Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 6, 22170

Wireless Cortical Brain-Machine Interface for Whole-Body Navigation in Primates

Affiliations

Wireless Cortical Brain-Machine Interface for Whole-Body Navigation in Primates

Sankaranarayani Rajangam et al. Sci Rep.

Abstract

Several groups have developed brain-machine-interfaces (BMIs) that allow primates to use cortical activity to control artificial limbs. Yet, it remains unknown whether cortical ensembles could represent the kinematics of whole-body navigation and be used to operate a BMI that moves a wheelchair continuously in space. Here we show that rhesus monkeys can learn to navigate a robotic wheelchair, using their cortical activity as the main control signal. Two monkeys were chronically implanted with multichannel microelectrode arrays that allowed wireless recordings from ensembles of premotor and sensorimotor cortical neurons. Initially, while monkeys remained seated in the robotic wheelchair, passive navigation was employed to train a linear decoder to extract 2D wheelchair kinematics from cortical activity. Next, monkeys employed the wireless BMI to translate their cortical activity into the robotic wheelchair's translational and rotational velocities. Over time, monkeys improved their ability to navigate the wheelchair toward the location of a grape reward. The navigation was enacted by populations of cortical neurons tuned to whole-body displacement. During practice with the apparatus, we also noticed the presence of a cortical representation of the distance to reward location. These results demonstrate that intracranial BMIs could restore whole-body mobility to severely paralyzed patients in the future.

Figures

Figure 1
Figure 1. Overview of the experimental design.
(A) The mobile robotic wheelchair, which seats a monkey, was moved from one of the three starting locations (dashed circles) to a grape dispenser. The wireless recording system records the spiking activities from the monkey’s head stage, and sends the activities to the wireless receiver to decode the wheelchair movement. (B) Schematic of the brain regions from which we recorded units tuned to either velocity or steering. Red dots correspond to units in M1, blue from PMd and green from the somatosensory cortex. (C) Three video frames show Monkey K drive toward the grape dispenser. The right panel shows the average driving trajectories (dark blue) from the three different starting locations (green circle) to the grape dispenser (red circle). The light blue ellipses are the standard deviation of the trajectories.
Figure 2
Figure 2. Translational and rotational velocity tuning for neurons during passive training and BMI-based navigation.
(A) Each diagram shows the normalized firing rate as a function of translational and rotational velocity commands (bottom-left), where 1 represents the maximum command value sent to the wheelchair (positive values represent forward commands for translation, and clock-wise (CW) for rotation). Each sequence shows the normalized firing rate at different time lag. Neuron A is tuned to backward movement during both passive training and BMI control. Neuron B is tuned to right turn movement during brain control, but with different tuning during passive training. (B) The tuning depth across time for the two neurons in (A). Note that the time lag of the peak tuning differs between passive training and BMI control. (C) Distribution of the tuning depth to velocity and the lag of peak velocity tuning for all the neural units recorded. The lag of peak velocity tuning moved from ~0 ms to ~ −500 ms, i.e. half a second earlier than the movement onset. (D) Neural units that were well tuned during passive navigation, were more likely to be well tuned during BMI navigation. Each dot represents one neural unit, and the solid line is the regression line. (E) For units that were better tuned, their tuning diagrams between passive and MI navigation were also better correlated in Monkey K, but not in Monkey M.
Figure 3
Figure 3. Behavioral improvement and the increase of decoder similarities across sessions.
(A) Both monkeys show significant improvement in the traveling time and distance as they learn. The circles represent the median and the error bars show the interquartile range of the medians. (B) Increased correlations between decoders trained in earlier sessions and the last session. (left, Monkey K; right, Monkey M (C) Both monkeys demonstrate impaired performance once their decoded movement commands were inverted (forward now becomes backward, and right turn becomes left turn). The bar graph shows the median and the error bars indicate the interquartile range of the medians.
Figure 4
Figure 4. Population responses as a function of distance to the reward.
(A) Population responses in Monkey K and M as a function of distance to the reward during passive navigation and BMI navigation (B) Color represents normalized firing rate, and the bar graph shows the average unit modulation, calculated by the average of the absolute values of all units’ normalized firing rate at each distance. When the cart is getting close to the target, the monkeys may reach to the grape, and these reach-related activities are shaded in red.

Similar articles

See all similar articles

Cited by 12 PubMed Central articles

See all "Cited by" articles

References

    1. Hoenig H., Giacobbi P. & Levy C. E. Methodological challenges confronting researchers of wheeled mobility aids and other assistive technologies. Disabil Rehabil Assist Technol 2, 159–168 (2007). - PubMed
    1. Aziz F., Arof H., Mokhtar N. & Mubin M. HMM based automated wheelchair navigation using EOG traces in EEG. J Neural Eng 11, 056018, 10.1088/1741-2560/11/5/056018 (2014). - DOI - PubMed
    1. Kaufmann T., Herweg A. & Kubler A. Toward brain-computer interface based wheelchair control utilizing tactually-evoked event-related potentials. J. Neuroeng. Rehabil. 11, 7, 10.1186/1743-0003-11-7 (2014). - DOI - PMC - PubMed
    1. Nicolas-Alonso L. F. & Gomez-Gil J. Brain computer interfaces, a review. Sensors (Basel) 12, 1211–1279, 10.3390/s120201211 (2012). - DOI - PMC - PubMed
    1. Vanacker G. et al. Context-based filtering for assisted brain-actuated wheelchair driving. Comput. Intell. Neurosci. 2007, 25130, 10.1155/2007/25130 (2007). - DOI - PMC - PubMed

Publication types

Feedback