is substantial empirical evidence from localfield potential studies in humans [11] and nonhuman primates[12] that this high frequency activity closely tracks populationfiring rates. The degree of control that can be achieved withthe large-scale population activity recorded with iEEG [13],[14] is unknown, however, especially with chronic trainingbeyond the time constraints of seizure monitoring. To ensurethat the risk of long-term electrode implantation is offset by thebenefit of stable long-term brain–machine interface (BMI) use,it would be advantageous to confirm at least basic control of theintended prosthetic at the time of implantation. Previous workin scalp EEG and iEEG has demonstrated 2-D and 3-D cursor

1534-4320 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

696 IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING, VOL. 22, NO. 3, MAY 2014

Fig. 1. Functional mapping of cue-averaged task-related high gamma activity in training set. (A) Reconstruction of the implanted grid location for Subject 1 isdepicted; the electrode used for reaching (number 25) is highlighted in red and corresponds to the channel circled in red in the activation maps below, while theelectrode used for grasping (number 11) is highlighted in blue and similarly corresponds to the electrode circled in blue below; the central sulcus is highlightedin green. (B) Reconstruction of the depth electrodes implanted in right hemisphere of Subject 2; electrodes used for reaching highlighted in red, electrodes usedfor grasping highlighted in blue (transparent medial view in inset). (C, D) Each task map displays the spatiotemporal distribution of significant increases (redspectrum) or decreases (blue spectrum) in high gamma energy relative to pre-cue baseline in 16 ms windows for Subject 1. Each row corresponds to a differentiEEG electrode in the frontoparietal grid displayed in (A). All times are relative to cue onset. (E) A differential map is shown for Subject 1, which is the result of aWilcoxon test between two conditions for each (channel, time) pair with FDR correction for comparisons across multiple time points within each channel. Channeland time pairs are in the red spectrum if forward reach is more activated than grasp, and in the blue spectrum if grasp is more activated than forward reach. Theaverage times of relevant behavioral events are marked with vertical lines and labeled (movement onset, MO; pressed target button, PT; released target button, RT;returned arm to home position, Home; released pressure bulb, Rest).

control where at least one dimension is controlled by behaviorunrelated to the task at hand (e.g., vocalization or tonguemovement) [13]–[16]. Although it has been demonstrated thattraining and “operant conditioning” can be used to learn BMIcontrol on the time-scale of months [17], it is unclear to whatextent an unnatural mapping will scale up to more complextasks in more complex environments. We therefore sought todetermine whether the command signals for forward reachingand grasping of the Johns Hopkins University Applied PhysicsLab (JHU/APL) Modular Prosthetic Limb (MPL) could bederived from high frequency (70–110 Hz) neural populationactivity associated with naturalistic reaching and graspingmovements, respectively. These commands were interpretedby the hardware in the MPL and converted to multi-axialanthropomorphic movements spanning two controllable jointsfor forward reaching and 10 controllable joints for grasping.

II. METHODS

A. Subject Info

The subjects for this study were 55-year-old (Subject 1) and30-year-old (Subject 2) right-handed males implanted withintracranial electrodes to map the ictal onset zone of medicallyresistant seizures prior to surgical resection. In Subject 1, an8 8 grid of subdural platinum-iridium electrodes (Adtech,Racine, WI, USA, 2.3-mm-diameter exposed surface, 1-cmspacing) were surgically implanted over right frontal-parietalregions (see Fig. 1), in addition to a 4 5 electrode gridover right lateral occipital cortex and a 1 8 electrode stripstretching from right mid-temporal regions to dorsolateralprefrontal cortex (both not shown). Subject 2 was implanted

with a 1 8 electrode strip (Adtech; Racine, WI, USA; asabove) across right frontoparietal cortex, six depth electrodeswith eight platinum macrocontacts each (Adtech; 2.41 mmlong, 6.5 mm center-to-center spacing) placed medially fromthe right premotor area to the posterior parietal lobe, andone hybrid depth with eight platinum macrocontacts (Adtech;1.57 mm long, 5 mm center-to-center) and 16 microcontacts(Adtech; 75 m diameter). Neuronavigation via the CranialNavigation Application (BrainLab; Westchester, IL, USA) wasused during placement of the depth electrodes in Subject 2.Anatomical reconstructions of the subjects’ brains with thelocation of implanted electrodes were generated by volumet-rically co-registering the pre-surgical MRI with a postsurgicalCT using BioImage [Fig. 1(A)] [18]. Subject 1’s seizures beganafter a bout of viral encephalitis with coma at 33 years ofage. His complex partial seizures were typically preceded bya somatosensory aura in his left hand with spread to the faceand subsequent shaking of the left hand, and were sometimesfollowed by secondary generalization. Subject 2 had previouslyundergone chronic recording with partial resection of his rightpost-central gyrus and superior parietal lobule. Both patientsgave informed consent for research testing, which was donein accordance with a protocol approved by the InstitutionalReview Board of the Johns Hopkins Medical Institutions.

B. Neural Signal Acquisition

Using a NeuroPort system (BlackRock Microsystems; SaltLake City, UT, USA), iEEG signals were initially sampled at30 KHzwith an analog bandpass filter with cutoffs of 0.3 Hz and7500 Hz. The NeuroPort system then applied a digital fourth-

FIFER et al.: SIMULTANEOUS NEURAL CONTROL OF SIMPLE REACHING AND GRASPING WITH THE MODULAR PROSTHETIC LIMB 697

Fig. 2. Schematics and photographs of experimental setup with MPL. (top) Schematic of the experimental setup is shown, with Subject 1 seated and interactingwith three behavioral sensors. MPL is to the front and right of the subject, in the same room as and in full view of the subject. Traces of the behavioral sensors,high gamma power, and MPL commands during a three trial segment are shown as an example. (A–C) Subject is seated on his hospital bed (not pictured, right ofview), with his arm at rest on a lap desk with inset pushbutton or “home switch.” Subject is holding but not actively grasping the squeeze bulb used to query graspstatus. On the subject’s hospital tray are a pushbutton for reach offset detection, and a laptop displaying a red bar indicating pressure exerted on the squeeze bulb.(A) In the background, the MPL is at its baseline state (rest posture). (B) Subject is executing a grasp movement, and (C) subject is executing a reach movement.

order Butterworth lowpass filter with a 250 Hz cutoff and down-sampled to 1000 Hz. Artifactual channels were visually identi-fied and excluded from all further analysis. Acquired iEEG sig-nals were broadcast over UDP to an experimental workstation,where they could be accessed for online spectral feature extrac-tion and model evaluation to drive the MPL.

