Browsing by Author "Nicolelis, Miguel A L"

The loss of the ability to walk as the result of neurological injury or disease critically impacts the mobility and everyday lifestyle of millions. The World Heath Organization (WHO) estimates that approximately 1% of the world's population needs the use of a wheelchair to assist their personal mobility. Advances in the field of brain-machine interfaces (BMIs) have recently demonstrated the feasibility of using neuroprosthetics to extract motor information from cortical ensembles for more effective control of upper-limb replacements. However, the promise of BMIs has not yet been brought to bear on the challenge of restoring the ability to walk. A future neuroprosthesis designed to restore walking would need two streams of information flowing between the user's brain and the device. First, the motor control signals would have to be extracted from the brain, allowing the robotic prosthesis to behave in the manner intended by the user. Second, and equally important would be the flow of sensory and proprioceptive information back to the user from the neuroprosthesis. Here, I contribute to the foundation of such a bi-directional brain machine interface for the restoration of walking in a series of experiments in two animal models, designed to show the feasibility of (1) extracting locomotor information from neuronal ensemble activity and (2) sending information back into the brain via cortical microstimulation.

In a set of experiments designed to investigate the extraction of locomotor parameters, I chronically recorded from ensembles of neurons in primary motor (M1) and primary somatosensory (S1) cortices in two adult female rhesus macaques as they walked bipedally, at various speeds, both forward and backward on a custom treadmill. For these experiments, rhesus monkeys were suitable because of their ability to walk bipedally in a naturalistic manner with training. I demonstrate that the kinematics of bipedal walking in rhesus macaques can be extracted from neuronal ensemble recordings, both offline and in real-time. The activity of hundreds of neurons was processed by a series of linear decoders to extract accurate predictions of leg joints in three dimensional space, as well as leg muscle electromyograms (EMGs). Using a multi-layered switching model allowed us to achieve increased extraction accuracy by segregating different behavioral modes of walking.

In a second set of experiments designed to investigate the usage of microstimulation as a potential artificial sensory channel, I instructed two adult female Aotus trivirgatus (owl monkeys) about the location of a hidden food reward using a series of cortical microstimulation patterns delivered to primary somatosensory (S1) cortex. The owl monkeys discriminated these microstimulation patterns and used them to guide reaching movements to one of two targets. Here, owl monkeys were used which were previously implanted with electrode arrays of high longevity and stability. These monkeys were previously trained on a somatosensory cued task, which allowed a quick transition to microstimulation cueing. The owl monkeys learned to interpret microstimulation patterns, and their skill and speed of learning new patterns improved over several months. Additionally, neuronal activity recorded on non-stimulated electrodes in motor (M1), premotor (PMD) and posterior parietal (PP) cortices allowed us to examine the immediate neural responses to single biphasic stimulation pulses as well as overall responses to the spatiotemporal pattern. Using this recorded neuronal activity, I showed the efficacy of several linear classification algorithms during microstimulation.

These results demonstrate that locomotor kinematic parameters can be accurately decoded from the activity of neuronal ensembles, that multichannel microstimulation is a viable information channel for sensorized prosthetics, and that the technical limitations of combining these techniques can be overcome. I propose that bi-directional BMIs integrating these techniques will one day restore the ability to walk to severely paralyzed patients.

Brain-machine interfaces (BMIs) offer the potential to assist millions of people worldwide suffering from immobility due to loss of limbs, paralysis, and neurodegenerative diseases. BMIs function by decoding neural activity from intact cortical brain regions in order to control external devices in real-time. While there has been exciting progress in the field over the past 15 years, the vast majority of the work has focused on restoring of motor function of a single limb. In the work presented in this thesis, I first investigate the expanded role of primary sensory (S1) and motor (M1) cortex during reaching movements. By varying target size during reaching movements, I discovered the cortical correlates of the speed-accuracy tradeoff known as Fitts' law. Similarly, I analyzed cortical motor processing during tasks where the motor plan is quickly reprogrammed. In each study, I found that parameters relevant to the reach, such as target size or alternative movement plans, could be extracted by neural decoders in addition to simple kinematic parameters such as velocity and position. As such, future BMI functionality could expand to account for relevant sensory information and reliably decode intended reach trajectories, even amidst transiently considered alternatives.

