From Prosthesis to proprioceptive Perception via neuromechanical modelling and Optogenetic cortical stimulation
Proprioception provides the brain with a readout of the musculoskeletal system kinetics and kinematics. The loss of these sensory inputs due to disease (e.g. diabetes, viral infection) or amputation profoundly affects motor control and leads to the loss of movement accuracy, as well as to chronic pain and the loss of perceived body ownership. In contrast to vision or touch, the cortical representations of proprioceptive inputs are intermingled at the surface of the cortex, and do not follow a topographical organization. So far this has hindered the successful restoration of proprioceptive feedback using mesoscale cortical stimulation patterns, in contrast to the promising results obtained by this method towards the direct cortical restoration of other sensory modalities. Achieving the restoration of proprioception via direct cortical stimulation holds great potential for significantly improving prosthesis control and enhancing sensory integration in neuroprosthetic systems. Here, we hypothesize that proprioceptive feedback can be restored in an animal mouse model via direct cortical stimulations, by combining (1) real-time prediction of proprioceptive signals using a biomechanical musculoskeletal model of the mouse forelimb and (2) deep learning to construct a topographically structured representation of the simulation-generated data flow. To test this hypothesis, we will implement such a proprioceptive feedback capability into a motorized mouse upper limb prosthesis, and we will train mice to solve manipulation tasks that depend on the control of the prosthesis. We will then explore the space of possible proprioceptive feedback designs, based on the ground truth of the behavioral performance achieved with the artificial proprioceptive feedback from the prosthesis.
3/6 complete
Imported from OpenAIRE Germany | Grant ANR-25-CE33-4035.
Open sourceOverview
Grant agreement ANR-25-CE33-4035 | ProPerO
Open sourceProprioception provides the brain with a readout of the musculoskeletal system kinetics and kinematics. The loss of these sensory inputs due to disease (e.g. diabetes, viral infection) or amputation profoundly affects motor control and leads to the loss of movement accuracy, as well as to chronic pain and the loss of perceived body ownership. In contrast to vision or touch, the cortical representations of proprioceptive inputs are intermingled at the surface of the cortex, and do not follow a topographical organization. So far this has hindered the successful restoration of proprioceptive feedback using mesoscale cortical stimulation patterns, in contrast to the promising results obtained by this method towards the direct cortical restoration of other sensory modalities. Achieving the restoration of proprioception via direct cortical stimulation holds great potential for significantly improving prosthesis control and enhancing sensory integration in neuroprosthetic systems. Here, we hypothesize that proprioceptive feedback can be restored in an animal mouse model via direct cortical stimulations, by combining (1) real-time prediction of proprioceptive signals using a biomechanical musculoskeletal model of the mouse forelimb and (2) deep learning to construct a topographically structured representation of the simulation-generated data flow. To test this hypothesis, we will implement such a proprioceptive feedback capability into a motorized mouse upper limb prosthesis, and we will train mice to solve manipulation tasks that depend on the control of the prosthesis. We will then explore the space of possible proprioceptive feedback designs, based on the ground truth of the behavioral performance achieved with the artificial proprioceptive feedback from the prosthesis.