Rozprawy doktorskie na temat „Neural interfaces”

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2018.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 161-174).
Neurological disorders affect up to a billion people worldwide, and their socioeconomic burden is projected to increase as the population ages. However, our ability to understand and to treat neural disorders is currently limited by the lack of tools capable of interfacing with the brain over extended periods of time. This is hypothesized to stem from the mismatch in mechanical and chemical properties between the neural probes and the neural tissues, which leads to foreign body response and functional device failure due to tissue scarring in the probe vicinity. To address the challenge, I developed fiber-based bioelectronic devices integrating diverse modalities within a single platform using thermal drawing process (TDP). All-polymer or hydrogel integrated probes with optical, electrical, and fluidic capabilities were developed all within the 100-200 [mu]m diameter, which allowed one-step surgery to the mouse brain and spinal cord for optogenetic experiments. This probe also addressed the challenge of biocompatibility and enabled the recording isolated action potentials for 3 months. In addition, I applied TPD to produce biocompatible polymer-based neural scaffold with various geometries (round, rectangular, micro-grooved) and dimensions between 50-200 [mu]m. This allowed for investigation of the enhancement of neurite growth as a function of fiber parameters. We found that the topographical features and the narrow channels generally led to enhanced growth. This thesis illustrated a variety of applications of multifunctional fiber-based devices in neuroscience and neural engineering, which anticipated to enable basic studies of the nervous system and future treatment of neurological disorders.
by Seongjun Park.
Ph. D.

El funcionament del cervell es basa en processos complexos, que encara no s’han descrit i comprès detalladament. En les últimes dècades, la neurociència ha experimentat un desenvolupament accelerat, impulsat per noves neurotecnologías que permeten monitoritzar les dinàmiques de l’activitat elèctrica al cervell amb una major resolució espai-temporal i una àrea de cobertura més àmplia. No obstant això, a causa de l’alta complexitat de les xarxes neuronals al cervell, que són compostes per poblacions neuronals fortament interconnectades en àmplies regions cerebrals, estem lluny de detectar una fracció significativa de les neurones que donen lloc a funcions complexes. Per tal d’investigar les dinàmiques neuronals a gran escala amb alta resolució espacial, s’han utilitzat diverses tecnologies, que inclouen la ressonància magnética funcional (fMRI), imatges amb marcadors sensibles al voltatge o registres electrofisiològics d’alt recompte de sensors. No obstant això, la resolució temporal del fMRI i els mètodes òptics es limita típicament a uns pocs hertzs, gairebé tres ordres de magnitud per sota de la dels potencials d’acció, i es limiten a les condicions en què el subjecte es troba immòbil. D’altra banda, els registres electrofisiològics basats en matrius de microelèctrodes proporcionen una alta resolució espai-temporal, el que permet detectar amb precisió dinàmiques ràpides de centenars de neurones individuals simultàniament en animals que es mouen lliurement. No obstant això, les interfícies de detecció neuroelectrónica presenten una limitació en el producte entre la resolució espacial i l’àrea de cobertura. A més, presenten una baixa sensibilitat a la banda de freqüència infra-lenta (<0.5Hz), que està relacionada amb la connectivitat funcional de llarg abast. En aquesta tesi es presenta una nova tecnologia basada en sensors actius de grafè, que permet incrementar l’àrea de cobertura i la resolució espacial dels registres electrofisiològics conservant una alta sensibilitat en una banda de freqüència àmplia, des de l’activitat infra-lenta fins a la de una sola cèl·lula electrogénica. Aquest desenvolupament tecnològic es divideix en tres etapes principals; en primer lloc, s’obté una comprensió més profunda de les característiques intrínseques del soroll i la resposta en freqüència d’aquests sensors basant-se en l’estat de l’art en tecnologia de sensors de grafè. En la segona etapa, es mostra un sistema quasi-comercial basat en matrius de sensors de grafè epi-cortical i transmissió sense fil per a la implantació crònica en rates. Amb aquest sistema, es demostra la reproductibilitat de les matrius de sensors de grafè, la seva estabilitat a llarg termini i la seva biocompatibilitat crònica. A més, es proporciona evidència preliminar per a una àmplia gamma de nous patrons electrofisiològics gràcies a la seva sensibilitat en la banda de freqüència infra-lenta. Finalment, en l’última etapa d’aquesta tesi, l’enfocament se centra en el desenvolupament de noves estratègies de multiplexació per augmentar el nombre de sensors a les sondes neuronals. Aquestes tres etapes principals de desenvolupament han portat a la demostració del potencial de les matrius de sensors de grafè multiplexats per al mapejat de les dinàmiques neuronals a gran escala en una banda de freqüència àmplia, en animals que es mouen lliurement, durant llargs períodes. La combinació d’aquestes capacitats fa que les matrius de sensors actius de grafè siguin una tecnologia prometedora per a interfícies cervell-ordinador d’alt ample de banda i una eina única per investigar el paper de l’activitat infra-lenta en la coordinació de les dinàmiques neuronals d’alta freqüència.