The second portion of my thesis work was the successful development of the first bimanual brain-machine interface. To reach this goal, I expanded the neural recordings system to enable bilateral, multi-site recordings from approximately 500 neurons simultaneously. In addition, I upgraded the experiment to feature a realistic virtual reality end effector, customized primate chair, and eye tracking system. Thirdly, I modified the tuning function of the unscented Kalman filter (UKF) to conjointly represent both arms in a single 4D model. As a result of widespread cortical plasticity in M1, S1, supplementary motor area (SMA), and posterior parietal cortex (PPC), the bimanual BMI enabled rhesus monkeys to simultaneously control two virtual limbs without any movement of their own body. I demonstrate the efficacy of the bimanual BMI in both a subject with prior task training using joysticks and a subject naïve to the task altogether, which simulates a common clinical scenario. The neural decoding algorithm was selected as a result of a methodical comparison between various neural decoders and decoder settings. I lastly introduce a two-stage switching model with a classify step and predict step which was designed and tested to generalize decoding strategies to include both unimanual and bimanual movements.

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.

Sensory neuroprosthetics offer a revolutionary approach to studying as well as treating patients suffering from sensory dysfunction resulting from neurological impairments. Devices, such as cochlear implants, which restore the functionality of defective peripheral sensory organs, have become increasingly more prevalent and provide greater autonomy and independence to patients. For those with damage to the sensory neural circuits themselves as a result of disease or injury, alternative treatment options must be implemented. Cortical prostheses that bypass the damaged circuitry and deliver sensory information directly to the brain offer an alternative option for these patients. This approach could be used to provide tactile sensation for a prosthetic limb, restore a sense of sight in those with cortical visual impairment, or recruit intact cortex to take on the lost functionality of damaged regions of the brain. Importantly, developing devices that best serve these patient populations requires deepening our understanding of the mechanisms underlying the brain’s ability to incorporate information from a sensory prosthesis. Much of the current literature, however, focuses on behavioral and perceptual endpoints rather than changes in the brain at the mesoscopic level.

To that end, this dissertation aims to address that gap by characterizing the emergence of distributed representations of artificial sensation following the use of a cortical sensory prosthesis. Prior research has shown that adult rats could use a microstimulation-based sensory neuroprosthesis that projected information about the infrared (IR) environment onto the barrel fields of primary somatosensory cortex (S1). Equipped with this prosthesis, rats quickly learned to perform a four-choice IR discrimination task with proficiency comparable to that attained in an analogous visual discrimination task. This research established a useful paradigm for studying how the brain adapts to incorporate new sensory information projected directly onto cortex. The original research presented in this dissertation thus utilizes this paradigm for investigating how brain regions distal to the site of stimulation represent the stimulation patterns delivered by the prosthesis.

For this dissertation, I first discuss the response of two areas directly coupled to S1: the ventral posteromedial nucleus of the thalamus (VPM) - the main input nucleus to S1 and recipient of extensive corticothalamic feedback from S1 - and the posteromedial nucleus of the thalamus (POm) - a modulatory nucleus in the paralemniscal whisker pathway. Specifically, I quantify the stimulation induced response in S1, VPM, and POm. Using recordings from hundreds of multi-units from each region, the proportion of units found to have post-stimulus responses statistically distinguishable from their corresponding baseline activities was 97%, 97%, and 99% for POm, VPM, and S1, respectively. This indicates that the region of the brain affected by electrical stimulation is not constrained to the site of stimulation, but in fact downstream correlates interconnected to the stimulated region of cortex show significant responses as well.

Next I compare the presence of IR receptive field maps and the relative distribution of preferred stimulus orientations. Previously, it has been demonstrated that S1 units develop preferred stimulation patterns. That is, individual units showed variable firing rates depending on the direction to the IR source. This work replicated that finding, but more importantly I found that emergent IR receptive field maps are found in VPM and POm as well. This shows that not only do the thalamic units respond to ICMS, but undergo experience-dependent plasticity that allows the thalamic nuclei to encode the stimulus and participate in the sensory processing of the artificial sensation. A mutual information analysis was preformed to quantify the degree to which these subcortical regions represent the pattern of stimulation delivered to the cortex. The proportion of units found to have significant mutual information values was 57%, 74%, and 69% for POm, VPM, and S1, respectively. These results indicate that the artificial sensory information is readily encoded in native sensory processing circuits. Furthermore it suggests that the cortex can impose significant influence over the receptive field characteristics of thalamic nuclei even in the adult rodent brain.

