Final Report Summary - NEUROPLAST (Implementation of a Novel Brain Machine Interface to Restore Limb Movement and Promote Recovery from Partial Spinal Cord Injury: Basic Studies and Clinical Application)
Spinal cord injury (SCI) causes alterations in the brain, as well as the sensorimotor and autonomic systems below the injury, and has devastating personal and socioeconomic costs. Although there is currently no cure, new research opportunities offer the prospect of accelerating both our understanding of the disorder, and the design of therapies to promote recovery. In this project I began to investigate, in both animal models and patients, novel rehabilitation techniques that consist of activating the paralyzed muscles electrically, driven by the subject’s own movement intent. I expect long-term use of this type of motor neuroprostheses will lead to unprecedented levels of restored movement, while the subject uses the system, and to maintained functional gain even without the neuroprosthesis, after therapy is complete.
This hypothesis is based on the mounting evidence that the nervous system has an important ability to reorganize and heal itself, a mechanism referred to as “neural plasticity.” This evidence comes from studies in cell culture, animals and, in more limited number, humans. Most of these observations are based on Hebb’s principle that “neurons that fire together wire together.” This idea is central to the project: our hypothesis is that neuroprostheses that help the subject to generate an intended motion will drive neural plasticity and foster lasting recovery after SCI –and other neurological diseases. Thus, my work on this project revolves around two aims: 1) animal studies to better understand how the brain forms motor commands and causes movement, and 2) development and validation of neuroprostheses to drive neural plasticity and foster recovery after long-term use, both in animal models and SCI patients. These goals are both relevant basic science questions and key for the development of novel rehabilitation techniques.
The classic approach to understand how the brain causes movement has been to try to identify lawful relationships between the activity of single neurons and behavioral parameters. Although this framework was the basis for many interesting studies, it is now apparent that there is not a single, robust lawful relationship between neural activity and behavior. Recent and accelerating technical developments provide the experimental tools for monitoring the simultaneous activity of hundreds or even thousands of neurons. Thanks to these tools, a number of groups are investigating movement planning and execution at the population level. To understand population level processing, we are focusing on the emerging “neural manifold” framework. The basic idea of this framework is that brain function is based on specific patterns of correlated activity that are determined by network connectivity, the neural modes (Fig. 1A,B), rather than on the independent modulation of single neurons.
Using this framework, we investigated the similarities between neural modes from different tasks in primary motor cortex (M1), the main output motor area of the brain. We found that the structure of the neural modes is largely preserved for different skilled arm and hand tasks (Fig. 1C). Moreover, the activation of some of these well-preserved modes followed very similar dynamics for different tasks; this is quite surprising given the large differences in motor output and single neuron activity across tasks. A subset of these modes even captured a constant mapping between cortical activity and commands to muscles. These findings, elusive when examining single neuron activity, provide insight into how the brain controls movement. They also suggest that movement intent estimated using neural modes may provide the most appropriate control signal for a motor neuroprosthesis.
We also applied this conceptual framework to study how the brain quickly adapts movement in the presence of an external force field (short-term learning). By examining the relationship between the activity of populations of neurons in premotor cortex (PMd) and M1, we found that fast behavioral adaptation is not mediated by fast plastic changes in the motor cortices. Instead, during force field adaptation, planning-related activity in premotor cortex explores new patterns within the region of the PMd manifold that does not directly effect on M1. These new PMd patterns recruit M1 in novel ways, thus driving movement adaptation. Interestingly, when we studied another classic motor adaptation task, the visuomotor rotation task, we found that in this case adaptation happens upstream of the motor cortices. This observation is in agreement with theoretical and experimental work in humans. Therefore, we identified a new population mechanism that mediates short-term learning. This mechanism also has implications for neuroprosthetics. It suggests that a neuroprosthesis should rely on activity patterns that lie in the existing manifold and that are “transmitted” through an existing mapping, so as to maximize ease of learning.
To further our current understanding of how neural populations in motor and sensory cortices control movement, I established a collaboration with the Pachitariu Group in Janelia Research Centre. Together, we have begun to investigate how populations of thousands of neurons compute movement-related information based on “local” and “global” neural modes. We are doing this using a wide field two photon microscope (a “mesoscope”) that records simultaneously from a large portion of mouse motor, sensory and retrosplenial cortices (Fig. 1D-F).