C. Experimental Procedures

Short offline data sets of 30 (Subject 1) or 50 (Subject 2) au-ditorily cued trials were collected each for forward reaches andgrasping movements. Audio cues of “reach” or “grasp” weredelivered via external speakers by E-Prime software (PST, Inc.,Sharpsburg, PA, USA). For Subject 2 only, the reach and grasp

trials were interspersed with “Reach and Grasp” trials, of whichthere were also 50. The onset of each trial was manually initi-ated by the experimenter to ensure that the preceding reach wascompleted and an additional varying delay had passed beforea cue was given. Behavioral states were detected using analogsensors sampled at 1000 Hz on the same hardware as the neuraldata: 1) the onset and offset of each reachingmovement were de-tected using a pushbutton embedded in a wooden lap desk, 2) thetermination of each reach on a distal target was detected usinga pushbutton, and 3) the onset and offset of each grasp were de-tected using a pneumatic squeeze bulb connected via flexibletubing to an electronic pressure sensor. A detailed schematicof the experimental setup is included in Fig. 2, and a video of


the training procedure is shown for Subject 1 in SupplementalVideo 1. Reaches by Subject 1 ranged in duration from 1.3 to1.8 s (median s) with response latencies ranging from330 to 500 ms (median ms), while grasps (i.e., as de-tected by the squeeze bulb) ranged in duration from 0.6 to 2.7 s(median s) with response latencies ranging from 380 to930 ms (median ms). Reaches by Subject 2 ranged induration from 1.9 to 4.8 s (median s), with response la-tencies ranging from 450 to 1450 ms (median ms), whilegrasps ranged in duration from 0.8 to 3.5 s (median s)with response latencies ranging from 640 to 2070 ms (median

ms).

D. iEEG Electrode Evaluation

Following collection of the reach and grasp datasets,event-related high gamma activations were analyzed. Theaudio cue played to the subject was split and fed into theBlackRock system; the beginning of this cue was detected andused as a stimulus onset (SO) marker. The 1024 ms prior to SOwas pooled into baseline distributions for each channel, whilethe 3072 ms following the onset of the audio cue was used as apost-stimulus epoch. The 1024 ms prior to SO and 3072 ms fol-lowing SO were segmented into 128 ms windows with 112 msoverlap. Each window was reduced to a single estimate of thehigh gamma analytic amplitude in a 16 ms bin using a Hilberttransform with an embedded, flat-top Gaussian bandpass filterwith cutoffs of 72 and 110 Hz. Separate distributions werecreated for each post-stimulus 16 ms time bin and channeland referenced to the channel baseline distributions usingtwo-sample tests with significance threshold . Thethresholds for p-value significance of these tests were correctedfor multiple comparisons within each channel using the falsediscovery rate (FDR) correction [19]. Any resulting significantp-values were then transformed, and any significant mod-ulation was labeled as an increase or a decrease. This resultingmatrix of statistical significance measures therefore containedtiming information about activation that was used to excludechannels which displayed modulation in response to the audiocue. This entire analysis was performed with custom MATLAB(MathWorks, Inc., Natick, MA, USA) software, from whichthe results were available within the experimental session (seeFig. 1).

E. BMI Model Training

For Subject 1, a final training set was recorded in which theverbal commands “reach,” “grasp,” and “reach and grasp” werepseudo-randomly chosen and played to the subject via externalspeakers with E-Prime; this training set contained 46 trials andlasted approximately five minutes. For Subject 2, the 150 trialsspanning approximately sixteen minutes collected for electrodeevaluation were used as a training set. Also for Subject 2, theinitially trained model was used to drive a virtual version ofthe MPL as visual feedback during an additional 120 trials (i.e.,40 each of “reach,” “grasp,” and “reach and grasp”). The iEEGand behavioral data recorded during this block were used as thetraining set for online testing.

Signals in each training set were first spatially filtered with acommon average [20] of all channels not excluded by visual in-spection because of artifact or noise. Autoregressive power wasextracted from the streamed signals using the Burg algorithmwith model order 16 on a 400 ms window. The logarithm ofthe spectral power from components between 72.5 and 110 Hzwere then averaged to yield an estimate of the broadband highgamma power. In offline data collection for model training pur-poses, feature extraction windows were overlapped by 300 ms.In Subject 1, one electrode each was chosen for reach

and grasp using information from the functional maps ofpost-stimulus activation. The high gamma log-powers duringmovement and rest movement were compared to manuallyestablish a threshold for movement classification. In Subject 2,four channels each were selected as model inputs to separatebinary linear discriminant analysis (LDA) classifiers for reachand grasp. In addition, transition probabilities were adjustedmanually before the testing session to smooth the output fromthe classifier. For this study, we used a probability of 0.95 forthe probability of a rest classification if currently at rest (i.e.,0.05 for a movement classification), and 0.8 for the probabilityof a movement classification if currently in the movement state(i.e., 0.2 for a rest classification).

F. JHU/APL Modular Prosthetic Limb

Developed by JHU/APL under the Defense Advanced Re-search Project Agency (DARPA) Revolutionizing ProstheticsProgram, the MPL (Supplemental Fig. 1) is an advanced upper-body extremity prosthetic and human rehabilitation device [21].The MPL has 17 controllable degrees-of-freedom (DoF) and 26articulating DoF in total (Supplemental Fig. 1, with specifica-tions and architecture details in Supplementary Methods). Tofacilitate control from neural decoded motion intent, the MPLhas a custom software interface, VulcanX, that receives move-ment/motion commands locally and sends them over a con-troller area network (CAN) bus to a limb controller (LC) boardin the hand of the MPL [22]. Three types of high-level con-trol commands, passed through VulcanX, are fused together toform individual actuator commands by the LC: 1) degree of mo-tion control (DOM) commands, which allow each degree of mo-tion to be controlled individually with position and/or velocitycommands; 2) endpoint control (EP) commands, which allowthe hand’s position and orientation to be controlled in Cartesianspace using a Jacobian-based algorithm for computing inversekinematics; and 3) reduced order control (ROC) commands,which allow pre-programmed hand grasp patterns to be actu-ated in a coordinated fashion as a single degree of freedom [23],[24]. EP velocity and ROC commands were utilized to controlreach and grasp, respectively, in this study.