Finally, I discuss the implementation of graph convolutional neural network (GCN) models to decode the stimulus features from the neural activity recorded during prosthetic use. The best performing GCN model was able to achieve a peak classification performance of 73.5% on a modified ordinal regression performance metric. Additionally, by allowing the model to learn the adjacency matrix for the neural graph data, the adjacency matrix inferred was found to provide a better representation of the underlying neural circuitry encoding the artificial sensation compared to standard techniques (i.e. cross correlation and mutual information). This further demonstrated the observation that thalamic units participated in the processing of the new sense. Because the adjacency matrix derived from training the GCNs reflects the nodes that best improve the predictions of the stimulation patterns, the adjacency matrix also serves as a method of deriving connectivity measures for the recorded units. The interpretation of these results represents a novel approach to determining functional interactions and the effective circuits involved in processing a new sensory modality.

Brain Machine Interfaces (BMI) strive to restore motor and sensory functions lost due to paralysis, amputation, and neurological diseases by interfacing brain circuitry to external actuators in form of a cursor on a computer screen or a robotic limb. There is a strong clinical need for sensory restoration as lack of somatosensory feedback leads to loss of fine motor control and one of the most common preferences for improvements according to individuals with upper-limb loss is the ability to require less visual attention to perform certain functions and to have a better control of wrist movement. One way to restore sensory functions is using electrical microstimulation of brain sensory areas as an artificial sensory channel; however, the ways of creating such artificial sensory inputs are poorly understood.

This dissertation presents the use of intracortical microstimulation (ICMS) to the primary somatosensory cortex (S1) to guide exploratory arm movements without visual feedback. Two rhesus monkeys were chronically implanted with multielectrode arrays in S1 and primary motor cortex (M1). The monkeys used a hand-held joystick to reach targets with a cursor on a computer screen. ICMS patterns were delivered to S1 when the cursor was placed over the target, mimicking the sense of touch. After the target or the cursor was made invisible, monkeys relied on ICMS feedback instead of vision to perform the task. For an invisible cursor, a random offset was added to the position of the invisible cursor to rule out the possibility that monkeys relied on joystick position felt through proprioception. Learning to perform these tasks was accompanied by changes in both the parameters of arm movements and representation of those parameters by M1 and S1 neurons at a population and individual neuronal levels.

Offline decoding of single neurons and population of neurons showed that overlapping, but not identical subpopulations of neurons represented movements when ICMS provided feedback instead of vision.

These results suggest that ICMS could be used as an essential source of sensation from prosthetic limbs.

Cortical representations of rhythmic and discrete movements are analyzed and used to create a novel neural decoding algorithm for brain-machine interfaces. This algorithm is then implemented to decode both cyclic movements and reach-and-hold movements in awake behaving rhesus macaques using their cortical activity alone. Finally, a healthy macaque wears and controls a lower body exoskeleton using the developed BMIas a proof of concept of a brain-controlled neuroprosthetic device for locomotion.

The spinal cord is an information highway continuously transmitting afferent and efferent signals to and from the brain. Although spinal cord stimulation has been used for the treatment of chronic pain for decades, its potential has not been fully explored. Spinal cord stimulation has never been used with the aim to transmit relevant information to the brain. Although, various locations along the sensory pathway have been explored for generating electrical stimulation induced sensory percepts, right from peripheral nerves, to thalamus to primary somatosensory cortex, the role of spinal cord has been largely neglected. In this dissertation, I have attempted to investigate if, electrical stimulation of dorsal columns of spinal cord called as Dorsal Column Stimulation (DCS) can be used as an effective technique to communicate therapeutic and somatosensory information to the brain.