In this project, I also integrated the first fully wireless cortically-controlled neuroprosthesis to restore movement. This neuroprosthesis comprises a wireless “neurosensor” that transmits neural data in real-time, a computer that maps them onto stimulation commands, and a wireless, fully controllable multichannel muscle stimulator (Fig. 2A). The neuroprosthesis also allows us to record muscle activity in free behaving monkeys with another wireless transmitter. During pilot experiments with the neuroprosthesis in the laboratory, we restored hand use to non-human primates (macaca mulatta) that had received a temporary nerve block. The final objective of this study is to investigate long-term adaptation to neuroprosthesis use. To this end, I developed a month-long, reversible paralysis model using an implanted mini-pump that delivered the nerve blocking agent tetrodotoxin. Our group has also begun to record neural and muscle activity while the monkeys move freely in a purposely built plastic cage (Fig. 2B,C). With this unique setup, we are beginning to investigate the basic mechanisms of motor control during free behavior using the neural manifold framework. This ongoing project, led by my Outgoing Host, Prof. Lee Miller (Northwestern University, Chicago, USA) has tremendous potential to impact both rehabilitation and neuroscience.
The reversible paralysis model has obvious ethical and practical advantages, but does not replicate some of the complications that follow an SCI, like spasticity or hyperreflexia. For this reason, I also collaborated in the development of a similar cortically-controlled neuroprosthesis to restore movement using muscle stimulation in rats (rattus norvegicus). The final goal of this ongoing study is to show that assisted movement that accurately replicates the motor intent induces neural plasticity and drives motor recovery after a real SCI.
Finally, in the last year of my Marie Curie, my collaborators and I also implemented a non-invasive version of this neuroprosthesis to restore hand function in SCI patients. This neuroprosthesis does not detect motor intent with intracortical electrodes, but through recordings of residual muscle activity. This activity is then used to activate the paralyzed muscles with preprogrammed muscle stimulation patterns. In future experiments, we will test if the association of motor intent and assisted movement will induce associative plasticity and drive lasting recovery after an incomplete SCI, as hypothesized.
In summary, this project has helped advanced our understanding of how the brain controls movement, and of how neural populations coordinate their activity to adapt to a changing environment. We have also developed new neuroprostheses to drive neural plasticity based on the association of movement intent and artificially generated movement. We have begun to test this approach in monkeys, and will soon do it in SCI rats and human patients. In future work, I intend to continue pursuing both these goals. Understanding the brain is perhaps the greatest scientific challenge of our time. I am convinced that many basic discoveries in neuroscience will impact our lives to a extent we cannot foresee.
This hypothesis is based on the mounting evidence that the nervous system has an important ability to reorganize and heal itself, a mechanism referred to as “neural plasticity.” This evidence comes from studies in cell culture, animals and, in more limited number, humans. Most of these observations are based on Hebb’s principle that “neurons that fire together wire together.” This idea is central to the project: our hypothesis is that neuroprostheses that help the subject to generate an intended motion will drive neural plasticity and foster lasting recovery after SCI –and other neurological diseases. Thus, my work on this project revolves around two aims: 1) animal studies to better understand how the brain forms motor commands and causes movement, and 2) development and validation of neuroprostheses to drive neural plasticity and foster recovery after long-term use, both in animal models and SCI patients. These goals are both relevant basic science questions and key for the development of novel rehabilitation techniques.
The classic approach to understand how the brain causes movement has been to try to identify lawful relationships between the activity of single neurons and behavioral parameters. Although this framework was the basis for many interesting studies, it is now apparent that there is not a single, robust lawful relationship between neural activity and behavior. Recent and accelerating technical developments provide the experimental tools for monitoring the simultaneous activity of hundreds or even thousands of neurons. Thanks to these tools, a number of groups are investigating movement planning and execution at the population level. To understand population level processing, we are focusing on the emerging “neural manifold” framework. The basic idea of this framework is that brain function is based on specific patterns of correlated activity that are determined by network connectivity, the neural modes (Fig. 1A,B), rather than on the independent modulation of single neurons.