G. Online Testing

Once the high gamma thresholds for movement were estab-lished, classification outputs from the trained models of reachand grasp movements were simultaneously used to actuate theMPL via the VulcanX interface. Whenever the classifiers pre-dicted that the subject was reaching and/or grasping, the MPLwas commanded to reach and/or grasp, respectively, at a set rate.If either the reach or grasp classifier predicted that the subject


was resting, the limb was commanded to return to its rest arm orhand posture, respectively, at an equal rate. For Subject 2 only,the return rate for reaching was adjusted to be 50% higher thanthe forward rate. High gamma log-power calculations were per-formed in 400 ms windows (i.e., as in training) computed asquickly as possible on the streaming iEEG signals to provideinputs to the trained model (i.e., 11 ms for Subject 1, slowedto 32 ms for Subject 2 purposefully to avoid inundation of theMPL). Both subjects completed three blocks of online trials byperforming the same overt movements with their native limbs asduring the training set. In Subject 1 only, the second and thirdblocks were separated by a battery of physical and imaginedmovements that were not analyzed as a part of this study.

H. Quantitative Evaluation of Control

The physical movement blocks lasted approximately 4, 11,and 13 min for Subject 1 and 11, 15, and 10 min for Subject 2(respectively). The MPL VulcanX control software created alog of commands sent to the limb with timestamps, which wascompared offline to the timestamps of salient cues and behav-ioral events recorded by the BlackRock system (e.g., subjectleaves the home switch, subject grasps the squeeze bulb, etc.).Trials were designated as starting 500 ms prior to the earliestof the reach and/or grasp onsets and ending 500 ms prior to theonset of the next trial. For each trial, we recorded the propor-tion of correct commands (e.g., the percentage of “grasp” com-mands with a positive velocity when a physical grasp was per-formed) in a window of equal length to the corresponding phys-ical movement duration for that trial. To account for variableresponse latencies by the subject and an inconsistent system la-tency, the start of the window relative to the onset of the trialwas selected individually for each trial to maximize the accu-racy. For reach-and-grasp trials, durations and latencies were se-lected separately for the reach and grasp components. As a con-trol, a window whose length equaled the average duration of thereaches or grasps was used to compute the peak reach or graspcommand accuracy in grasp-only and reach-only trials, respec-tively. Accuracy for each trial was computed as the average ofthe single trial sensitivity (i.e., proportion of reach or grasp com-mands within the selected movement window) and the singletrial specificity (i.e., proportion of rest commands outside of theselected movement window). The median reach command ac-curacies for reach-only versus grasp-only and reach-and-graspversus grasp-only and the grasp command accuracies for grasp-only versus reach-only and grasp-only versus reach-and-graspwere compared using a nonparametric two-sidedWilcoxon ranksum test.

III. RESULTS

Both subjects were able to attain a high degree of subjectivecontrol over reaching and grasping with the MPL across the ex-perimental session with no model adaptation while moving theirnative limbs (Supplemental Video 2 and 3). Furthermore, bothsubjects were able to achieve a level of performance throughoutthe experimental session that was qualitatively similar to thefirst block.We investigated the spectrogram of modulation time-locked

to salient stimuli and behavioral events to validate our choice of

the high gamma band for online control. As shown in the func-tional mapping results (Fig. 1), the electrodes used for control oftheMPL exhibited robust high gammamodulation. Fig. 3 showsthe time-frequency response of the reaching electrode duringreach-only and reach-and-grasp trials of the online task, as wellas the grasping electrode during grasp-only and reach-and-grasptrials. High gamma modulation in the reaching electrode oc-curred within the frequency range of 80–160 Hz for Subject 1, incontrast with the more spectrally restricted 60–120 Hz modula-tion in the grasping electrode. Subject 2 displayed activation at alower frequency range, centralized around 40–90 Hz. These fre-quency ranges of power modulation show that while our choiceof 72.5–110 Hz for control may not have exactly matched theneurophysiological response to the task, it did capture a sub-stantial amount of the power modulation for both tasks. Thetemporal envelope of activation was relatively restricted in thereaching electrode for both subjects, with mean power modu-lation peaking roughly 200 ms before the onset of movement.Subject 2 had similarly tight timing in grasp-related corticalactivation. In contrast, power modulation in the grasping elec-trode of Subject 1 began an average of 300 ms prior to move-ment onset and peaked more than 300 ms after movement onset.The reach-related high gamma power modulation also differedfrom grasp-related power modulation in the presence of twodistinct temporal peaks, time-locked to outward reach and thesubsequent return to rest. Fig. 3 (bottom row) provides verifi-cation that gamma power modulations in the grasp and reachelectrodes were markedly lower during execution of reach andgrasp, respectively.During online control in Subject 1, we observed that control

of grasping was less reliable for reach-and-grasp trials thangrasp-only trials. Fig. 3 (middle row) shows that high gammamodulation in the reaching and grasping electrodes duringreach-and-grasp trials was qualitatively reduced relative toreach-only and grasp-only trials. To evaluate this effect, loghigh gamma power was extracted in 300 ms around the onsetof movement. Statistical analysis revealed that log power in thegrasping electrode around the onset of grasp was significantlyhigher in grasp-only than in reach-and-grasp trials ( ,Wilcoxon test); the log power in the reaching electrode aroundthe onset of reach was not significantly different in reach-onlyand reach-and-grasp trials, however. Identical analyses per-formed in Subject 2 did not reveal any significant differences inmovement-related power modulation between reach-and-grasptrials and either the reach-only or grasp-only trials in any ofthe electrodes used for control ( , Wilcoxon test,Bonferroni-corrected).To evaluate the high gamma power modulation associated

with movement state in reach-only trials, log power was alsoextracted in time windows around the onset of stable hold andthe onset of return, in addition to a baseline window precedingthe cue. For Subject 1, log-power in the reach electrode was sig-nificantly higher in the reach window and return window thanin the hold window, all of which were significantly higher in thebaseline window ( , one-way ANOVA with Tukey’shonestly significant difference post-hoc). In all four electrodesused for reaching control in Subject 2, median hold activity waslower than median reaching and returning activity; the differ-