To study the long term effects of DCS, I employed the 6-hydroxydopamine (6-OHDA) rodent model of Parkinson’s Disease (PD). Twice a week DCS for 30 minutes resulted in a dramatic recovery of weight and behavioral symptoms in rats treated with striatal infusions of 6-OHDA. The improvement in motor symptoms was accompanied by higher dopaminergic innervation in the striatum and increased cell count of dopaminergic neurons in the substantia nigra pars compacta (SNc). These results suggest that DCS has a chronic therapeutic and neuroprotective effect, increasing its potential as a new clinical option for treating PD patients. Thus, I was able to demonstrate the long-term efficacy of DCS, as a technique for therapeutic intervention.

Subsequently, I investigated if DCS can be used as a technique to transmit artificial somatosensory information to the cortex and trained rats to discriminate multiple artificial tactile sensations. Rats were able to successfully differentiate 4 different tactile percepts generated by varying temporal patterns of DCS. As the rats learnt the task, significant changes in the encoding of this artificial information were observed in multiple brain areas. Finally, I created a Brainet that interconnected two rats: an encoder and a decoder, whereby, cortical signals from the encoder rat were processed by a neural decoder while it performed a tactile discrimination task and transmitted to the spinal cord of the decoder using DCS. My study demonstrated for the first time, a cortico-spinal communication between different organisms.

My obtained results suggest that DCS, a semi-invasive technique, can be used in the future to send prosthetic somatosensory information to the brain or to enable a healthy brain to directly modulate neural activity in the nervous system of a patient, facilitating plasticity mechanism needed for efficient recovery.

Forebrain cholinergic projection systems innervate the entire cortex and hippocampus. These cholinergic systems are involved in a wide range of cognitive and behavioral functions, including learning and memory, attention, and sleep-waking modulation. However, the in vivo physiological mechanisms of cholinergic functions, particularly their fast dynamics and the consequent modulation on the hippocampus and cortex, are not well understood. In this dissertation, I investigated these issues using a number of convergent approaches.

First, to study fast acetylcholine (ACh) dynamics and its interaction with field potential theta oscillations, I developed a novel technique to acquire second-by-second electrophysiological and neurochemical information simultaneously with amperometry. Using this technique on anesthetized rats, I discovered for the first time the tight in vivo coupling between phasic ACh release and theta oscillations on fine spatiotemporal scales. In addition, with electrophysiological recording, putative cholinergic neurons in medial setpal area (MS) were found with firing rate dynamics matching the phasic ACh release.

Second, to further elucidate the dynamic activities and physiological functions of cholinergic neurons, putative cholinergic MS neurons were identified in behaving rats. These neurons had much higher firing rates during rapid-eye-movement (REM) sleep, and brief responses to auditory stimuli. Interestingly, their firing promoted theta/gamma oscillations, or small-amplitude irregular activities (SIA) in a state-dependent manner. These results suggest that putative MS cholinergic neurons may be a generalized hippocampal activation/arousal network.

Third, I investigated the hypothesis that ACh enhances cortical and hippocampal immediate-early gene (IEG) expression induced by novel sensory experience. Cholinergic transmission was manipulated with pharmacology or lesion. The resultant cholinergic impairment suppressed the induction of arc, a representative IEG, suggesting that ACh promotes IEG induction.

In conclusion, my results have revealed that the firing of putative cholinergic neurons promotes hippocampal activation, and the consequent phasic ACh release is tightly coupled to theta oscillations. These fast cholinergic activities may provide exceptional opportunities to dynamically modulate neural activity and plasticity on much finer temporal scales than traditionally assumed. By the subsequent promotion of IEG induction, ACh may further substantiate its function in neural plasticity and memory consolidation.

Brain-inspired computing architectures such as neural networks and neuromorphic chips have demonstrated promise in performing complex pattern recognition tasks by coarsely mimicking synaptic activity in software and hardware. In this dissertation, we take a departure from these more traditional methods which are confined by what we know about the dynamics of synaptic computation and introduce a brain-inspired hybrid analog-digital computing paradigm involving magnetic fields. We first review biomagnetic fields - a wide array of topics is covered to spark the interest of the reader in the field of neuro-biomagnetism and to provide a general overview of the field that explains (1) various techniques to measure, quantify, and model the magnetic signals generated by neurons; (2) how magnetic stimulation can affect neurons; and (3) the clinical relevance of these findings. These highlight the importance of magnetism in biology and neural signal processing and provide motivation for engineering magnetically-based computational devices. We then introduce a new hybrid analog-digital computing device inspired by the interplay between neural activity and its induced magnetic fields. We show that magnetic fields can interact nonlinearly in analog in a ferromagnetic medium. Specifically, the magnetic flux induced by two alternating magnetic fields can be employed to perform an absolute difference, or smooth XOR, operation. The physical structure of the analog device is based on a white matter tractography analysis; hence, we call it the neuromagnetic reactor. We also describe our design of a scalable implementation of a perceptron in hardware, which provides a digital 0-1 output. We demonstrate in a synthetic environment that these two systems together allow an organism to learn from and react appropriately to its environment. Although the design presented here is a proof-of-concept, it can be improved to yield not only new ways to study brain function but also new brain-inspired computing architectures based on magnetic fields.

