Framework for the Development of Neuroprostheses: From Basic Understanding by Sciatic and Median Nerves Models to Bionic Legs and Hands

Neuroprostheses based on electrical stimulation are becoming a therapeutic reality, dramatically improving the life of disabled people. They are based on neural interfaces that are designed to create an intimate contact with neural cells. These devices speak the language of electron currents, while the human nervous system uses ionic currents to communicate. A deep understanding of the complex interplay between these currents, during the electrical stimulation, is essential for the development of optimized neuroprostheses. Neural electrodes can have different geometries, placement within the nervous system, and the stimulation protocols (paradigms of use). This high-dimensional problem is not tractable by an empiric, brute-force approach and should be tackled by exact computational models, making use of our accumulated knowledge. In pursuit of this goal, a hybrid finite element method-NEURON modeling-is used for a solution of electrical field generated by stimulation, within the different neural structures having anisotropic conductivity, and a corresponding neural response computation. In this work, an important correction of perineurium electrical conductivity is computed. Models of median and sciatic nerves, innervating the hand and foot areas, relevant to the development of bionic hands and legs with sensory feedback, are implemented. The obtained results have the potential to optimize the design of neural interfaces, in terms of shape and number of stimulating contacts. Guidelines for the neurosurgical planning are proposed, by indicating the optimal number of implants for a specific nerve to obtain the best efficacy with the lowest invasiveness. The interpretation is proposed for one of the basic problems of neural interfaces, consisting in the change of the stimulation threshold due to fibrotic reaction of tissue. We show that it is possible to use human microstimulation as an experimental setup for testing of afferent stimulation paradigms, which can be translated to further chronic implants. In the future, models will have a key role in the decision of the most appropriate design of customized neuroprostheses, their optimal modality of use, understanding the effects that occur during their use, and minimizing animal and human experimentation.

Agile in-litero experiments

Renaud Luc Richardet

In neuroscience, as in many other scientific domains, the primary form of knowledge dissemination is through published articles in peer-reviewed journals. One challenge for modern neuroinformatics is to design methods to make the knowledge from the tremendous backlog of publications accessible for search, analysis and its integration into computational models. In this thesis, we introduce novel natural language processing (NLP) models and systems to mine the neuroscientific literature. In addition to in vivo, in vitro or in silico experiments, we coin the NLP methods developed in this thesis as in litero experiments, aiming at analyzing and making accessible the extended body of neuroscientific literature. In particular, we focus on two important neuroscientific entities: brain regions and neural cells. An integrated NLP model is designed to automatically extract brain region connectivity statements from very large corpora. This system is applied to a large corpus of 25M PubMed abstracts and 600K full-text articles. Central to this system is the creation of a searchable database of brain region connectivity statements, allowing neuroscientists to gain an overview of all brain regions connected to a given region of interest. More importantly, the database enables researcher to provide feedback on connectivity results and links back to the original article sentence to provide the relevant context. The database is evaluated by neuroanatomists on real connectomics tasks (targets of Nucleus Accumbens) and results in significant effort reduction in comparison to previous manual methods (from 1 week to 2h). Subsequently, we introduce neuroNER to identify, normalize and compare instances of identify neuronsneurons in the scientific literature. Our method relies on identifying and analyzing each of the domain features used to annotate a specific neuron mention, like the morphological term 'basket' or brain region 'hippocampus'. We apply our method to the same corpus of 25M PubMed abstracts and 600K full-text articles and find over 500K unique neuron type mentions. To demonstrate the utility of our approach, we also apply our method towards cross-comparing the NeuroLex and Human Brain Project (HBP) cell type ontologies. By decoupling a neuron mention's identity into its specific compositional features, our method can successfully identify specific neuron types even if they are not explicitly listed within a predefined neuron type lexicon, thus greatly facilitating cross-laboratory studies. In order to build such large databases, several tools and infrastructureslarge-scale NLP were developed: a robust pipeline to preprocess full-text PDF articles, as well as bluima, an NLP processing pipeline specialized on neuroscience to perform text-mining at PubMed scale. During the development of those two NLP systems, we acknowledged the need for novel NLP approaches to rapidly develop custom text mining solutions. This led to the formalization of the agile text miningagile text-mining methodology to improve the communication and collaboration between subject matter experts and text miners. Agile text mining is characterized by short development cycles, frequent tasks redefinition and continuous performance monitoring through integration tests. To support our approach, we developed Sherlok, an NLP framework designed for the development of agile text mining applications.