Fig. 3. Average change of power spectral densities (PSD) relative to baseline, aligned to movement onset. (A) Reach and grasp electrodes are shown for Subject1, and (B) two representative electrodes are shown for Subject 2. First vertical dashed line in all plots corresponds to the average time the audio cue began. For eachtrial, the baseline was chosen from before the onset of the cue (leftmost dashed line). Solid line denotes movement onset (MO). In reach trials, the dashed linesafter the solid line correspond to the average time of the reach completion (pressing target button, PT), release of the target button (RT), and return to home (restingon the home switch), from left to right. Rightmost dashed line in the grasp trials corresponds to the average time of grasp completion. PSD’s were computed viaautoregressive spectral analysis. Window size did not allow for accurate calculations at 0–7.5 Hz, so these frequencies are not displayed.

ence was significant in three out of four electrodes ( ,one-way ANOVA, with Tukey’s honestly significant differencepost-hoc). Reaching, returning, and intermediate hold windowssimilarly exhibited higher levels of high gamma activity thanbaseline windows in Subject 2.Classification accuracy for both reaching and grasping started

and remained high throughout all three blocks of the online task.The mean reach classification accuracy across all trials was 86%(Subject 1) and 82% (Subject 2) for reach-only trials; the reachaccuracy across reach-and-grasp trials was 83% (Subject 1) and89% (Subject 2). The mean grasp classification accuracy acrossall grasp-only trials was 81% (Subject 1) and 96% (Subject2); the grasp accuracy across reach-and-grasp trials was 55%(Subject 1) and 88% (Subject 2). The evolution of classifica-tion accuracies showed no significant effect of block ( ,one-way ANOVA) in either subject. The trial-by-trial reach andgrasp accuracies are depicted in Fig. 4. Reach accuracies weresignificantly higher than chance for both reach-only trials andreach-and-grasp trials ( , Wilcoxon test with Bonferronicorrection), while grasp accuracies were significantly higherthan chance for grasp-only trials ( , Wilcoxon test withBonferroni correction), but not reach-and-grasp trials in Sub-

ject 1 only ( , Wilcoxon test). Grasp accuracies weresignificantly higher in grasp-only trials than in reach-and-grasptrials in both subjects ( , Wilcoxon test). Reach accu-racies were not significantly higher in reach-only trials than inreach-and-grasp trials for Subject 1 ( , Wilcoxon test),although reach accuracies were higher in reach-and-grasp trialsthan in reach-only trials for Subject 2 ( , Wilcoxon test).To investigate whether reaching and grasping were indeed in-

dependent, sham sensitivities were calculated as a control; reachsensitivities were calculated during grasp-only trials and graspsensitivities were calculated during reach-only trials (Fig. 4).Since no physical reaches took place in grasp-only trials, norphysical grasps during reach-only trials, the average reach andgrasp durations were used as surrogates. Peak reach sensitiv-ities were significantly higher in cued reach-only and reach-and-grasp trials than in cued grasp-only trials for both Subjects( , Wilcoxon test); reach sensitivities were significantlyhigher in reach-only trials than in reach-and-grasp trials for Sub-ject 2 ( , Wilcoxon test) but the difference was not sig-nificant in Subject 1 ( , Wilcoxon test). Peak grasp sen-sitivities were higher in cued grasp-only and reach-and-grasptrials than in cued reach-only trials for both subjects ( ,


Fig. 4. Limb performance accuracy metrics. (A, B) Accuracies are shown for reaching and grasping during trials where reach and grasp were executed simul-taneously. (C, D) The reach and grasp accuracies are shown for reach and grasp only trials, respectively. Vertical dashed lines in A–D denote separate blocks.Distributions are shown and summarized with boxplots of the peak sensitivities for grasps in Subject 1 (E), reaches in Subject 1 (F), grasps in Subject 2 (G), andreaches in Subject 2 (H). Each distribution is comprised of the peak sensitivities from each trial. Bars above the boxplots with asterisks mark distributions withsignificantly different medians ( , Wilcoxon test).

Wilcoxon test); grasp sensitivities were significantly higher ingrasp-only trials than in reach-and-grasp trials for Subject 1( , Wilcoxon test) but the difference was not signifi-cant in Subject 2 ( , Wilcoxon test).

IV. DISCUSSION

We were able to provide two human subjects with control ofthe MPL using a control scheme that exploited individual func-tional anatomy, i.e., the population responses in cortical regionsused for control of each subject’s native arm. This allowed oursubjects to achieve control without extensive training. To iden-tify iEEG control sites and characterize their response selec-tivity, we used iEEG functional mapping during reaching andgrasping. By using electrodes over cortical areas that were dif-ferentially activated during reaching and/or grasping, we wereable to afford the patient independent control over the reachingand grasping functionalities of the MPL arm. We showed that

these two movements, when executed individually, elicited cor-tical responses in the gamma band that generalized to their si-multaneous execution, although the same responses occurredwith a reduced magnitude.Additionally, the subject’s control over the arm did not

wane over the course of three separate blocks using thresholdsderived from a short training block. Models were equally effec-tive across blocks with no adaptation or re-training, providingevidence that control was achieved by accurately detecting thenaturalistic circuits for reaching and grasping, not via adapta-tion or operant conditioning. Reach and grasp commands werecontrolled independently, suggesting functional segregationof these movements at the spatial scale of clinically routineiEEG electrodes. There is abundant evidence from experimentsin non-human primates that reaching and grasping engagedifferent networks of cortical areas [25]. As in non-humanprimates, human premotor cortices engaged by reaching arelikely dorsal to those engaged by grasping [26]. As expected,


the iEEG site activated by and used for control of reaching wasdorsal to the site activated by and used for control of grasping.Our BMI used event-related high gamma power augmenta-

tion as an index of task-related neural activity during phys-ical movements. This choice was based on a body of litera-ture which demonstrates that high gamma band modulation isan index of cortical processing in humans [27]–[30] and re-cent experimental evidence that high gamma power changesare strongly and positively correlated with the firing rates ofneuronal populations in close proximity to recording electrodes[11], [31]–[33]. Our findings are consistent with empirical evi-dence that compared with power changes in other frequencies,high gamma power augmentation has high spatial selectivitywith respect to task-related cortical activation, such that adja-cent iEEG electrodes can yield signals with greater indepen-dence at higher frequencies [34], [35]. High gamma responsesare also robust enough to be detected in single trials [36], a nec-essary requirement for BMI applications. Furthermore, severalstudies have shown that high gamma features extracted fromhuman iEEG outperform corresponding lower frequency fea-tures for offline motor decoding [1], [37] and online BMI con-trol [6], [14], [38].This study focused on the control of reach and grasp in the