El funcionamiento del cerebro se basa en procesos complejos, que aún no se han descrito y comprendido detalladamente. En las últimas décadas, la neurociencia ha experimentado un desarrollo acelerado, impulsado por nuevas neurotecnologías que permiten monitorear la dinámica de la actividad eléctrica en el cerebro con una mayor resolución espacio-temporal y un área de cobertura más amplia. Sin embargo, debido a la alta complejidad de las redes neuronales en el cerebro, que están compuestas por poblaciones neuronales fuertemente interconectadas en amplias regiones cerebrales, estamos lejos de monitorear una fracción significativa de neuronas que dan lugar a funciones complejas. Con el fin de investigar las dinámica neuronales a gran escala con alta resolución espacial, se han utilizado diversas tecnologías, que incluyen la resonancia magnética funcional (fMRI), imágenes con marcadores sensibles al voltaje o registros electrofisiológicos de alto conteo de sensores. Sin embargo, la resolución temporal del fMRI y los métodos ópticos se limita típicamente a unos pocos hercios, casi tres órdenes de magnitud por debajo de la de los potenciales de acción, y se limitan a condiciones en los que el sujeto se encuentra inmóvil. Por otro lado, los registros electrofisiológicos basados en matrices de microelectrodos proporcionan una alta resolución espacio-temporal, lo que permite detectar con precisión dinámicas rápidas de cientos de neuronas individuales simultáneamente en animales que se mueven libremente. Sin embargo, las interfaces de detección neuroelectrónica presentan una limitación en el producto entre la resolución espacial y el área de cobertura. Además, presentan una baja sensibilidad en la banda de frecuencia infra-lenta (<0.5Hz), que está relacionada con la conectividad funcional de largo alcance. En esta tesis se presenta una nueva tecnología basada en sensores activos de grafeno, que permite incrementar el área de cobertura y la resolución espacial de los registros electrofisiológicos conservando una alta sensibilidad en una amplia banda de frecuencia, desde la actividad infra-lenta hasta la de una sola célula electrogénica. Este desarrollo tecnológico se divide en tres etapas principales; en primer lugar, se obtiene una comprensión más profunda de las características intrínsecas del ruido y la respuesta en frecuencia de estos sensores basándose en el estado del arte en tecnología de sensores de grafeno. En la segunda etapa, se muestra un sistema cuasi-comercial basado en matrices de sensores de grafeno epi-cortical y transmisión inalámbrica para implantación crónica en ratas. Con este sistema, se demuestra la reproducibilidad de las matrices de sensores de grafeno, su estabilidad a largo plazo y su biocompatibilidad crónica. Además, se proporciona evidencia preliminar para una amplia gama de nuevos patrones electrofisiológicos debido a su sensibilidad en la banda de frecuencia infra-lenta. Finalmente, en la última etapa de esta tesis, el enfoque se centra en el desarrollo de nuevas estrategias de multiplexación para aumentar el número de sensores en las sondas neuronales. Estas tres etapas principales de desarrollo han llevado a la demostración del potencial de las matrices de sensores de grafeno multiplexados para el mapeado de las dinámicas neuronales a gran escala en una amplia banda de frecuencia en animales que se mueven libremente durante largos períodos. La combinación de estas capacidades hace que las matrices de sensores activos de grafeno sean una tecnología prometedora para interfaces cerebro-ordenador de alto ancho de banda y una herramienta única para investigar el papel de la actividad infra-lenta en la coordinación de las dinámicas neuronales de alta frecuencia.
Brain function is based on highly complex processes, which remain yet to be described and understood in detail. In the last decades, neuroscience has experienced an accelerated development, prompted by novel neurotechnologies that allow monitoring the dynamics of electrical activity in the brain with a higher spatio-temporal resolution and wider coverage area. However, due to the high complexity of neural networks in the brain, which are composed of strongly interconnected neural populations across large brain regions, we are far from monitoring a significant fraction of neurons mediating complex functions. In order to investigate large-scale brain dynamics with high spatial resolution several technologies have been extensively used, including functional magnetic resonance imaging (fMRI), voltage-sensitive dye imaging or high sensor-count electrophysiological recordings. However, the temporal resolution of fMRI and optical methods is typically limited to few hertz, almost three orders of magnitude below that of action potentials, and are limited to head-fixed conditions. On the other hand, electrophysiological recordings based on micro-electrode arrays provide a high spatio-temporal resolution, allowing to accurately detect fast dynamics from hundreds of individual neurons simultaneously in freely moving animals. However, neuroelectronic sensing interfaces present a trade-off between spatial resolution and coverage area. Moreover, they present a poor sensitivity in the infra-slow frequency band ($<0.5$\,$Hz$), which is related to long-range functional connectivity. In this thesis, a novel technology based on graphene active sensors is presented, which allows to increase the coverage area and spatial resolution of electrophysiological recordings while preserving a high sensitivity in a wide frequency band, from infra-slow to single electrogenic cell activity. This technological development is divided into three main stages; first, a deeper understanding of the intrinsic noise characteristics and frequency response of these sensors is obtained by building on prior graphene sensor technology. In the second stage, a quasi-commercial system based on epi-cortical graphene sensor arrays and a wireless headstage for chronic implantation in rats is shown. Using this system, the reproducibility of the graphene sensor arrays, their long-term stability and their chronic biocompatibility are demonstrated. Furthermore, preliminary evidence is provided for a wide range of novel electrophysiological patterns owing to their sensitivity in the infra-slow frequency band. Finally, in the last stage of this thesis, the focus is centred on the development of new multiplexing strategies to upscale the number of sensors on the neural probes. These three main development stages have led to the demonstration of the potential of multiplexed graphene sensor arrays for mapping of large-scale brain dynamics in a wide frequency band in freely moving animals over long periods. The combination of these capabilities makes graphene active sensor arrays a promising technology for high bandwidth brain computer interfaces and a unique tool to investigate the role of infra-slow activity on the coordination of higher frequency brain dynamics.
