Browsing by Subject "brain-machine interface"

Brain-machine interfaces (BMIs) use neuronal activity to control external actuators. As such, they show great promise for restoring motor and communication abilities in persons with paralysis or debilitating neurological disorders.

While BMIs aim to enact normal sensorimotor functions, so far they have lacked afferent feedback in the form of somatic sensation. This deficiency limits the utility of current BMI designs and may hinder the translation of future clinical BMIs, which will need a means of delivering sensory signals from prosthetic devices back to the user.

This dissertation describes the development of brain-machine-brain interfaces (BMBIs) capable of bidirectional communication with the brain. The interfaces consisted of efferent and afferent modules. The efferent modules decoded motor intentions from the activity of populations of cortical neurons recorded with chronic multielectrode recording arrays. The activity of these ensembles was used to drive the movements of a computer cursor and a realistic upper-limb avatar. The afferent modules encoded tactile feedback about the interactions of the avatar with virtual objects through patterns of intracortical microstimulation (ICMS).

I first show that a direct intracortical signal can be used to instruct rhesus monkeys about the direction of a reach to make with a BMI. Rhesus monkeys placed an actuator over an instruction target and obtained, from the target's artificial texture, information about the correct reach path. Initially these somatosensory instructions took the form of vibrotactile stimulation of the hands. Next, ICMS of primary somatosensory cortex (S1) in one monkey and posterior parietal cortex (PPC) in another was substituted for this peripheral somatosensory signal. Finally, the monkeys made direct brain-controlled reaches using the activity of ensembles of primary motor cortex (M1) cells, conditional on the ICMS cues. The monkey receiving ICMS of S1 was able to achieve the same level of proficiency with ICMS as with the stimulus delivered to the skin of the hand. The monkey receiving ICMS of PPC was unable to perform the task above chance. This experiment indicates that ICMS of S1 can form the basis of an afferent prosthetic input to the brain for guiding brain-controlled prostheses.

I next show that ICMS of S1 can provide feedback about the interactions of a virtual-reality upper-limb avatar and virtual objects, enabling active touch. Rhesus monkeys initially controlled the avatar with the movements of their arms and used it to search through sets of up to three objects. Feedback in the form of temporal patterns of ICMS occurred whenever the avatar touched a virtual object. Monkeys learned to use this feedback to find the objects with particular artificial textures, as encoded by the ICMS patterns, and select those associated with reward while avoiding selecting the non-rewarded objects. Next, the control of the avatar was switched to direct brain-control and the monkeys continued to move the avatar with motor commands derived from the extracellular neuronal activity of M1 cells. The afferent and efferent modules of this BMBI were temporally interleaved, and as such did not interfere with each other, yet allowed effectively concurrent operation. Cortical motor neurons were measured while the monkey passively observed the movements of the avatar and were found to be modulated, a result that suggests that concurrent visual and artificial somatosensory feedback lead to the incorporation of the avatar into the monkey's internal brain representation.

Finally, I probed the sensitivity of S1 to precise temporal patterns of ICMS. Monkeys were trained to discriminate between periodic and aperiodic ICMS pulse trains. The periodic pulse-trains consisted of 200 Hz bursts at a 10 Hz secondary frequency. The aperiodic pulse trains had a distorted periodicity and consisted of 200 Hz bursts at a variable instantaneous secondary frequency. The statistics of the aperiodic pulse trains were drawn from a gamma distribution with equal mean inter-burst intervals to the periodic pulse trains. The monkeys were able to distinguish periodic pulse trains from aperiodic pulse trains with coefficients of variation of 0.25 or greater. This places an upper-bounds on the communication bandwidth that can be achieved with a single channel of temporal ICMS in S1.

In summary, rhesus monkeys were augmented with a bidirectional neural interface that allowed them to make reaches to objects and discriminate them by their textures--all without making actual movements and without relying on somatic sensation from their real bodies. Both action and perception were mediated by the brain-machine-brain interface. I probed the sensitivity of the afferent leg of the interface to precise temporal patterns of ICMS. Moreover, I describe evidence that the BMBI controlled avatar was incorporated into the monkey's internal brain representation. These results suggest that future clinical neuroprostheses could implement realistic feedback about object-actuator interactions through patterns of ICMS, and that these artificial somatic sensations could lead to the incorporation of the prostheses into the user's body schema.

