Sensorimotor control of fine grasping in non-human primates

Performance site: Fribourg

Introduction

Information about body position and movement (proprioception) is critical to the central nervous system in order to adapt the generation of movement to its environment. We use our limbs to manipulate objects, but also to communicate with and understand our environment. During these interactions, we use among others touch and proprioception as feedback to adjust our movements and actions with the aim of fulfilling a task, such as opening a door or carrying a bag of groceries. However, while touch as feedback modality has been extensively studied, proprioception and its importance for movement control and postural adjustment are far less understood.

Bachelor/semester projects 

Project 1: Kinematic correlates of fine grasping

Kinematic (movement in 3D) and kinetic (force) information define the type and nature of different grasping movements such as spherical, cylindrical or pinch grips. Here, we will analyze those data recorded from a non-human primate performing a reach-and-grasp task. A robotic arm presents a target object to the animal, which is prompted by a sound and light cue to reach for it, grasp, pull, and then release it. The kinematics of the arm and hand as well as the grasping force applied to the objects and the pulling force exerted by the animal are recorded in real time.

The project will consist in extracting and reconstructing the hand and arm joint position, speed, and acceleration as well as the grasping and pulling forces. Different data analyses techniques are then used to compare those signals for objects of different shapes and size (thus implying different grasp types) and for various level of resistance applied during the pulling movement.

Master projects

Project 1: S1 cortical dynamics of precision grasping

Reaching movements have been associated with oscillatory dynamics in M1 (Churchland 2012), while hand movements and specifically grasping do not seem to exhibit the same neural dynamics in M1 but also not in S1 (Suresh SfN 2018). One possible explanation of those fundamental differences could be that reaching movements are governed by a stronger reflex-based component than grasping movements, explaining the oscillatory dynamics that can be found in M1 cortical activity during reaching but not during grasping. If this hypothesis was true, grasping movements would be regulated by a stronger direct corticospinal connection, and the cervical spinal cord would exhibit an increased density of those connections at the lower cervical segments that are known to relate to hand and fingers, while segments that are more rostral innervate the arm.

In order to assess this hypothesis, we first validate that reaching but not grasping exhibit rotational dynamics through an oscillatory component in M1 and S1 population activity in our experimental dataset through dynamical system analysis. We recorded arm and hand kinematics and intracortical activity in S1 and M1 of a non-human primate during reaching and precision grasping tasks. Based on those results, we will design electrophysiological and anatomical studies to investigate the corticospinal projections along the cervical spinal cord.

The master thesis work will focus on the neural dynamics analysis. Applicants must be proficient in Matlab and have basic knowledge in biological signal processing.

Start: From January 2019. Contact: sophie.wurth@epfl.ch

Project 2: Information content of local field potentials during precision grasping

Local field potentials (LFP) represent the low frequency component of intracortical neural recordings and contain information about the neural activity at a population and network level in contrast to single unit activity. Despite being a coarser recording, the information content in LFP is extremely high as it captures underlying network dynamics (Magri 2012, Pesaran 2012). LFP are typically sampled at about an order of magnitude lower than multiunit activity, and represent thereby an excellent candidate for use in brain-computer interfaces as they ease the burden of data transmission in real-time.

Here we compare the information content of LFP with multi- and single-unit activity from neural recordings in M1 and S1 to identify different precision grasps during unconstrained behavior in non-human primates by looking at the decoding strength of grasp types and/or hand and finger joint kinematics using LFP vs. multiunit activity.

Data analysis thesis. Applicants must be proficient in Matlab and have basic knowledge in biological signal processing.

Start: From January 2019. Contact: sophie.wurth@epfl.ch

Project 3: Interaction M1 and S1 during precision grasping

Activity in M1 and PM during motor tasks is relatively well documented (Saleh 2012, Vargas-Irwin 2010, Davare 2011 among others). The contribution of S1 in that it provides feedback to M1 during movement execution is known, however the interactions between both cortices before and during movement execution are not fully understood. To investigate those relationships, we analyze the coherence between M1 and S1 activation during reaching and grasping and look at mutual information of both with respect to movement kinematics.

Data analysis thesis. Applicants must be proficient in Matlab and have basic knowledge in biological signal processing.

Start: From January 2019. Contact: sophie.wurth@epfl.ch

 

Project 4: Representation of peripheral nerve stimulation (PNS) in S1

The mechanisms of signaling hand proprioception via the peripheral nerves and how these signals are encoded in the primary somatosensory cortex remain elusive. In this framework, we will compare the responses obtained in S1 by tendon vibration or during passive finger and hand joint movements with selective intraneural subthreshold PNS to recruit large myelinated afferent fibers (Ia fibers mediating proprioception) within the median and radial nerves.

We will analyze how the natural (through passive movement or tendon vibration) and artificial (through PNS) fiber recruitment is represented in the primary somatosensory cortex representing the hand and forearm. We will further assess whether the elicited activity can be modulated (amplitude, pulse width, frequency) and we will determine the stability of these responses over time. The data collection will be carried out during anesthetized experiments in non-human primates (NHP) implanted with intracortical arrays in S1 and M1, intraneural electrodes in the median and ulnar nerves, and intramuscular EMG wire electrodes in hand and forearm muscles.

Data analysis thesis plus participation in data collection. Applicants must be proficient in Matlab and have basic knowledge in biological signal processing.

Start: From March 2019. Contact: sophie.wurth@epfl.ch

 

Project 5:  Proprioceptive feedback mediated by PNS during behavior

The goal of this project is to understand whether we are able to elicit proprioceptive signals by subthreshold peripheral nerve stimulation (PNS) during behavioral experiments in non-human primates. For this, we will trigger PNS during the reaching phase of the movement and record arm and hand EMG signals and joint kinematics to see whether the stimulation modifies the movement, and if so, how this change is represented in S1 compared to unperturbed trials.

Data analysis thesis plus participation in data collection. Applicants must be proficient in Matlab and have basic knowledge in biological signal processing.

Start: From March 2019. Contact: sophie.wurth@epfl.ch

Project 6: Deep learning-based personalized stimulation strategy for hand movement restoration

The goal of this project is to develop a deep-learning algorithm to predict the optimal patterns of electrical stimulation that must be applied to arm peripheral nerves to recruit the muscles of interest. This algorithm will make use of a simplified model of peripheral nerve stimulation with intraneural electrode and combine it to in vivo data collected in the primate model to personalize and optimize the stimulation strategies that must be applied in order to recruit hand muscles in a certain manner (i.e, optimization of frequency, amplitude, multipolar interactions …).

By the end of the project, the algorithm will be tested and validated in vivo. Such model development would not only improve stimulation paradigm design in primates but could be easily translated to humans.

Computational thesis plus in vivo validation. Applicants must be proficient in python and basic knowledge in machine learning algorithms is a plus.

Start: From January 2019. Contact: marion.badi-dubois@epfl.ch

References:

Ethier, C., Oby, E.R., Bauman, M.J., and Miller, L.E. (2012). Restoration of grasp following paralysis through brain-controlled stimulation of muscles. Nature 485, 368–371