Universitat Autònoma de Barcelona. Programa de Doctorat en Enginyeria Electrònica i de Telecomunicació

The thesis describes novel processing routes that have been developed to fabricate neural interfaces. A process has been investigated that uses microfabrication techniques to fabricate a multi-channel regenerative implant that can record nerve impulses in the peripheral nervous system (PNS), called the Spiral Peripheral Nerve Interface (SPNI). It is shown both theoretically and experimentally that the implant improves the ability to record signals in the PNS via micro-channels that act as axonal amplifiers. New processing routes are introduced to create robust interconnections from the SPNI to external electronics via ‘Microflex’ technology. To incorporate the new interconnection technology the SPNI had to be modified. During this modification the strain in the device was given specific consideration, for which a new bending model is presented. Modelling is used to show that electrochemical impedance spectroscopy can be used to assess the quality of the fabrication process. Electrochemical and mechanical tests show that the interconnection technology is suitable for a neural interfaces but the fabrication of perfectly sealed micro-channels was not evident. Thus, the SPNI was further improved by the introduction of a silicone sealing layer in the construction of the micro-channel array that was implemented using a novel adhesive bonding technique.

From investigating movement in an insect to deciphering cognition in a human brain to treating Parkinson's disease, hearing loss, or even blindness, electronic implants are an essential tool for understanding the brain and treating neural diseases. Currently, the stimulating and recording resolution of these implants remains low. For instance, they can record all the neuron activity associated with movement in an insect, but are quite far from recording, at an individual neuron resolution, the large volumes of brain tissue associated with cognition. Likewise, there is remarkable success in the cochlear implant restoring hearing due to the relatively simple anatomy of the auditory nerves, but are failing to restore vision to the blind due to poor signal fidelity and transmission in stimulating the more complex anatomy of the visual nerves. The critically important research needed to improve the resolution of these implants is to optimize the neuron-electrode interface. This thesis explores geometrical and material modifications to both stimulating and recording electrodes which can improve the neuron-electrode interface. First, we introduce a fractal electrode geometry which radically improves the restored visual acuity achieved by retinal implants and leads to safe, long-term operation of the implant. Next, we demonstrate excellent neuron survival and neurite outgrowth on carbon nanotube electrodes, thus providing a safe biomaterial which forms a strong connection between the electrode and neurons. Additional preliminary evidence suggests carbon nanotubes patterned into a fractal geometry will provide further benefits in improving the electrode-neuron interface. Finally, we propose a novel implant based off field effect transistor technology which utilizes an interconnecting fractal network of semiconducting carbon nanotubes to record from thousands of neurons simutaneously at an individual neuron resolution. Taken together, these improvements have the potential to radically improve our understanding of the brain and our ability to restore function to patients of neural diseases.

Thesis: S.B., Massachusetts Institute of Technology, Department of Materials Science and Engineering, 2015.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 48-50).
The understanding of the brain would be revolutionized by a tool that can measure intra- and extra-cellular electrical potentials on a parallelized scale, without disrupting the neural physiology. Existing technologies do not sufficiently carry out these functions. Using a thermal drawing process (TDP), multimaterial fibers comprised of polymer-metal composites can be fabricated to create flexible, microelectrode arrays. These fibers can be further processed after the TDP, using selective etching to reduce the diameter of the probe. These devices have been implanted and have been used to record neural activity in vivo while evoking minimal tissue response. Additionally, electrodeposition of biocompatible metals onto the fiber-electrode tips can be implemented to increase the signal-to-noise ratio (SNR). Here, I describe the electroplating of gold onto the fiber-tips of tin and tin-indium electrodes, which were drawn using TDP. By adjusting the electrodeposition conditions, the electrode tip geometries can be tuned to optimize the interface between the device tips and neuronal membranes.
by Christina M. Tringides.
S.B.