Perception arises from sensory inputs detected by peripheral organs and processed in the brain by complex neuronal circuits required for the integration of external information with internal states such as expectation and attention. Stimulus discrimination requires activation of primary sensory areas in the brain, but expectation is traditionally associated with the activation of higher-order brain areas. Sensory information obtained by tactile organs is represented along the primary areas that comprise the trigeminal thalamocortical pathway. In anesthetized animals, neuronal activity in the somatosensory system has been extensively described over the past century. However, it is still unclear how the different thalamocortical structures contribute to active tactile discrimination and represent relevant features of the stimulus. It is also unknown whether expectation modulates tactile representations in these regions. In this dissertation, I investigated neuronal ensemble activity recorded from freely behaving rats performing a whisker-based tactile discrimination t-+ask. Multielectrode arrays were chronically implanted to record simultaneously from the main stages of the trigeminal thalamocortical pathways involved in whisking: the primary somatosensory cortex (S1), the ventral posterior medial nucleus of the thalamus (VPM), the posterior medial complex (POm) and the zona incerta (ZI). In Chapter 1 I describe the behavior of rats performing the tactile discrimination task, which requires animals to associate two different tactile stimuli with two corresponding choices of spatial trajectory in order for reward to be delivered. I found that both cortical and thalamic neurons are dynamically engaged during execution of the task. The data reveal a very complex mosaic of responses comprising single or multiple periods of inhibition and excitation. Thalamocortical activity was modulated during whisker stimulation as well as after stimulus removal, up until reward delivery. To investigate whether reward expectation plays a role in tactile processing at early processing stages, I also recorded neuronal activity from rats performing a freely-rewarded version of the tactile discrimination task. Comparing data from regularly-rewarded and freely-rewarded sessions, I show in chapter 2 that the activity of single neurons in the primary somatosensory thalamocortical loop is strongly modulated by reward expectation. Stimulus-related information coded by primary thalamocortical neurons is high when a correct association between stimulus and response is crucial for reward, but decreases significantly when the association is irrelevant. These results indicate that tactile processing in primary somatosensory areas of the thalamus and cerebral cortex is directly affected by reward expectation.

Traditionally, brain-machine interfaces (BMI) recorded from neurons in cerebral

cortical regions associated with voluntary motor control including primary motor

(M1), primary somatosensory (S1), and dorsal premotor (PMd) cortices. Wheelchair

BMI where users’ desired velocity commands are decoded from these cortical neu-

rons can be used to restored mobility for the severely paralyzed. In addition,

spatial information in these areas during navigation can potentially can incorpo-

rated to bolster BMI performance. However, the study of spatial representation

and navigation in the brain has traditionally been centered on the hippocampal

structures and the parietal cortex, with the majority of the studies conducted in

rodents. Under this classical model, S1, M1, and PMd would not contain allocen-

tric spatial information. In this dissertation I show that a significant number of

neurons in these brain areras do indeed represent body position and orientation

in space during brain-controlled wheelchair navigation.

First, I describe the design and implementation of the first intracortical BMI

for continuous wheelchair navigation. Two rhesus monkeys were chronically im-

planted with multichannel microelectrode arrays that allowed wireless recordings

from ensembles of premotor and sensorimotor cortical neurons. While monkeys

remained seated in the robotic wheelchair, passive navigation was employed to

train a linear decoder to extract wheelchair velocity from cortical activity. Next,

monkeys employed the wireless BMI to translate their cortical activity into the

ivwheelchair’s translational and rotational velocities. Over time, monkeys improved

their ability to navigate the wheelchair toward the location of a grape reward. The

presence of a cortical representation of the distance to reward location was also

detected during the wheelchair BMI operation. These resutls demonstrate that

intracranial BMIs have the potential to restore whole-body mobility to paralyzed

patients.

