A high-performance brain–computer interface

Abstract

Video LegendsThese videos show the behavioral conditions from our different experiments. The experiment room itself was visibly dark, but the video camera could image the scene with infrared light. Bright text and numbers were added as annotations and were not seen by the monkey. Each successful trial was followed by a juice reward, indicated by a short tone followed by a click corresponding to the action of a juice dispenser. These auditory signals were heard by the monkey during the experiment. For the duration of the experiment, a piece of reflective tape was lightly wrapped around one phalanx of the monkey's left hand to track the monkey's arm movements. This tape is especially bright in the infrared videos but was not visible to the monkey during the experiments. Video 3 SuppVideo3BCIFast.mpg Movie showing fast-paced prosthetic cursor trials during BCI experiments. We showed three separate sequences, each of five cursor trials followed by a real reach. These three sequences were not performed in succession. Rather, the three sequences of five cursor trials were spliced from different times during the same experiment to better demonstrate the speed of our system. The prosthetic cursor is not visible in this video because it was flashed briefly so as to not slow the overall presentation of trials.

Free-paced high-performance brain–computer interfaces

“…We made two changes: (1) we ignored Baseline, since it provided no information about the target and (2) for Plan and Go, we used data within [150,350] ms after the start of each period instead of [50,250] ms as before. This further delay in the time window reflects the additional time required for plan activity to fully form and stabilize in PMd, as discussed above in association with our recent fixed-pace results (termed T skip in Santhanam et al (2006); and normalized variance reduction time in Churchland et al (2006c)). As with the state detection model, we measured the mean and variance of the neuron spike counts for each target during both Plan and Go periods.…”

“…With these limits identified, we can now design complementary state estimators which use trial timing parameters identified in Santhanam et al (2006) yet maintain overall prosthetic performance.…”

“…Following a variable delay period (200-1000 ms), the central touch and fixation targets disappear; this is the 'go' cue indicating that the reach may begin. The movement ends with the monkey acquiring the target and holding for 200 ms. We recorded neural activity from a 96-channel electrode (Cyberkinetics Neurotechnology Inc.) implanted in the arm representation of PMd (see supplementary materials by Santhanam et al (2006) for the exact placement) and automatically spike-sorted data using methods described previously by Santhanam et al (2004Santhanam et al ( , 2006.…”

Prediction of upper limb muscle activity from motor cortical discharge during reaching

“…The brain-machine interface field has been dominated by applications in which neural signals recorded from motor or premotor areas are used to predict kinematic signals: the position of a cursor on a computer screen (Serruya et al 2002, Taylor et al 2002, Santhanam et al 2006, Kennedy and Bakay 1998, Wolpaw and McFarland 2004 or the endpoint of a robotic limb (Serruya et al 2002, Chapin et al 1999, Taylor et al 2003, Carmena et al 2003. We have demonstrated that similar techniques can be used to predict the activity of individual muscles of the arm and hand.…”

Brain-Computer Interfaces in Medicine

“…12,16 Ongoing studies in a number of laboratories are working toward achieving natural control of devices such as a prosthetic arm using electrode microarrays implanted in the motor cortex or other cortical areas of nonhuman primates. [109][110][111][112][113][114] Plans are under way in several centers to translate these studies into human trials.…”