The authors note, "For the 'hybrid' location discrimination task, we report data obtained from 27 electrodes, 16 of which were in area 1; the 11 electrodes in area 3b were divided evenly across the two animals (6 and 5). We had previously tested all of the electrodes, including those in area 3b, in the detection and discrimination tasks (as shown in Fig. 3) and found them all to yield approximately equivalent performance (see Fig 3A). We noticed in the hybrid location discrimination task, however, that one of the animals performed much more poorly based on stimulation of area 3b than it did based on stimulation of area 1 (while the other animal performed better based on stimulation of area 1). Having no reason to question any of the arrays, we attributed this discrepancy to differences across animals and arrived at the conclusion, based on pooled data from both animals, that stimulation of the two areas yields equivalent performance in the 'hybrid location discrimination' task. The overall conclusion, then, was that stimulation of neurons in area 3b and 1 evokes percepts that are equally localized on the skin."Shortly after publication of the paper, we repeated detection experiments across the arrays and found that the animal could no longer detect stimulation through the array in area 3b that had yielded poor performance in the hybrid location discrimination task. It is therefore likely that this array had failed between the time we conducted the initial detection and discrimination experiments and the time we conducted the hybrid location discrimination task (which required 2-3 months of retraining). If this is the case, and we eliminate data from that bad array, then the median performance on hybrid trials is 83% (up from the 80% that was originally reported), which is still statistically poorer than that on the location-matched mechanical trials [median difference between performance on mechanical and hybrid trials was 3.3% rather than 5.6%, t (119) = 6.1, P < 0.001] (see the corrected Fig. 2). Thus, we probably underestimated overall performance on hybrid trials, and thus the degree to which artificial percepts are localized, in the original publication. Importantly, however, performance on hybrid trials based on stimulation of area 3b was significantly better than performance based on stimulation of area 1 [median Δp = 0.028 and 0.054 for areas 3b and 1, respectively; t test: t (76) = 2.8, P < 0.01]. Thus, based on the data obtained from only one animal, it seems as though stimulation of area 3b elicits more localized percepts than does stimulation of area 1, as might be expected given that neurons in area 3b tend to have smaller receptive fields than their counterparts in area 1 (1, 2)."As a result of this error, Fig. 2 and its legend appeared incorrectly. The corrected figure and its corresponding legend appear below. On both mechanical and hybrid trials, the relative locations of stimuli applied to widely spaced digits were more accurately discriminated than were the relative locations of stimuli applie...

Quadratic assignment problems arise in a wide variety of domains, spanning operations research, graph theory, computer vision, and neuroscience, to name a few. The graph matching problem is a special case of the quadratic assignment problem, and graph matching is increasingly important as graph-valued data is becoming more prominent. With the aim of efficiently and accurately matching the large graphs common in big data, we present our graph matching algorithm, the Fast Approximate Quadratic assignment algorithm. We empirically demonstrate that our algorithm is faster and achieves a lower objective value on over 80% of the QAPLIB benchmark library, compared with the previous state-of-the-art. Applying our algorithm to our motivating example, matching C. elegans connectomes (brain-graphs), we find that it efficiently achieves performance.

No abstract

To increase the ability of brain-machine interfaces (BMIs) to control advanced prostheses such as the modular prosthetic limb (MPL), we are developing a novel system: the Hybrid Augmented Reality Multimodal Operation Neural Integration Environment (HARMONIE). This system utilizes hybrid input, supervisory control, and intelligent robotics to allow users to identify an object (via eye tracking and computer vision) and initiate (via brain-control) a semi-autonomous reach-grasp-and-drop of the object by the MPL. Sequential iterations of HARMONIE were tested in two pilot subjects implanted with electrocorticographic (ECoG) and depth electrodes within motor areas. The subjects performed the complex task in 71.4% (20/28) and 67.7% (21/31) of trials after minimal training. Balanced accuracy for detecting movements was 91.1% and 92.9%, significantly greater than chance accuracies (p < 0.05). After BMI-based initiation, the MPL completed the entire task 100% (one object) and 70% (three objects) of the time. The MPL took approximately 12.2 seconds for task completion after system improvements implemented for the second subject. Our hybrid-BMI design prevented all but one baseline false positive from initiating the system. The novel approach demonstrated in this proof-of-principle study, using hybrid input, supervisory control, and intelligent robotics, addresses limitations of current BMIs.

Tactile sensation is critical for effective object manipulation, but current prosthetic upper limbs make no provision for delivering somesthetic feedback to the user. For individuals who require use of prosthetic limbs, this lack of feedback transforms a mundane task into one that requires extreme concentration and effort. Although vibrotactile motors and sensory substitution devices can be used to convey gross sensations, a direct neural interface is required to provide detailed and intuitive sensory feedback. In light of this, we describe the implementation of a somatosensory prosthesis with which we elicit, through intracortical microstimulation (ICMS), percepts whose magnitude is graded according to the force exerted on the prosthetic finger. Specifically, the prosthesis consists of a sensorized finger, the force output of which is converted into a regime of ICMS delivered to primary somatosensory cortex through chronically implanted multi-electrode arrays. We show that the performance of animals (Rhesus macaques) on a tactile task is equivalent whether stimuli are delivered to the native finger or to the prosthetic finger.

No abstract

We present a neuromorphic silicon chip that emulates the activity of the biological spinal central pattern generator (CPG) and creates locomotor patterns to support walking. The chip implements ten integrate-and-fire silicon neurons and 190 programmable digital-to-analog converters that act as synapses. This architecture allows for each neuron to make synaptic connections to any of the other neurons as well as to any of eight external input signals and one tonic bias input. The chip's functionality is confirmed by a series of experiments in which it controls the motor output of a paralyzed animal in real-time and enables it to walk along a three-meter platform. The walking is controlled under closed-loop conditions with the aide of sensory feedback that is recorded from the animal's legs and fed into the silicon CPG. Although we and others have previously described biomimetic silicon locomotor control systems for robots, this is the first demonstration of a neuromorphic device that can replace some functions of the central nervous system in vivo.

Intracranial electroencephalographic (iEEG) signals from two human subjects were used to achieve simultaneous neural control of reaching and grasping movements with the Johns Hopkins University Applied Physics Lab (JHU/APL) Modular Prosthetic Limb (MPL), a dexterous robotic prosthetic arm. We performed functional mapping of high gamma activity while the subject made reaching and grasping movements to identify task-selective electrodes. Independent, online control of reaching and grasping was then achieved using high gamma activity from a small subset of electrodes with a model trained on short blocks of reaching and grasping with no further adaptation. Classification accuracy did not decline (p<0.05, one-way ANOVA) over three blocks of testing in either subject. Mean classification accuracy during independently executed overt reach and grasp movements for (Subject 1, Subject 2) were (0.85, 0.81) and (0.80, 0.96) respectively, and during simultaneous execution they were (0.83, 0.88) and (0.58, 0.88) respectively. Our models leveraged knowledge of the subject's individual functional neuroanatomy for reaching and grasping movements, allowing rapid acquisition of control in a time-sensitive clinical setting. We demonstrate the potential feasibility of verifying functionally meaningful iEEG-based control of the MPL prior to chronic implantation, during which additional capabilities of the MPL might be exploited with further training.