From investigating movement in an insect to deciphering cognition in a human brain to treating Parkinson's disease, hearing loss, or even blindness, electronic implants are an essential tool for understanding the brain and treating neural diseases. Currently, the stimulating and recording resolution of these implants remains low. For instance, they can record all the neuron activity associated with movement in an insect, but are quite far from recording, at an individual neuron resolution, the large volumes of brain tissue associated with cognition. Likewise, there is remarkable success in the cochlear implant restoring hearing due to the relatively simple anatomy of the auditory nerves, but are failing to restore vision to the blind due to poor signal fidelity and transmission in stimulating the more complex anatomy of the visual nerves. The critically important research needed to improve the resolution of these implants is to optimize the neuron-electrode interface. This thesis explores geometrical and material modifications to both stimulating and recording electrodes which can improve the neuron-electrode interface. First, we introduce a fractal electrode geometry which radically improves the restored visual acuity achieved by retinal implants and leads to safe, long-term operation of the implant. Next, we demonstrate excellent neuron survival and neurite outgrowth on carbon nanotube electrodes, thus providing a safe biomaterial which forms a strong connection between the electrode and neurons. Additional preliminary evidence suggests carbon nanotubes patterned into a fractal geometry will provide further benefits in improving the electrode-neuron interface. Finally, we propose a novel implant based off field effect transistor technology which utilizes an interconnecting fractal network of semiconducting carbon nanotubes to record from thousands of neurons simutaneously at an individual neuron resolution. Taken together, these improvements have the potential to radically improve our understanding of the brain and our ability to restore function to patients of neural diseases.

En els últims anys s’han produït desenvolupaments tecnològics innovadors en el camp dels implants neuronals per a aplicacions mèdiques. La comprensió de el cervell humà es considera com un dels majors reptes científics del nostre temps; com a conseqüència, estem sent testimonis d’una intensificació de la investigació en el desenvolupament de les interfícies cervell-màquina (IMC) per llegir i estimular l’activitat cerebral. No obstant això, els implants neuronals actualment disponibles ofereixen una eficàcia clínica modesta, en part a causa de les limitacions que plantegen la invasivitat dels materials. Aquests materials comprometen la resolució de la interfície, el rendiment i l’estabilitat a llarg termini dels implants. El desenvolupament d’una electrònica flexible que utilitzi materials biocompatibles és clau per al desenvolupament d’implants neuronals mínimament invasius, que puguin implantar-se de forma crònica. Un camp d’investigació molt prometedor, és l’ús de materials bidimensionals, com el grafè, per a aplicacions bioelectròniques. El transistor d’efecte de camp en solució de grafè (gSGFET) és una de d’aquestes noves tecnologies neuronals emergents. Aquests dispositius poden superar les limitacions esmentades anteriorment gràcies a les extraordinàries propietats del grafè, com ara la seva alta flexibilitat mecànica, estabilitat electroquímica, biocompatibilitat i alta sensibilitat. En aquesta tesi doctoral, s’han fabricat matrius de gSGFET i s’han optimitzat iterativament en termes de sensibilitat i relació senyal / soroll, adoptant mètodes de microfabricació a escala d’oblia. S’ha caracteritzat el soroll 1 / f en els gSGFETs i s’ha optimitzat amb un tractament UVO de la interfície metall / grafè i desacoblant el grafè del substrat utilitzant diferents nanomaterials com ara l’encapsulació del grafè amb nitrur de bor hexagonal (hBN), monocapes autoacoblades i grafè bicapa. A més, s’han fabricat amb èxit sondes neuronals epicorticals i intracorticals flexibles, que contenien matrius de gSGFET, i s’han fet enregistraments de microelectrocorticografia in vivo en rosegadors. S’han inserit dispositius intracorticals flexibles en el cervell utilitzant un protocol de reforç de la capa posterior del dispositiu amb proteïna de fibroïna de seda biorresistent. Els resultats presentats en aquesta tesi demostren la superior resolució espai-temporal dels gSGFET en comparació amb la tecnologia estàndard de microelèctrodes; en particular, la capacitat de mapejar amb alta fidelitat, l’activitat de molt baixa freqüència (ISA, <0,1 Hz) juntament amb els senyals en el típic ample de banda dels LFP. Avui dia se sap que l’activitat cerebral de molt baixa freqüència, contribueix a la fisiopatologia de diversos trastorns neurològics com el vessament cerebral, la lesió cerebral traumàtica, la migranya i l’epilèpsia. No obstant això, aquesta activitat rares vegades es registra a causa de les limitacions tècniques intrínseques dels elèctrodes convencionals acoblats a la CA. S’han obtingut mesures neuronals amb sondes de profunditat flexibles i multicanal de grafè (gDNP) en models animals desperts amb convulsions i epilèpsia. S’ha detectat i cartografiat l’AIS a través de diferents capes corticals i regions subcorticals, registrant simultàniament l’activitat epilèptica en bandes de freqüència més convencionals (1-600Hz). A més, com a part d’aquesta tesi s’ha demostrat també l’estabilitat i funcionalitat de registres a llarg termini, així com la biocompatibilitat dels gDNPs. La tecnologia bioelectrònica basada en grafè aquí descrita té el potencial d’esdevenir una eina de referència per a l’electrofisiologia d’ample de banda complet. Es preveu que aquesta tecnologia tingui un gran impacte en una comunitat àmplia i multidisciplinària que inclogui investigadors en neurotecnologia, enginyers biomèdics, neurocientífics que estudien la dinàmica cortical de banda ampla associada amb el comportament espontani i /o els estats cerebrals, així com investigadors clínics interessats en el paper de l’activitat de molt baixa freqüència en epilèpsia, els accidents cerebrovasculars i la migranya.
