Advances in brain-machine interfaces by closing the loop between single-neuron operant conditioning and sensory reafference – NEUROWHISK

In Task 1, we have developed a whisker discrimination task in a modified T-maze. This has required building an automated maze in which the animal behavior is controlled by motorized doors, position detectors, motorized stimuli and reward distributors. This new setup has allowed to train rats during months, thus probing their learning capabilities.
In Task 3, we are building a new setup for head-fixed mice incorporating multi-channel electrophysiology and light delivery to the cortex. We can deliver light either through an optic fiber implanted on the skull and coupled to a laser diode system, or using a mirror-based light projector allowing patterned light stimulation.

In the behavioral discrimination task developed for the tactile sense of the rat, the animals successfully differentiated the patterned surface from a smooth surface, and their performance dropped when the pattern was changed to an irregular one. We showed that this behavior was dependent on the integrity of the whiskers, as well as the integrity of the primary somatosensory cortex. These results are being summarized in a manuscript that will be submitted to Journal of Neuroscience (Kerekes, Daret, Shulz & Ego-Stengel, in preparation). This novel task can be applied to exploring further the tactile discrimination capabilities of the rat, and adapted to the mouse.
Tasks 2 and 3 are still being developed, but should yield the majority of the results of the project, focusing on whether sensory feedback increases the learning and performance of a neuroprosthesis control.

With this project, we hope to contribute both to our knowledge of brain physiology, and to the development of efficient neuroprostheses. Our research directly addresses the potential of adaptation in neural networks, a question of utmost importance in the quest for useful neuroprostheses. The ability to record the activity of neuronal populations which have been disconnected from body effectors, and use it to drive artificial actuators replacing the muscles, is a promising avenue. We hope that our research project will help bridge the gap between purely fundamental research and its applicability, thus advancing towards the goal of bringing back autonomy to motor-impaired patients.

The results obtained in the behavioral task (Task 1) will be published soon (Kerekes et al, will be submitted to J Neuroscience) and will be presented as posters at International Congresses (FENS 2016, SfN 2016). Several invited seminars have also taken place in order to introduce our planned project of Tasks 2 & 3.

Our research project is aimed at understanding the neuronal activity patterns underlying learning of a new behavioral task, specifically, the control of a motor neuroprosthesis incorporating sensory feedback. Throughout their life, animals learn new information and integrate it with prior knowledge. This process is necessary in order to adapt the organism's behavior optimally to the ever-changing environment. To support such learning, the adult brain undergoes modifications on multiple timescales. Our general aim is to understand what type of neural-based events and plasticity induction rules guide dynamical changes in primary sensory and motor cortical neuronal networks during the acquisition of new perceptual experiences, and how they trigger long-lasting modifications in the immediate and downstream networks.
One challenging interdisciplinary approach is provided by the neuroprosthetics framework, in which robotic devices are interfaced with body parts in order to restore the loss of sensory perception or the capacity of action, eventually combined with sensory reappropriation of the impaired limb. In addition to the formidable hope they represent for sensory- and motor-impaired subjects, neuroprostheses also constitute ideal models for testing hypotheses of neural coding and central plasticity mechanisms. On the motor side, by transforming neural patterns into output control signals executable by external actuators, they allow to extract a more or less explicit read-out of the cortical activity states associated to predictable movements of body parts, allowing in certain cases to learn more about the neural code itself of action planning. On the sensory side, conveying information to the brain about external stimuli requires a more causal approach akin to coding external information with events interpretable directly by the activated neural assembly, ie writing the neural sensory code. In both instances, the ultimate success of the attempts probably involves a large degree of plasticity of the brain structures interfaced with the neuroprostheses, and one can use this to explore the nature and extent across multiple cortical networks of the underlying adaptive mechanisms.
In the last 15 years, the design of efficient neuroprothetic devices has become a major challenge towards the long-term goal of restoring autonomy to sensory and/or motor-impaired patients. Current devices are however often rejected by the patients, mostly because they have limited capabilities and do not “feel” like body parts. In order to improve both dexterity and embodiment of neuroprostheses, researchers need to incorporate sensory feedback until the use of the device operates naturally with a feedback quality close to that normally provided by the simulated sensory modality. In this project, we propose to investigate directly the neuronal activity of sensory and motor cortical assemblies recorded simultaneously in a freely-behaving rodent while the animal is learning abstract rules to control a particular neuroprosthesis, with or without integrated sensory feedback. Following our previous work in the field, the spiking activity of individual motor cortex neurons will be trained, by operant conditioning, to control an actuator delivering a reward to the animal. Tactile feedback signals will be delivered either by presenting stimuli to the whiskers of the animal, or by electrical intracortical microstimulation in the primary sensory cortex. We will determine the impact of this sensory feedback on the efficiency and reliability of the neuroprosthesis. We hope that our results will help understand fundamental principles of neural processing and plasticity, to be exploited in the realization of ergonomic neuroprostheses.