EPFL2016

Electronic dura mater soft, multimodal neural interfaces

Arthur Edouard Hirsch

Neuroprosthetic devices are engineered to study, support or replace impaired functions of the nervous system. The neural interface is an essential element of neuroprosthetic systems as it allows for transduction of signals and stimuli of desired functions (recording, stimulation, neuromodulation). A persistent challenge for translating neuroprosthetics from the laboratory to the clinic is the lack of long-term biointegration of neural interfaces. This thesis aims at improving biointegration of neural interfaces by reducing the mechanical mismatch between implant and neural tissue. In this thesis, the design, fabrication and characterization of soft surface neural interfaces is described. These soft neural interfaces, termed electronic dura mater or e-dura, were designed to mimic the mechanical properties of dura mater. In contrast with conventional neural technologies, e-dura neural interfaces were made of soft and compliant materials. They conform to the circumvolutions of the brain and spinal cord and follow their dynamic deformation without damaging the surrounding neural tissues. These soft multimodal neural interfaces were fabricated on silicone substrates using techniques imported from the microfabrication industry and incorporate compliant electrodes, stretchable electrical interconnects and a micro-catheter for drug delivery. Evaluation of the e-dura biointegration with spinal tissues demonstrated reduced foreign body reaction, compared to stiff polyimide based implants. Additionally, mechanical tests on an in-vitro spinal surrogate provided insights on the complex biomechanical coupling between implants and neural tissue. E-dura interfaces, implanted in rodents, maintained their functionality over extended periods and provided high-resolution neuronal recordings and concurrent delivery of electrical and chemical neuromodulation. Eventually, the use of gallium thin films was explored to create highly conductive and stretchable interconnects for integration of active electronic components in e-dura neural interfaces.

EPFL2018

Neuroprosthetics to support impaired hand functions: A multidisciplinary approach combining brain-machine interfaces and wearable exoskeletons

Luca Randazzo

Neuroprosthetics, the discipline that aims at interfacing neural systems to artificially engineered devices, has witnessed in recent years important advancements towards the ultimate goal of augmenting and restoring human functions through technology and artificial systems. Research in this field covers a wide range of efforts, from advancing our basic understanding of the neural processes involved into human sensory, motor and cognitive functions, to developing artificial intelligence algorithms for their decoding, to the physical interfacing between artificial devices and the human nervous system and biomechanics. Ultimately, this discipline aims at fusing these systems in order to restore, replace and rehabilitate lost and impaired functions due to motor disabling conditions or traumatic accidents. In this thesis, I present a multidisciplinary approach that aimed at developing a neuroprosthesis for the assistance and restoration of hand functions impaired by neurological disorders or traumatic accidents, such as cerebrovascular accidents and spinal cord injuries. These efforts encompassed the conceptualization and development of a robotic hand exoskeleton and brain-machine interface (BMI) algorithms for its control. Specifically, this thesis focused on the design, development and testing of (i) a novel mechatronic hand exoskeleton to assist hand motions within domestic and clinical settings, (ii) non-invasive BMI approaches based on electroencephalography (EEG) to decode neural correlates of intended hand movements, and on (iii) the closed-loop integration of the proposed exoskeleton and BMI for the sake of providing continuous feedback about ongoing neural modulations through hand motions and to trigger sensorimotor rehabilitation within clinical scenarios. Results showed that the proposed mechatronic system can successfully control hand opening and closing within a fully wearable, portable and lightweight package. The system was tested with users who suffered from motor disabling impairments, showing that it could help them in performing several activities of daily living for the first time since their accidents. From a brain-machine interfacing perspective, this work shows how imagined hand movements can be decoded, through EEG, in parallel with exoskeleton-induced motions, with important implications for the development of more embodied human-machine interaction protocols. Finally, this work shows how the closed-loop control of exoskeleton motions by means of the decoded ongoing neural activity enhances the discriminability of sensorimotor neural patterns and improves the performance of the brain-machine control channel. Overall, the results presented here represent important advancements within the field of neuroprosthetics, with interesting implications for the development of assistive exoskeletal technologies and non-invasive brain-machine interfaces aimed at controlling such systems in clinical or domestic settings, for both assistive and neurorehabilitative purposes.

EPFL2018

Neuroprosthetics to support impaired hand functions: A multidisciplinary approach combining brain-machine interfaces and wearable exoskeletons

Luca Randazzo

EPFL2018

Electronic dura mater soft, multimodal neural interfaces

Arthur Edouard Hirsch

EPFL2018

Agile in-litero experiments

Renaud Luc Richardet

EPFL2016