En los últimos años se han producido nuevos desarrollos tecnológicos en el campo de los implantes neuronales para aplicaciones médicas. La comprensión del cerebro humano se considera uno de los mayores desafíos científicos de nuestro tiempo; como consecuencia, estamos siendo testigos de una intensificación de la investigación en el desarrollo de las interfaces cerebro-máquina (IMC) para leer y estimular la actividad cerebral. No obstante, los implantes neuronales actualmente disponibles ofrecen una eficacia clínica modesta, en parte debido a las limitaciones que plantea la invasividad de los materiales. Esos materiales comprometen la resolución de la interfaz, el rendimiento y la estabilidad a largo plazo de los implantes neurales. El desarrollo de una electrónica flexible que utilice materiales biocompatibles es clave para la realización de implantes neuronales mínimamente invasivos que puedan implantarse de forma crónica. Un campo de investigación muy prometedor es el uso de materiales bidimensionales, como el grafeno, para aplicaciones bioelectrónicas. El transistor de efecto de campo en solución de grafeno (gSGFET) es una de dichas nuevas tecnologías neurales emergentes. Estos dispositivos pueden superar las limitaciones mencionadas anteriormente gracias a las extraordinarias propiedades del grafeno, como su alta flexibilidad mecánica, estabilidad electroquímica, biocompatibilidad y sensibilidad. En esta tesis doctoral, se han fabricado matrices de gSGFET y se han optimizado iterativamente en términos de sensibilidad y relación señal/ruido, adoptando métodos de microfabricación a escala de oblea. Se ha caracterizado el ruido 1/f en los gSGFETs y optimizado haciendo un tratamiento UVO en la interfaz metal/grafeno y desacoplando el canal de grafeno del sustrato utilizando diferentes nanomateriales como la encapsulación con nitruro de boro hexagonal (hBN), monocapas autoensambladas y bicapas de grafeno. Además, se han fabricado con éxito sondas neurales epicorticales e intracorticales flexibles con matrices de gSGFET y se han utilizado durante las medidas de microelectrocorticografía in vivo en roedores. Se han insertado dispositivos intracorticales flexibles en el cerebro utilizando un protocolo de refuerzo de la capa posterior de los dispositivos con proteína de fibroína de seda biorresistente. Los resultados presentados en esta tesis demuestran la superior resolución espacio-temporal de los gSGFET en comparación con la tecnología estándar de microelectrodos; en particular, referente a la capacidad de mapear con alta fidelidad, la actividad de muy baja frecuencia (ISA, < 0,1 Hz) junto con las señales en el típico ancho de banda LFP. Hoy en día se sabe que la actividad cerebral de muy baja frecuencia, contribuye a la fisiopatología de varios trastornos neurológicos como el derrame cerebral, la lesión cerebral traumática, la migraña y la epilepsia. Sin embargo, esta actividad rara vez se registra debido a las limitaciones técnicas intrínsecas de los electrodos convencionales acoplados a la CA. Se han obtenido registros con sondas neuronales de profundidad de grafeno (gDNP) en modelos animales de epilepsia. Se detectó ISA a través de diferentes capas corticales y regiones subcorticales, registrando simultáneamente la actividad epiléptica en bandas de frecuencia más convencionales (1-600Hz). Además, se ha demostrado también la evaluación de la estabilidad y funcionalidad en registros crónicos, así como la biocompatibilidad del gDNP. La tecnología bioelectrónica basada en el grafeno aquí descrita tiene el potencial de convertirse en una herramienta de referencia para la electrofisiología de ancho de banda completo. Se prevé que esta tecnología tenga un gran impacto en una comunidad amplia y multidisciplinaria que incluya investigadores en neurotecnología, ingenieros biomédicos, neurocientíficos que estudien la dinámica cortical de banda ancha asociada con el comportamiento espontáneo y/o los estados cerebrales, así como investigadores clínicos interesados en la actividad de baja frecuencia en la epilepsia, los accidentes cerebrovasculares y la migraña.