Second, building upon the finding of cortical representation of the distance

to reward location, I found that during wheelchair BMI navigation the discharge

rates of M1, S1, and PMd neurons correlated with the two-dimensional (2D) room

position and the direction of the wheelchair and the monkey head. The activities

of these cells were phenomenologically similar to place cells and head direction

(HD) cells found in rat hippocampus and entorhinal cortices. I observed 44.6%

and 33.3% of neurons encoding room position in the two monkeys, respectively,

and the overlapping populations of 41.0% and 16.0% neurons encoding head di-

rection. These observations suggest that primary sensorimotor and premotor cor-

tical areas in primates are likely involved in allocentrically representing body po-

sition in space during whole-body navigation, which is an unexpected finding

given the classical model of spatial processing that attributes the representation of

allocentric space to the hippocampal formations.

Finally, I found that allocentric representation of body position in space was

not clear during passive wheelchair navigation. Two rhesus monkeys were pas-

sively transported in an experimental space with different reward locations while

neuronal ensemble activities from M1 and PMd were recorded wirelessly. The ac-

tivities of the recorded cells did not clearly represent the position and direction

of the wheelchair. These results suggest active navigation might be a prerequisite

for primary sensorimotor and PMd participation in the allocentric representation

of space.

In summary, dorsal premotor and primary sensorimotor cortical correlates of

body position and orientation in space were found in rhesus monkeys during

the operation of an intracortical wheelchair BMI for navigation. These findings

contradict the classical dichotomy of localized spatial processing, support a dis-

tributed model of spatial processing in the primate brain, and suggest both con-

text and species differences are important in neural processing. The incorporation

of the allocentric spatial information present in these cortical areas during brain-

controlled wheelchair navigation can potentially improve future BMI navigation

performance.

Neural activities of the motor cortices have been traditionally known to represent motor information such as velocity of the movement and muscle force. Recent studies show that motor cortices, including primary motor cortex (M1), also represent non-traditional information such as observed movements of others and reward-related signal. However, how the neurons simultaneously multiplex such non-traditional information along with traditional motor parameters and whether the multiplexing leads to significant interactions are not well understood. Furthermore, understanding how the non-traditional information are encoded and they interact with motor information may help the development of more error-resistant, autonomous brain-to-machine interface and the understanding of underlying mechanism behind joint action and motor skill learning. In this dissertation, we investigate in detail how the observed movements and reward are simultaneously multiplexed along with traditional motor information and how each pair of neural representations interact with each other. First, regarding movement observation, we show that significant fraction of M1 neurons simultaneously encode the presence and direction of the movement of others along with those of self-movements. Neurons respond differently to joint action than to self-movements and show an interaction effect from the two representations of observed and executed movements rather than simple averaging of the two. Some neurons that separately encode observed and executed movements turn to suppress the representation of observed movements in joint action. In simultaneous actions, the representation of self-executed movement gets weaker, which suggests an interaction between two information and may possibly lead to behavioral interference. Preferred directions also change to be decoupled for noncongruent joint actions as to allow simultaneous multiplexing of both information with phase difference, while being synced for congruent ones. Conditional probabilities from the distribution of encoding neurons suggest a shared circuitry for movement observation, execution, and simultaneous actions. Shared circuitry with interactions between representations may explain why people can perform movements freely while watching others move; yet if the interaction between the two goes up due to simultaneous occurrence, it may result in interferences in behavior. Second, regarding the multiplexing of reward-related signal with movement signals, we show that both signals are multiplexed in individual and population neurons in M1 and S1. The activity of neural population in M1 and S1 distinguished whether the reward timing before the delivery of the reward. Furthermore, reward per se, reward anticipation, and reward prediction error (RPE) were encoded along with the motor information. The encoding of the reward-related signal interacted with the motor information in that the preferred direction changed when the reward was omitted. Change of spatial tuning of neurons due to reward prediction error signifies that there is interaction between the neural representation of reward and motor information, which may impact and underlie motor skill learning. In conclusion, both observed movements and reward are simultaneously multiplexed with traditional motor information. Co-representation of the two non-traditional information then leads to interaction between them and the motor information. Such interaction suggest that such simultaneous multiplexing may lead to behavioral interferences and motor skill learning.