Brain-machine interfaces (BMI) are systems which connect brains directly to machines or computers for communication. BMI-controlled prosthetic devices use algorithms to decode neuronal recordings into movement commands. These algorithms operate using models of how recorded neuronal signals relate to desired movements, called models of tuning. Models of tuning have typically been linear in prior work, due to the simplicity and speed of the algorithms used with them. Neuronal tuning has been shown to slowly change over time, but most prior work do not adapt tuning models to these changes. Furthermore, extracellular electrical recordings of neurons' action potentials slowly change over time, impairing the preprocessing step of spike-sorting, during which the neurons responsible for recorded action potentials are identified.

This dissertation presents a non-linear adaptive Bayesian filter and an adaptive spike-sorting method for BMI decoding. The adaptive filter consists of the n-th order unscented Kalman filter and Bayesian regression self-training updates. The unscented Kalman filter estimates desired prosthetic movements using a non-linear model of tuning as its observation model. The model is quadratic with terms for position, velocity, distance from center of workspace, and velocity magnitude. The tuning model relates neuronal activity to movements at multiple time offsets simultaneously, and the movement model of the filter is an order n autoregressive model.

To adapt the tuning model parameters to changes in the brain, Bayesian regression self-training updates are performed periodically. Tuning model parameters are stored as probability distributions instead of point estimates. Bayesian regression uses the previous model parameters as priors and calculates the posteriors of the regression between filter outputs, which are assumed to be the desired movements, and neuronal recordings. Before each update, filter outputs are smoothed using a Kalman smoother, and tuning model parameters are passed through a transition model describing how parameters change over time. Two variants of Bayesian regression are presented: one uses a joint distribution for the model parameters which allows analytical inference, and the other uses a more flexible factorized distribution that requires approximate inference using variational Bayes.

To adapt spike-sorting parameters to changes in spike waveforms, variational Bayesian Gaussian mixture clustering updates are used to update the waveform clustering used to calculate these parameters. This Bayesian extension of expectation-maximization clustering uses the previous clustering parameters as priors and computes the new parameters as posteriors. The use of priors allows tracking of clustering parameters over time and facilitates fast convergence.

To evaluate the proposed methods, experiments were performed with 3 Rhesus monkeys implanted with micro-wire electrode arrays in arm-related areas of the cortex. Off-line reconstructions and on-line, closed-loop experiments with brain-control show that the n-th order unscented Kalman filter is more accurate than previous linear methods. Closed-loop experiments over 29 days show that Bayesian regression self-training helps maintain control accuracy. Experiments on synthetic data show that Bayesian regression self-training can be applied to other tracking problems with changing observation models. Bayesian clustering updates on synthetic and neuronal data demonstrate tracking of cluster and waveform changes. These results indicate the proposed methods improve the accuracy and robustness of BMIs for prosthetic devices, bringing BMI-controlled prosthetics closer to clinical use.

Brain-machine interfaces (BMIs) are devices that transform neural activity into commands executed by a robotic actuator. For paraplegics who have suffered spinal cord injury and for amputees, BMIs provide an avenue to regain lost limb mobility by providing a direct connection between the brain and an actuator. One of the most important aspects of a BMI is the decoding algorithm, which interprets patterns of neural activity and issues an appropriate kinematic action. The decoding algorithm relies heavily on a neural tuning function for each neuron which describes the response of that neuron to an external stimulus or upcoming motor action. Modern BMI decoders assume a simple parametric form for this tuning function such as cosine, linear, or quadratic, and fit parameters of the chosen function to a training data set. While this may be appropriate for some neurons, tuning curves for all neurons may not all take the same parametric form; hence, performance of BMI decoding may suffer because of an inappropriate mapping from firing rate to kinematic. In this work, we develop a non-parametric model for the identification of non-linear tuning curves with arbitrary shape. We also develop an associated variational Bayesian (VB) inference scheme which provides a fast, big data-friendly method to obtain approximate posterior distributions on model parameters. We demonstrate our model's capabilities on both simulated and experimental datasets.