Recent years have witnessed novel technology developments of neural implants for medical applications which are expected to pave the way to unveil functionalities of the central nervous system. Understanding the human brain is commonly considered one of the biggest scientific challenges of our time; as a consequence, we are witnessing an intensified research in the development of brain-machine-interfaces (BMIs), which would allow us to both read and stimulate brain activity. Nevertheless, currently available neural implants offer a modest clinical efficacy, partly due to the limitations posed by the invasiveness of the implants materials and technology and by the metals used at the electrical interface with the tissue. Such materials compromise the interfacing resolution, the performance and the long term stability of neural implants. Development of flexible electronics using biocompatible materials is key for the realisation of minimally invasive neural implants, which can be chronically implanted without causing rejection from the immune system. A relatively young yet very promising research field, that is increasingly drawing attention is the use of two dimensional materials, such as graphene, for bioelectronic applications. Graphene solution-gated field effect transistor (gSGFET) is one of several emerging new neural technologies. These devices can overcome the above-mentioned limitations thanks to the outstanding properties of graphene, such as mechanical flexibility, electrochemical inertness, biocompatibility and high sensitivity. In this PhD thesis, arrays of gSGFETs have been fabricated and iteratively optimized in terms of sensitivity and signal-to-noise ratio, adopting wafer-scale micro-fabrication methods. The 1/f noise in gSGFETs has been characterised and the optimisation of both, contact and channel noises was achieved by UVO-treatment at the metal/graphene interface, as well as by decoupling the graphene channel from the substrate, using different nanomaterials such as graphene encapsulation with hexagonal boron nitride (hBN), self assembled monolayers and double transferred graphene. Moreover, flexible and ultra-thin epicortical and intracortical neural probes, containing arrays of gSGFETs, have been successfully fabricated and used during in vivo microelectrocorticography recordings in anaesthesized and awake rodents. Flexible intracortical devices were inserted into the brain using a back-coating stiffening protocol with bioresobable silk fibroin protein, developed during this PhD thesis. The results presented in this PhD demonstrate the superior spatio-temporal resolution of gSGFETs compared to standard microelectordes technology; particularly the ability to map with high fidelity, infraslow activity (ISA, < 0.1 Hz) together with signals in the typical local field potential bandwidth. Today it is known that infraslow brain activity, including spreading depolarisations, contribute to the pathophysiology of several neurological disorders such as stroke, traumatic brain injury, migraine and epilepsy. However, this activity is seldom recorded due to intrinsic technical limitations of conventional AC-coupled electrodes. To demonstrate the usefulness of the developed flexible gSGFET arrays technology, recordings have been obtained with multichannel flexible graphene depth neural probes (gDNP) in relevant awake animal models of seizures and established epilepsy. ISA was detected and mapped through different cortical layers and subcortical regions, whilst simultaneously recording epileptiform activity in more conventional frequency bands (1-600Hz). Furthermore, the assessment of the long term recording stability and functionality, as well as biocompatibility of the gDNP has also been demonstrated as part of this thesis. The graphene based bioelectronic technology here described has the potential to become a gold standard tool for full bandwidth electrophysiology. This technology is envisioned to have a great impact on a broad and multidisciplinary community including neurotechnology researchers, biomedical engineers, neuroscientists studying wide-band cortical dynamics associated with spontaneous behaviour and/or brain states, as well as clinical researchers interested in the role of infraslow activity in epilepsy, stroke and migraine.

Electronic book production is a developing field which is still in its infancy. As such, there is still relatively little material available in the form of design principles or guidelines for the production of such books. It is also extremely complex, in that electronic book designers can take advantage of a number of delivery techniques which are not available to authors of traditional paper-based books. Such techniques include: multimedia (the delivery of text, pictures, sound, and moving pictures); and hypermedia (the linking of reactive information items to form non-linear structures). This research investigates some of the key issues in the design of end-user interfaces to electronic books. Essentially, this centres on three basic problems: the use of metaphors in the design of interfaces to electronic books; models for the design of multimedia pages; and the provision of various knowledge corpus structures. Interface metaphors are investigated through the implementation and evaluation of the book metaphor. Applications were developed which either embedded or did not embed the book metaphor. Subjects used these applications while undertaking a number of information access tasks. Both qualitative and performance data werecollected and some significant results were obtained. Five page models were developed (referred to as: simple; tiled; overlay; oversize; and dynamic) which were used to design a number of page structures. These page structures were evaluated using qualitative measures of user reactions to the various page structures. Seven interface dimensions were measured and again significant results were obtained. To measure the effects of knowledge corpus structure on the design of electronic books three different book structures were created: linear; tree; and network. These were investigated in the light of some common information access tasks. The results indicated that some knowledge corpus structures were more appropriate for certain types of task.