Brain-machine interfaces (BMIs) use recordings from the nervous system to extract volitional and motor parameters for controlling external actuators, such as prosthetics, thereby bypassing or replacing injured tissue. As such, they show enormous promise for restoring mobility, dexterity, or communication in paralyzed patients or amputees. Recent advancements to the BMI paradigm have made the brain -- machine communication channel bidirectional, enabling the prosthetic to inform the user about touch, temperature, strain, or other sensory information; these devices are hence called brain-machine-brain interfaces (BMBIs).

In the first chapter an intraoperative BMI is investigated in human patients undergoing surgery for implantation of a deep brain stimulation (DBS) treatment electrodes. While the BMI was marginally effective, we found high levels of behavioral and tremor tuning among cells recorded from the surgical targets, the subthalamic nucleus (STN) and ventral intermediate nucleus (VIM) of the thalamus. Notably, this tremor or behavior tuning was not mutually exclusive with oscillatory behavior, suggesting that physiological tuning persists even in the face of pathological oscillations. We then used nonlinear means for extracting tremor tuning, and found a significant population, consistent with double-frequency or co-modulation to tremor within the basal ganglia. Synchrony was then assessed over long and short timescales between pairs of neurons, and it was found that tremor tuning implies synchrony: all units exhibiting tremor tuning showed synchrony to at least one other unit.

BMBIs rely on a host of both scientific knowledge and technology for effective function, and this technology is currently in intensive research. In this dissertation two technologies for BMBIs, corresponding to the two directions of communication, are designed, described, and tested. The first one is a high compliance, digitally controlled, high-side current-regulated microstimulator for intracortical microstimulation (ICMS). The device is validated on the bench, tested in monkeys, and used for multiple experimental setups. Due to careful control of parasitic charge injection, the microstimulator is ideally suited for interleaving stimulation and recording as employed in some BMBIs.

The second technology described is a wireless, scalable, 128 channel neural recording system. The device features aggressive digital filtering to maximize signal quality, has spike sorting and compression on the transceiver, can be fully configured over the air through a custom wireless bridge and client software, and can run for over 30 hours on one battery. This system has been tested in a monkey while in its home cage, where the wireless system permitted unfettered, continuous recording and continuous access to a simplified BMI. A full description of the development and device is described, as well as results showing convincing 1D and suggestive 2D BMI control.

In recent years, large-scale brain recordings in nonhuman primates have been a driving force for both fundamental neuroscience and the field of brain-machine interfaces (BMIs). This required monkey implants connected to external amplifiers and computers with ever increasing number of cables. As shown with our recent demonstration of 2,000 neurons recorded in one monkey, a tethered recording system begins to get bulky and complex, particularly for our BMI and neurophysiological research. To address this problem, we developed a multichannel wireless recording framework. The system was been tested in freely moving rhesus monkey by integrating wireless neural recordings with external computers performing BMI decoding, behavioral manipulanda and optical tracking. This technology can be applied to primate behavior research and, in the near future, wireless, fully implantable human neuroprosthetics, which is of great significance to those suffering from locomotor deficiencies, such as those brought on by spinal cord injury and stroke. Aided with these advances, I was able to study monkeys in unrestrained locomotion while their cortical activity was continuously monitored. I also explored unrestrained behaviors and how they showed distinct transitions in neural dynamics as monkeys engaged in different behavioral activities or learned new motor skills, such as bipedal walking. I was able to decode them many of these behavioral states from cortical activity with neural classifiers. Lastly, monkeys were able to perform BMI tasks continuously for many hours, allowing us to prove the relevance of unrestrained noise in BMI performance. Lastly, I present my role in developing two brain actuated movement platforms, a robotic exoskeleton under the guise of the WalkAgain project, and a microelectrode BMI enabled wheelchair. This body of work should assist those on the path to the next generation of clinical neuroprostheses and neural communication systems.