MPL since they are fundamental to upper limb use, whichprovides a proof of concept for the systems-level integrationgroundwork necessary for more complicated and dexteroustasks. Reaching and grasping movements were decoded foractuation of the MPL with high accuracy and stability; further-more, this was achieved in a clinical epilepsy monitoring settingunder time constraints that did not allow for long-term trainingor testing. Although this prohibited testing the long-termstability of MPL control, it did demonstrate the feasibility ofobtaining MPL control within a compressed timeframe, whichcould have important clinical benefits. Specifically, it would behighly advantageous to demonstrate acceptable brain control ofa neuroprosthetic at the time of surgical implantation, in orderto verify the placement of electrodes and troubleshoot anytechnical difficulties at the time of the operation. Noninvasivemethods of functional mapping (e.g., fMRI) can be used toperform gross surgical planning, but intraoperative verificationof control with iEEG would be extremely useful to refinethe final implantation site. This would help to avoid the needfor re-implantation because the patient is unable to controlthe neuroprosthetic. This would be both costly and increasesurgical risk. Although the total time for our experiment waslonger than that of an awake craniotomy, most of this time wasdue to experimental setup and troubleshooting, and thus couldbe reduced with additional practice.We observed during online testing with the subject that it

was fairly common for the MPL to exhibit a secondary reachor grasp as the subject returned to the resting position. Thiscorresponded to a burst of high gamma activity as the subjectinitiated return of his limb to the home switch or as the sub-ject relaxed his hand after squeezing the bulb. This was bestdemonstrated in the reaching trials by post-hoc offline analysisof the high gamma power in windows associated with reachingto, holding at, and returning from the distal reach target, whichdemonstrated a higher degree of modulation for reaching and

returning than for the intermediate holding in the reaching elec-trode for Subject 1 and for a subset of the reaching electrodes inSubject 2.This report provides additional evidence for the potential

utility of iEEG as a source of control signals for BMIs. Al-though the participants in this study did not suffer from upperlimb paralysis, we believe that the technique of rapid trial-av-eraged spatiotemporal mapping of high gamma modulationcan be used to identify sites that are activated when subjectswith motor impairments attempt to perform movements. Thesepatients often have residual motor function and could attemptto move with assistance, be moved passively, or observe upperlimb movements in a trial-based framework.A large amount of decoding and BMI success has been

achieved using command signals derived from iEEG [6], [7],[13], [14], [38]. Although iEEG macroelectrodes [13], [14],iEEG microelectrodes [39], and multi-electrode arrays [40],[41] have all been used to demonstrate effective BMI controlin small populations, no large-scale longitudinal studies havecompared the tissue response and control performance betweenthese classes of implants. Much previous work has illustrated asignificant redundancy of motor encoding at the single neuronlevel [42], suggesting that population activity could be usefulfor prosthetic control. Nevertheless, there is evidence fromstudies in motor, perceptual, and cognitive systems that therichness of encoding increases with improvements in spatialresolution (i.e., iEEG macroelectrodes exhibit coarser encodingthan iEEG microelectrodes, and iEEG microelectrodes exhibitcoarser encoding than local field potentials from multi-elec-trode arrays) [43]. It is possible that as the spatial resolutionof iEEG implants improves and more comparative studiesare done between iEEG and multi-electrode arrays, that iEEGimplants for BMI control will be an attractive option for somepatients [44], [45]. In the meantime, iEEG recordings in pa-tients undergoing epilepsy surgery will continue to serve as aplatform for demonstrating the degree of useful control thatcan be achieved without extensive training, prior to chronicimplantation of iEEG electrodes for BMIs.

ACKNOWLEDGMENT

The authors would like to thank H. Benz, A. Korzeniewska,Z. Huff, and G. Milsap for lab meeting discussions of our ap-proach and analysis, H. Conner for building the stand used tohold the MPL, and C. Schuman for constructing the mountinginterface.

REFERENCES

[1] T. Pistohl, A. Schulze-Bonhage, A. Aertsen, C. Mehring, and T. Ball,“Decoding natural grasp types from human ECoG,” Neuroimage, vol.59, pp. 248–260, Jan. 2, 2012.

[2] T. Pistohl, T. S. Schmidt, T. Ball, A. Schulze-Bonhage, A. Aertsen,and C. Mehring, “Grasp detection from human ECoG during naturalreach-to-grasp movements,” PLoS One, vol. 8, p. e54658, 2013.

[3] C. A. Chestek, V. Gilja, C. H. Blabe, B. L. Foster, K. V. Shenoy, J.Parvizi, and J. M. Henderson, “Hand posture classification using elec-trocorticography signals in the gamma band over human sensorimotorbrain areas,” J. Neural. Eng., vol. 10, p. 026002, Jan. 31, 2013.

[4] N. R. Anderson, T. Blakely, G. Schalk, E. C. Leuthardt, and D. W.Moran, “Electrocorticographic (ECoG) correlates of human armmove-ments,” Exp. Brain. Res., vol. 223, pp. 1–10, Nov. 2012.


[5] R. Vinjamuri, D. J. Weber, Z. H. Mao, J. L. Collinger, A. D. Degen-hart, J. W. Kelly, M. L. Boninger, E. C. Tyler-Kabara, and W. Wang,“Toward synergy-based brain-machine interfaces,” IEEE Trans. Inf.Technol. Biomed., vol. 15, no. 5, pp. 726–736, Sep. 2011.

[6] T. Yanagisawa, M. Hirata, Y. Saitoh, H. Kishima, K. Matsushita, T.Goto, R. Fukuma, H. Yokoi, Y. Kamitani, and T. Yoshimine, “Electro-corticographic control of a prosthetic arm in paralyzed patients,” Ann.Neurol., vol. 71, pp. 353–361, Mar. 2012.

[7] W. Wang, J. L. Collinger, A. D. Degenhart, E. C. Tyler-Kabara, A. B.Schwartz, D. W. Moran, D. J. Weber, B. Wodlinger, R. K. Vinjamuri,R. C. Ashmore, J. W. Kelly, and M. L. Boninger, “An electrocortico-graphic brain interface in an individual with tetraplegia,” PLoS One,vol. 8, p. e55344, 2013.

[8] E. C. Leuthardt, G. Schalk, D. Moran, and J. G. Ojemann, “Theemerging world of motor neuroprosthetics: A neurosurgical perspec-tive,” Neurosurgery, vol. 59, pp. 1–14, Jul. 2006.