Afin d’étudier le cerveau humain dans le but d’aider les patients souffrant de maladies neurologiques, on a besoin d’une interface cérébrale entièrement implantable pour permettre l’accès direct aux neurones et enregistrer et analyser l’activité neuronale. Dans cette thèse, des interfaces cerveau-machine implantables (IMC) à très faible puissance basées sur plusieurs circuits et innovations de systèmes ont été étudiées pour être utilisées comme analyseur neuronal. Un tel système est destiné à recueillir l’activité neuronale émise par centaines de neurones tout en les activant à la demande en utilisant des moyens d’actionnement tels que l’électro- et / ou la photo-stimulation. Un tel système doit fournir plusieurs canaux d’enregistrement, tout en consommant très peu d’énergie, et présente une taille extrêmement réduite pour la sécurité et la biocompatibilité. Typiquement, un microsystème d’interfaçage avec le cerveau comprend plusieurs blocs, tels qu’un bloc analogique d’acquisition (AFE), un convertisseur analogique-numérique (ADC), des modules de traitement de signal numérique et un émetteur-récepteur de données sans fil. Un IMC extrait les signaux neuronaux du bruit, les numérise et les transmet à une station de base sans interférer avec le comportement naturel du sujet. Cette thèse se concentre sur les blocs analogiques d’acquisition à très faible consommation à utiliser dans l’IMC. Cette thèse présente des frontaux avec plusieurs stratégies innovantes pour consommer moins d’énergie tout en permettant des données de haute résolution et de haute qualité. Premièrement, nous présentons une nouvelle structure frontale utilisant un schéma de réutilisation du courant. Cette structure est extensible à un très grand nombre de canaux d’enregistrement, grâce à sa petite taille de silicium et à sa faible consommation d’énergie. L’AFE à réutilisation de courant proposée, qui comprend un amplificateur à faible bruit (LNA) et un amplificateur à gain programmable (PGA), utilise une nouvelle topologie de miroir de courant entièrement différentielle utilisant moins de transistors et améliorant plusieurs paramètres de conception, tels que la consommation d’énergie et du bruit, par rapport aux mises en oeuvre de circuit d’amplificateur de réutilisation de courant précédentes. Ensuite, dans la deuxième partie de cette thèse, nous proposons un nouveau convertisseur sigmadelta multicanal qui convertit plusieurs canaux indépendamment en utilisant un seul amplificateur et plusieurs condensateurs de stockage de charge. Par rapport aux techniques conventionnelles, cette méthode applique un nouveau schéma de multiplexage entrelacé, qui ne nécessite aucune phase de réinitialisation pour l’intégrateur lors du passage à un nouveau canal, ce qui améliore sa résolution. Lorsque la taille des puces n’est pas une priorité, d’autres approches peuvent être plus attrayantes, et nous proposons une nouvelle stratégie d’économie d’énergie basée sur un nouveau convertisseur sigma-delta à très basse consommation conçu pour réduire la consommation d’énergie. Ce nouveau convertisseur utilise une architecture basse tension basée sur une topologie prédictive innovante qui minimise la non-linéarité associée à l’alimentation basse tension.
Studying brain functionality to help patients suffering from neurological diseases needs fully implantable brain interface to enable access to neural activities as well as read and analyze them. In this thesis, ultra-low power implantable brain-machine-interfaces (BMIs) that are based on several innovations on circuits and systems are studied for use in neural recording applications. Such a system is intended to collect information on neural activity emitted by several hundreds of neurons, while activating them on demand using actuating means like electro- and/or photo-stimulation. Such a system must provide several recording channels, while consuming very low energy, and have an extremely small size for safety and biocompatibility. Typically, a brain interfacing microsystem includes several building blocks, such as an analog front-end (AFE), an analog-to-digital converter (ADC), digital signal processing modules, and a wireless data transceiver. A BMI extracts neural signals from noise, digitizes them, and transmits them to a base station without interfering with the natural behavior of the subject. This thesis focuses on ultra-low power front-ends to be utilized in a BMI, and presents front-ends with several innovative strategies to consume less power, while enabling high-resolution and high-quality of data. First, we present a new front-end structure using a current-reuse scheme. This structure is scalable to huge numbers of recording channels, owing to its small implementation silicon area and its low power consumption. The proposed current-reuse AFE, which includes a low-noise amplifier (LNA) and a programmable gain amplifier (PGA), employs a new fully differential current-mirror topology using fewer transistors. This is an improvement over several design parameters, in terms of power consumption and noise, over previous current-reuse amplifier circuit implementations. In the second part of this thesis, we propose a new multi-channel sigma-delta converter that converts several channels independently using a single op-amp and several charge storage capacitors. Compared to conventional techniques, this method applies a new interleaved multiplexing scheme, which does not need any reset phase for the integrator while it switches to a new channel; this enhances its resolution. When the chip area is not a priority, other approaches can be more attractive, and we propose a new power-efficient strategy based on a new in-channel ultra-low power sigma-delta converter designed to decrease further power consumption. This new converter uses a low-voltage architecture based on an innovative feed-forward topology that minimizes the nonlinearity associated with low-voltage supply.

