Implant cérébral | EPFL Graph Search

Les implants cérébraux, également appelés « BrainChips » ou « Brain implant » en anglais, sont des systèmes électroniques implantés dans le cerveau, afin de lire ou contrôler certains signaux cérébraux. Ils sont également connus sous le nom de « puce électronique cérébrale ». Description Les implants cérébraux sont des implants électroniques constitués d'électrodes ou de grilles d'électrodes, permettant de lire et contrôler des signaux cérébraux. Les électrodes peuvent être placées de façon très précises à l'aide de bras robotisés. Les implants cérébraux sont généralement alimentés par une batterie, qui peut être rechargée à distance par induction électromagnétique. Des systèmes interfacés avec des ordinateurs sont également appelés « interface neuronale directe » ou en anglais « Brain-Computer Interface ». L'implantation ne laisse pas forcément de cicatrices visibles car ils peuvent être implantés en passant par la gorge, le nez, ou du côté de l'œil, en utilisant des

Development of a peripheral neuroprosthesis to elicit fine hand movements in primates

Marion Aimée Geneviève Badi-Dubois

Neurological disorders such as spinal cord lesion and stroke disturb sensorimotor pathways and result in severe motor deficits that can dreadfully impact the quality of life of affected individuals. Strategies aiming at restoring hand function after upper-limb paralysis have relied on functional electrical stimulation (FES) of the remnant muscles but have shown limited functional benefits. Other electrical stimulation approaches have employed peripheral nerve electrodes to engage multiple distal muscles from a single proximal location on the nerve. Interestingly, intrafascicular interfaces penetrating the nerve have shown impressive capabilities for the restoration of specific somatosensory percepts in humans following limb amputation. In this thesis, I developed a peripheral neuroprosthesis based on intrafascicular implants for the restoration of fine hand movements in the pre-clinical primate model. The first part of this work focuses on building an adaptable technological platform for the design of arm and hand sensorimotor tasks in monkeys. I demonstrated the significance of our platform for fundamental and translational studies by studying a rich collection of physiological signals during unconstrained reaching and grasping tasks. In the second part of this work, I mobilized this experimental framework for the functional evaluation of intrafascicular stimulation paradigms in monkeys. More specifically, I used complementary neuroanatomical and computational analyses to tailor an intrafascicular interface to the arm motor nerves of monkeys. I then demonstrated that intrafascicular stimulation promotes selective hand muscle activation, evokes a large diversity of functional hand movements, and generates sustained and controllable levels of force. Finally, I coupled intrafascicular stimulation to motor cortical signals to enable a temporarily paralyzed monkey to perform a functional grasping task. The last part of this thesis is dedicated to a pilot study investigating the use of low-dimensional cortical dynamics for the direct control of a neuroprosthetic device. It provides exploratory routes for the design of an adaptive brain-control paradigm to drive peripheral implants using robust, voluntary commands. This thesis brings evidence of the high selectivity and functionality of intraneural implants for motor applications and demonstrates their potential for eliciting precise hand movements. These findings open promising perspectives for the development of peripheral intraneural neuroprosthetic systems to restore functional and dexterous hand control after paralysis.

EPFL2021

Interactions of Peripheral Nerve Cells with Soft Micropillar Interfaces

Cédric Xavier Paulou

Regenerative peripheral nerve interfaces are particularly invasive implant devices. They rely on the growth of axons from transected nerve stumps through electrode-bearing structures in order to establish highly efficient communication channels with the nervous system. The functioning and lifetime of these devices is challenged by the host response from the nervous tissue, which complicates their use in long-term applications. The elicited foreign body reaction is characterized by chronic inflammation and excessive deposition of collagen. The difference in mechanical properties between implant material and nervous tissue, which are among the softest in the body, is thought to be an important contributor to these adverse effects. With the motivation of improving the bio-integration of implanted nerve interfaces, this thesis evaluate micropillar arrays as a texture to modulate the behaviour of peripheral nerve cells on polydimethylsiloxane (PDMS) silicone surfaces. Thanks to their flexibility, micropillars can reduce the effective stiffness perceived by cells, while providing topographic cues at the microscopic level. Importantly, these mechanical and topographic cues can be easily tailored by modification of their geometrical dimensions. Micropillar arrays with varying dimensions were fabricated by soft lithography process. Their mechanical behaviour under bending was simulated by finite element method (FEM) analysis to approximate their spring constants. The softest micropillar configuration had an effective surface stiffness of 0.7 kPa, representing a reduction of 3 orders of magnitude compared to bulk PDMS. The effect of micropillar diameter (from 1.2 to 4.2 um), modulating the surface stiffness, and the effect of interpillar spacing (from 1.0 to 7.6 um), modifying micropillars density, were investigated in vitro. Dissociated cell cultures from dorsal root ganglion (DRG), comprising neurons and glial cells, were probed for different parameters after 7 days of incubation. While neurons spread and established neurite networks on all configurations, fewer glial cells were found on the softest and lowest density of micropillars. Neurites explored this pseudo-3D environment and preferentially anchored to the top of pillar shafts, well above the bottom surface. The matrix of pillars provided physical cues for growing neurites, which strongly aligned with the densest arrays. The morphology of neuronal bodies was affected by the most flexible pillars, which induced smaller and rounder somas. Macrophages and adipose-derived stem cells showed significantly increased attachment to micropillar topographies compared to flat surface. The effect of micropillars was investigated on nerve regeneration in vivo. Nerve guidance conduits patterned with micropillars on their lumen surface were sutured to the transected sciatic nerve of rats. While the axonal regeneration was slightly affected by the micropillar texture, macrophages were massively recruited to the micropillar surface in comparison to the flat control. Their phenotype was also affected, with predominantly classically activated macrophage in contact with the micropillars. Altogether, these results demonstrate the possibility to modulate the behaviour of a variety of cell types in vitro and in vivo, through the patterning of micropillars on the surface. This approach is amenable to implementation on silicone neural implants for mitigating tissue-material interactions.

EPFL2015

Closed-Loop Neural Interfaces with Embedded Machine Learning

Uisub Shin, Mahsa Shoaran, Bingzhao Zhu

Neural interfaces capable of multi-site electrical recording, on-site signal classification, and closed-loop therapy are critical for the diagnosis and treatment of neurological disorders. However, deploying machine learning algorithms on low-power neural devices is challenging, given the tight constraints on computational and memory resources for such devices. In this paper, we review the recent developments in embedding machine learning in neural interfaces, with a focus on design trade-offs and hardware efficiency. We also present our optimized tree-based model for low-power and memory-efficient classification of neural signal in brain implants. Using energy-aware learning and model compression, we show that the proposed oblique trees can outperform conventional machine learning models in applications such as seizure or tremor detection and motor decoding.

IEEE2020