[9] W. J. Freeman, M. D. Holmes, B. C. Burke, and S. Vanhatalo, “Spatialspectra of scalp EEG and EMG from awake humans,” Clin. Neuro-physiol., vol. 114, pp. 1053–1068, Jun. 2003.

[10] T. Ball, M. Kern, I. Mutschler, A. Aertsen, and A. Schulze-Bonhage,“Signal quality of simultaneously recorded invasive and non-invasiveEEG,” Neuroimage, vol. 46, pp. 708–716, Jul. 1, 2009.

[11] J. R. Manning, J. Jacobs, I. Fried, and M. J. Kahana, “Broadband shiftsin local field potential power spectra are correlated with single-neuronspiking in humans,” J. Neurosci., vol. 29, pp. 13613–13620, Oct. 28,2009.

[12] S. Ray, N. E. Crone, E. Niebur, P. J. Franaszczuk, and S. S. Hsiao,“Neural correlates of high-gamma oscillations (60–200Hz) inmacaquelocal field potentials and their potential implications in electrocorticog-raphy,” J. Neurosci., vol. 28, pp. 11526–11536, Nov. 5, 2008.

[13] E. C. Leuthardt, G. Schalk, J. R. Wolpaw, J. G. Ojemann, and D. W.Moran, “A brain-computer interface using electrocorticographic sig-nals in humans,” J. Neural. Eng., vol. 1, pp. 63–71, Jun. 2004.

[14] G. Schalk, K. J. Miller, N. R. Anderson, J. A. Wilson, M. D. Smyth, J.G. Ojemann, D. W. Moran, J. R. Wolpaw, and E. C. Leuthardt, “Two-dimensional movement control using electrocorticographic signals inhumans,” J. Neural. Eng., vol. 5, pp. 75–84, Mar. 2008.

[15] J. R. Wolpaw, D. J. McFarland, G. W. Neat, and C. A. Forneris, “AnEEG-based brain-computer interface for cursor control,” Electroen-cephalogr. Clin. Neurophysiol., vol. 78, pp. 252–259, Mar. 1991.

[16] D. J. McFarland, W. A. Sarnacki, and J. R. Wolpaw, “Electroen-cephalographic (EEG) control of three-dimensional movement,” J.Neural. Eng., vol. 7, p. 036007, Jun. 2010.

[17] A. G. Rouse, J. J. Williams, J. J. Wheeler, and D. W. Moran, “Corticaladaptation to a chronic micro-electrocorticographic brain computer in-terface,” J. Neurosci., vol. 33, pp. 1326–1330, Jan. 23, 2013.

[18] J. S. Duncan, X. Papademetris, J. Yang, M. Jackowski, X. Zeng, and L.H. Staib, “Geometric strategies for neuroanatomic analysis fromMRI,”Neuroimage, vol. 23, pp. S34–S45, 2004.

[19] Y. Klipper-Aurbach, M. Wasserman, N. Braunspiegel-Weintrob, D.Borstein, S. Peleg, S. Assa, M. Karp, Y. Benjamini, Y. Hochberg, andZ. Laron, “Mathematical formulae for the prediction of the residualbeta cell function during the first two years of disease in childrenand adolescents with insulin-dependent diabetes mellitus,” Med.Hypotheses, vol. 45, pp. 486–490, Nov. 1995.

[20] D. J. McFarland, L. M. McCane, S. V. David, and J. R. Wolpaw,“Spatial filter selection for EEG-based communication,” Electroen-cephalogr. Clin. Neurophysiol., vol. 103, pp. 386–394, 1997.

[21] M. S. Johannes, J. D. Bigelow, J. M. Burck, S. D. Harshbarger, M.V. Kozlowski, and T. Van Doren, “An overview of the developmentprocess for the modular prosthetic limb,” Johns Hopkins Univ. Appl.Phys. Lab. Tech. Dig., vol. 30, 2011.

[22] A. Harris, K. Katyal, M. Para, and J. Thomas, “Revolutionizing pros-thetics software technology,” in Proc. 2011 IEEE Int. Conf. Syst., Man,Cybern., 2011, pp. 2877–2884.

[23] M. M. Bridges, M. P. Para, and M. J. Mashner, “Control system archi-tecture for the modular prosthetic limb,” Johns Hopkins Univ. Appl.Phys. Lab. Tech. Dig., vol. 30, 2011.

[24] M. S. Fifer, S. Acharya, H. L. Benz, M. Mollazadeh, N. E. Crone, andN. V. Thakor, “Toward electrocorticographic control of a dexterousupper limb prosthesis: Building brain-machine interfaces,” IEEEPulse, vol. 3, pp. 38–42, Jan. 2012.

[25] G. Rizzolatti, G. Luppino, and M. Matelli, “The organization of thecortical motor system: New concepts,” Electroencephalogr. Clin. Neu-rophysiol., vol. 106, pp. 283–296, Apr. 1998.

[26] F. Filimon, “Human cortical control of handmovements: Parietofrontalnetworks for reaching, grasping, and pointing,”Neuroscientist, vol. 16,pp. 388–407, Aug. 2010.

[27] N. E. Crone, D. L.Miglioretti, B. Gordon, and R. P. Lesser, “Functionalmapping of human sensorimotor cortex with electrocorticographicspectral analysis. II. Event-related synchronization in the gammaband,” in Brain, Dec. 1998, vol. 121, pp. 2301–2315, (pt 12).

[28] N. E. Crone, L. Hao, J. Hart, Jr., D. Boatman, R. P. Lesser, R. Irizarry,and B. Gordon, “Electrocorticographic gamma activity during wordproduction in spoken and sign language,” Neurology, vol. 57, pp.2045–2053, Dec. 11, 2001.

[29] K. J. Miller, E. C. Leuthardt, G. Schalk, R. P. Rao, N. R. Anderson,D. W. Moran, J. W. Miller, and J. G. Ojemann, “Spectral changes incortical surface potentials during motor movement,” J. Neurosci., vol.27, pp. 2424–2432, Feb. 28, 2007.

[30] S. Ray, E. Niebur, S. S. Hsiao, A. Sinai, and N. E. Crone, “High-frequency gamma activity (80–150 Hz) is increased in human cortexduring selective attention,” Clin. Neurophysiol., vol. 119, pp. 116–133,Jan. 2008.

[31] J. Liu and W. T. Newsome, “Local field potential in cortical area MT:Stimulus tuning and behavioral correlations,” J. Neurosci., vol. 26, pp.7779–7790, July 26, 2006.