Using this framework, we investigated the similarities between neural modes from different tasks in primary motor cortex (M1), the main output motor area of the brain. We found that the structure of the neural modes is largely preserved for different skilled arm and hand tasks (Fig. 1C). Moreover, the activation of some of these well-preserved modes followed very similar dynamics for different tasks; this is quite surprising given the large differences in motor output and single neuron activity across tasks. A subset of these modes even captured a constant mapping between cortical activity and commands to muscles. These findings, elusive when examining single neuron activity, provide insight into how the brain controls movement. They also suggest that movement intent estimated using neural modes may provide the most appropriate control signal for a motor neuroprosthesis.
We also applied this conceptual framework to study how the brain quickly adapts movement in the presence of an external force field (short-term learning). By examining the relationship between the activity of populations of neurons in premotor cortex (PMd) and M1, we found that fast behavioral adaptation is not mediated by fast plastic changes in the motor cortices. Instead, during force field adaptation, planning-related activity in premotor cortex explores new patterns within the region of the PMd manifold that does not directly effect on M1. These new PMd patterns recruit M1 in novel ways, thus driving movement adaptation. Interestingly, when we studied another classic motor adaptation task, the visuomotor rotation task, we found that in this case adaptation happens upstream of the motor cortices. This observation is in agreement with theoretical and experimental work in humans. Therefore, we identified a new population mechanism that mediates short-term learning. This mechanism also has implications for neuroprosthetics. It suggests that a neuroprosthesis should rely on activity patterns that lie in the existing manifold and that are “transmitted” through an existing mapping, so as to maximize ease of learning.
To further our current understanding of how neural populations in motor and sensory cortices control movement, I established a collaboration with the Pachitariu Group in Janelia Research Centre. Together, we have begun to investigate how populations of thousands of neurons compute movement-related information based on “local” and “global” neural modes. We are doing this using a wide field two photon microscope (a “mesoscope”) that records simultaneously from a large portion of mouse motor, sensory and retrosplenial cortices (Fig. 1D-F).
In this project, I also integrated the first fully wireless cortically-controlled neuroprosthesis to restore movement. This neuroprosthesis comprises a wireless “neurosensor” that transmits neural data in real-time, a computer that maps them onto stimulation commands, and a wireless, fully controllable multichannel muscle stimulator (Fig. 2A). The neuroprosthesis also allows us to record muscle activity in free behaving monkeys with another wireless transmitter. During pilot experiments with the neuroprosthesis in the laboratory, we restored hand use to non-human primates (macaca mulatta) that had received a temporary nerve block. The final objective of this study is to investigate long-term adaptation to neuroprosthesis use. To this end, I developed a month-long, reversible paralysis model using an implanted mini-pump that delivered the nerve blocking agent tetrodotoxin. Our group has also begun to record neural and muscle activity while the monkeys move freely in a purposely built plastic cage (Fig. 2B,C). With this unique setup, we are beginning to investigate the basic mechanisms of motor control during free behavior using the neural manifold framework. This ongoing project, led by my Outgoing Host, Prof. Lee Miller (Northwestern University, Chicago, USA) has tremendous potential to impact both rehabilitation and neuroscience.
The reversible paralysis model has obvious ethical and practical advantages, but does not replicate some of the complications that follow an SCI, like spasticity or hyperreflexia. For this reason, I also collaborated in the development of a similar cortically-controlled neuroprosthesis to restore movement using muscle stimulation in rats (rattus norvegicus). The final goal of this ongoing study is to show that assisted movement that accurately replicates the motor intent induces neural plasticity and drives motor recovery after a real SCI.
Finally, in the last year of my Marie Curie, my collaborators and I also implemented a non-invasive version of this neuroprosthesis to restore hand function in SCI patients. This neuroprosthesis does not detect motor intent with intracortical electrodes, but through recordings of residual muscle activity. This activity is then used to activate the paralyzed muscles with preprogrammed muscle stimulation patterns. In future experiments, we will test if the association of motor intent and assisted movement will induce associative plasticity and drive lasting recovery after an incomplete SCI, as hypothesized.
In summary, this project has helped advanced our understanding of how the brain controls movement, and of how neural populations coordinate their activity to adapt to a changing environment. We have also developed new neuroprostheses to drive neural plasticity based on the association of movement intent and artificially generated movement. We have begun to test this approach in monkeys, and will soon do it in SCI rats and human patients. In future work, I intend to continue pursuing both these goals. Understanding the brain is perhaps the greatest scientific challenge of our time. I am convinced that many basic discoveries in neuroscience will impact our lives to a extent we cannot foresee.