[32] S. Ray, N. E. Crone, E. Niebur, P. J. Franaszczuk, and S. S. Hsiao,“Neural correlates of high-gamma oscillations (60–200Hz) inmacaquelocal field potentials and their potential implications in electrocorticog-raphy,” J. Neurosci., vol. 28, pp. 11526–11536, 2008.

[33] S. Ray and J. H. Maunsell, “Different origins of gamma rhythm andhigh-gamma activity in macaque visual cortex,” PLoS Biol., vol. 9, p.e1000610, Apr. 2011.

[34] A. Flinker, E. F. Chang, N. M. Barbaro, M. S. Berger, and R. T. Knight,“Sub-centimeter language organization in the human temporal lobe,”Brain Lang., vol. 117, pp. 103–109, Jun. 2011.

[35] J. P. Lachaux, N. Axmacher, F. Mormann, E. Halgren, and N. E. Crone,“High-frequency neural activity and human cognition: Past, presentand possible future of intracranial EEG research,” Prog. Neurobiol.,vol. 98, pp. 279–301, Sept. 2012.

[36] A. Flinker, E. F.Chang,H.E.Kirsch,N.M.Barbaro,N. E.Crone, andR.T.Knight, “Single-trial speech suppression of auditory cortex activity inhumans,” J. Neurosci., vol. 30, pp. 16643–16650, Dec. 8, 2010.

[37] J. Kubanek, K. J. Miller, J. G. Ojemann, J. R. Wolpaw, and G. Schalk,“Decoding flexion of individual fingers using electrocorticographicsignals in humans,” J. Neural. Eng., vol. 6, p. 66001, Dec. 2009.

[38] T. Yanagisawa, M. Hirata, Y. Saitoh, T. Goto, H. Kishima, R. Fukuma,H. Yokoi, Y. Kamitani, and T. Yoshimine, “Real-time control of a pros-thetic hand using human electrocorticography signals,” J. Neurosurg.,vol. 114, pp. 1715–1722, June 2011.

[39] Z. C. Chao, Y. Nagasaka, and N. Fujii, “Long-term asynchronous de-coding of arm motion using electrocorticographic signals in monkeys,”Front Neuroeng., vol. 3, p. 3, 2010.

[40] L. R. Hochberg, D. Bacher, B. Jarosiewicz, N. Y. Masse, J. D. Simeral,J. Vogel, S. Haddadin, J. Liu, S. S. Cash, P. van der Smagt, and J. P.Donoghue, “Reach and grasp by peoplewith tetraplegia using a neurallycontrolled robotic arm,”Nature, vol. 485, pp. 372–375,May 17, 2012.

[41] J. L. Collinger, B. Wodlinger, J. E. Downey, W. Wang, E. C. Tyler-Kabara, D. J. Weber, A. J. McMorland, M. Velliste, M. L. Boninger,and A. B. Schwartz, “High-performance neuroprosthetic control by anindividual with tetraplegia,” Lancet, Dec. 13, 2012.

[42] N. S. Narayanan, E. Y. Kimchi, and M. Laubach, “Redundancy andsynergy of neuronal ensembles in motor cortex,” J. Neurosci., vol. 25,pp. 4207–4216, Apr. 27, 2005.

[43] M. W. Slutzky, L. R. Jordan, T. Krieg, M. Chen, D. J. Mogul, and L. E.Miller, “Optimal spacing of surface electrode arrays for brain-machineinterface applications,” J. Neural. Eng., vol. 7, p. 26004, Apr. 2010.

[44] J. Viventi, D. H. Kim, L. Vigeland, E. S. Frechette, J. A. Blanco, Y. S.Kim, A. E. Avrin, V. R. Tiruvadi, S. W. Hwang, A. C. Vanleer, D. F.Wulsin, K. Davis, C. E. Gelber, L. Palmer, J. Van der Spiegel, J. Wu,J. Xiao, Y. Huang, D. Contreras, J. A. Rogers, and B. Litt, “Flexible,foldable, actively multiplexed, high-density electrode array for map-ping brain activity in vivo,” Nat. Neurosci., vol. 14, pp. 1599–1605,Dec. 2011.

[45] S. Thongpang, T. J. Richner, S. K. Brodnick, A. Schendel, J. Kim, J.A. Wilson, J. Hippensteel, L. Krugner-Higby, D. Moran, A. S. Ahmed,D. Neimann, K. Sillay, and J. C. Williams, “A micro-electrocorticog-raphy platform and deployment strategies for chronic BCI applica-tions,” Clin. EEG Neurosci., vol. 42, pp. 259–265, Oct. 2011.


Matthew S. Fifer received the B.S. degree inbiomedical engineering from the University ofVirginia, Charlottesville, VA, USA, in 2009. He iscurrently working toward the Ph.D. degree at JohnsHopkins University, Baltimore, MD, USA, under thementorship of Dr. Nathan Crone.He has been primarily focused on neural decoding

of movement parameters from intracranial EEG sig-nals and translating that knowledge into online con-trol of the JHU/APL Modular Prosthetic Limb.

Guy Hotson received the B.S. degree in computerengineering from Santa Clara University, SantaClara, CA, USA, in 2011. He is currently workingtoward the Ph.D. degree in electrical and computerengineering at Johns Hopkins University, Baltimore,MD, USA, in Dr. Nitish Thakor’s Neuroengineeringand Biomedical Instrumentation Lab.His focus is on algorithm research and develop-

ment for an online motor-BMI using intracranialEEG signals.

Brock A. Wester received the B.S. degree in com-puter engineering from the Georgia Institute of Tech-nology, Atlanta, GA, USA, in 2004, and a joint Ph.D.degree in biomedical engineering from the GeorgiaInstitute of Technology and Emory University, At-lanta, GA, USA, in 2010.He serves as both a Project Manager and Section

Supervisor for the Applied Neuroscience sectionof the Research and Exploratory Development De-partment of the Johns Hopkins University AppliedPhysics Laboratory, Laurel, MD, USA. His research

areas include prosthetics, brain–computer interfaces, virtual environments,MEMS microfabrication and packaging, and traumatic brain and blast injury.

David P. McMullen received the B.S. degree inpsychology with a neuroscience concentration fromDuke University, Durham NC, USA, in 2008. Heis currently working toward the M.D. degree atRutgers Robert Wood Johnson Medical School, NewBrunswick, NJ, USA, during which he conducted aHoward Hughes Medical Institute Research Fellow-ship at Johns Hopkins University, Baltimore, MD,USA.His research focuses on brain–machine interfaces

and neuromodulation. In particular, his current re-search explores novel control paradigms for the JHU/APL Modular ProstheticLimb using electrocorticography-based BMIs.

Yujing Wang received the B.S. degree in biomed-ical engineering from Tsinghua University, Beijing,China, in 2011. She worked with Dr. Nathan Crone inhis Cognitive Neurophysiology Lab and received theM.S. degree in biomedical engineering from JohnsHopkins University, Baltimore, MD, USA, in 2013.She is currently working toward the Ph.D. degree inbioengineering at University of Maryland, CollegePark, MD, USA, and focusing on functionality of themammalian auditory system in Dr. Shihab Shamma’sNeural System Lab.

Matthew S. Johannes received the B.S. and Ph.D.degrees in mechanical engineering and materials sci-ence from Duke University, Durham, NC, USA, in2001 and 2007, respectively.He serves as a Project Manager in the Research

and Exploratory Development Department of theJohns Hopkins University Applied Physics Labora-tory, Laurel, MD, USA, managing the developmentand clinical integration of the Modular ProstheticLimb. His research areas include prosthetics,brain–computer interfaces, open architectures for

ground robotic systems, bimanual robotic manipulation, MEMS, and systemengineering and integration.

Kapil D. Katyal received the B.S. degree in com-puter science from Pennsylvania State University,State College, PA , USA, and the M.S. degree inelectrical engineering from Columbia University,New York, NY, USA.He currently works as a Senior Professional Staff

member at the Johns Hopkins University AppliedPhysics Lab, Laurel, MD, USA, where his researchinterests include embedded systems, robotic andprosthetic control, autonomous grasping, sensorprocessing, and 3-D computer vision/perception.

John B. Helder received the B.S. degree in electricalengineering and the M.S. degree in electrical andcomputer engineering from the Georgia Institute ofTechnology, Atlanta, GA, USA, in 2006 and 2007,respectively.He serves as both a Research Engineer and

Assistant Section Supervisor for the Signal LocatingSystems Section in the Asymmetric Operation De-partment of the Johns Hopkins University AppliedPhysics Laboratory, Laurel, MD, USA. His researchareas include software defined radios, low power

electronics, and miniaturization.

Matthew P. Para received the B.S. degrees in me-chanical engineering and in computer science, andthe M.S. degree in mechanical engineering from theUniversity of Maryland, College Park, MD, USA, in2004 and 2010, respectively.He works as the Controls Lead for the Revolu-

tionizing Prosthetics program in the Robotics andAutonomous Systems group of the Research andExploratory Development Department of the JohnsHopkins University Applied Physics Laboratory,Laurel, MD, USA. His research interests include

prosthetics, brain–computer interfaces, virtual environments, robotic systems,and autonomous control.

R. Jacob Vogelstein received the Sc.B. degree inneural engineering from Brown University, Provi-dence, RI, USA, and the Ph.D. degree in biomedicalengineering from the Johns Hopkins UniversitySchool of Medicine, Baltimore, MD, USA.He serves as a Program Manager in the Research

and Exploratory Development Department of theJohns Hopkins University Applied Physics Lab-oratory, Laurel, MD, USA, where he oversees theApplied Neuroscience portfolio. In addition, heholds a joint appointment as an Assistant Research

Professor in theDepartment of Electrical and Computer Engineering at the JohnsHopkins University’s Whiting School of Engineering. His research focuses areneuromimetic computing, brain–computer interfaces, and neuroprosthetics.


William S. Anderson received the B.S. degreein physics from Texas A&M University, CollegeStation, TX, USA, the M.A. and Ph.D. degrees inphysics from Princeton University, Princeton, NJ,USA, and the M.D. degree from the Johns HopkinsUniversity School of Medicine, Baltimore, MD,USA.He is currently an Assistant Professor of Neu-

rosurgery the Johns Hopkins University School ofMedicine, Baltimore, MD, USA. His research inter-ests include computational modeling of epileptiform

neural networks, cortical and subcortical recording techniques, and closed-looptherapeutic neurostimulation systems.

Nitish V. Thakor is a Professor of BiomedicalEngineering. Electrical and Computer Engineering,and Neurology at Johns Hopkins University anddirects the Laboratory for Neuroengineering. He isalso the Director the Singapore Institute for Neu-rotechnology (SINAPSE) at the National Universityof Singapore. His technical expertise is in the fieldof Neuroengineering, including neural diagnosticinstrumentation, neural microsystems, neural signalprocessing, optical imaging of the nervous system,neural control of prosthesis, and brain–machine

interface. He is currently the Editor-in-Chief of Medical and BiologicalEngineering and Computing.

Dr. Thakor was the Editor-in-Chief of IEEE TRANSACTIONS ON NEURALSYSTEMS AND REHABILITATION ENGINEERING from 2005 to 2011. He is a recip-ient of a Research Career Development Award from the National Institutes ofHealth and a Presidential Young Investigator Award from the National ScienceFoundation, and is a Fellow of the American Institute of Medical and Biolog-ical Engineering, Founding Fellow of the Biomedical Engineering Society, andFellow of International Federation of Medical and Biological Engineering. Heis a recipient of the award of Technical Excellence in Neuroengineering fromIEEE Engineering in Medicine and Biology Society, Distinguished AlumnusAward from Indian Institute of Technology, Bombay, India, and a CentennialMedal from the University of Wisconsin School of Engineering.

Nathan E. Crone is an Associate Professor of Neu-rology in The Johns Hopkins University School ofMedicine, Baltimore, MD, USA, and directs the Cog-nitive Neurophysiology Laboratory. He leads a long-standing research program using intracranial EEG tostudy human brain mechanisms of language, atten-tion, and motor function. This research is carried outin patients undergoing intracranial EEG monitoringprior to epilepsy surgery. He serves as an AttendingNeurologist in the Johns Hopkins Hospital where hisclinical work focuses on the management of drug-

resistant epilepsy, electroencephalography, and intraoperative monitoring. Histechnical expertise is in the fields of cognitive neuroscience and neurophys-iology, including the rapid characterization and application of human neuro-physiological responses for both functional brain mapping and brain–machineinterfaces.

Documents

Simultaneous Neural Control of Simple Reaching and Grasping With the Modular Prosthetic Limb Using Intracranial EEG