Brain-Computer-Interfaces (BCIs) provide a novel way of communication by interpreting different types of brain states. This principle of reading minds makes BCIs a challenging but at the same time fascinating topic among the different disciplines of electrophysiology and biomedical-signal-processing. This work describes the development of a non-invasive BCI approach using steadystate- visual-evoked-potentials (SSVEP) as a mental strategy. SSVEP based BCIs require an external visual stimulation, which in this work is transmitted by a LCD-screen. Consequently, a visual reactive BCI is integrated as a plug-in into the open source project MNE-CPP, which provides an extensive library for brain monitoring and processing. MNE-Scan, as a standalone software from MNE-CPP, contains the necessary real-time source localization and is used as a framework for the BCI. Moreover, an expansion with a screen keyboard device shows the BCI’s practicability. The work’s result delivers a functioning SSVEP BCI approach with an average detection accuracy of 86 %. However, it is shown, that a transition from a BCI on sensor level to a BCI on source level is challenging and requires a certain pre-development, whereby a first approach of the BCI only was realized on sensor level in this work.
Tesis

Brain-computer interface (BCI) technology provides a means of communication that allows individuals with severely impaired movement to communicate with assistive devices using the electroencephalogram (EEG) or other brain signals. Dealing with the unknown embedded noise within the raw EEG and the inherent lower bandwidth of BCI are still two of the major challenges in making BC! practical for day-to-day use. The raw EEG signal recorded non-invasively during motor i ~~~y (MI) is intrinsically " ." embedded with non-Gaussian noise while the actual noise-free EEG has so far not been attained. Therefore, a filtering approach is needed to remove noise. A novel quantum mechanics motivated alternative neural information processing architecture using the Schrodinger wave equation (SWE) is proposed to filter and thereby enhance the information from the otherwise noisy EEG signals. This novel filtering approach is constructed using a layer of neurons within the neural network framework, referred to as the Recurrent Quantum Neural Network (RQNN) that recurrently computes a time-varying probability density function (pdf) for the measurement of the observed signal. The raw EEG sample is encoded in terms of a particle-like wave packet that can be used to accurately filter noise from the EEG using an unsupervised learning scheme without making any assumption about the underlying distribution. The RQNN enhanced EEG signal is more easily classified than the raw signal. Another major challenge in two-class BC! systems is also addressed in this thesis, namely the inherent lower bandwidth of the communication channel that may lead to a sluggish response in suitably controlling a mobile robotic device. An intelligent and adaptive user interface, which plays a very important role as a front-end display for the BCI user is proposed. The framework of the proposed intelligent Adaptive User Interface (iAUI) i.e., the brain-robot interface is consistent for a range of applications e.g., for controlling either a mobile robot or a robotic arm. The iAUI for mobile robot offers a real-time prioritized list of all the options for selection by the user. Prioritized update ofthe iAUI is possible by utilizing information obtained from the sonar sensors mounted on the mobile robot. Through iAUI the user is always offered the most likely choice, thereby improving the information transfer rate. Similarly, the interface for controlling the robot arm displays the list of available objects for user selection depending on the real-time information from the camera view of the robot arm. Results on multiple participants show that both the main contributions, the RQNN filtering and the iAUI address to a large extent the issues of dealing with unknown embedded noise within the raw EEG and the inherent lower bandwidth of BCI. xv Abstract

The automatic detection of code errors is a ubiquitous part of the quality assurance process performed during software development. However, graphical errors that may occur in user interfaces are often detected manually. This report examines if deep neural networks (DNNs), may be used to automatically detect two common types of anomalies present in a graphical user interface. The results point towards this being the case for the particular dataset used in this report.
Automatisk detektering av kodfel är standard i kvalitetsarbetet som utförs vid mjukvaruveckling. Grafiska fel som kan uppstå i användargränssnitt upptäcks dock ofta manuellt. Den här rapporten undersöker ifall djupa neurala nätverk kan användas för att automatiskt detektera två vanliga fel som uppstår i användargränssnitt. Resultaten indikerar att så är fallet åtminstone för det specifika dataset som används.

This research investigates the use of conducting polymer coatings on platinum (Pt) electrodes for use in neuroprostheses. Conducting polymers aim to provide an environment conducive to neurite outgrowth and attachment at the electrode sites, producing intimate contact between neural cells and stimulating electrodes. Conducting polymers were electropolymerised onto model Pt electrodes. Conventional polymers polypyrrole (PPy) and poly-3,4-ethylenedioxythiphene (PEDOT) doped with polystyrenesulfonate (PSS) and para-toluenesulfonate (pTS)were investigated. Improvement of material properties was assessed through the layering of polymers with multi-walled carbon nanotubes (MWNTs). The ability to incorporate cell attachment bioactivity into polymers was examined through the doping of PEDOT with anionic laminin peptides DCDPGYIGSR and DEDEDYFQRYLI. Finally, nerve growth factor (NGF), was entrapped in PEDOT during polymerisation and tested for neurite outgrowth bioactivity against the PC12 cell line. Each polymer modification was assessed for electrical performance over multiple reduction-oxidation cycles, conductivity and impedance spectroscopy, mechanical adherence and hardness, and biological response. Scanning electron microscopy was used to visualise film topography and x-ray photon spectroscopy was employed to examine chemical constitution of the polymers. For application of electrode coatings to neural prostheses, optimal bioactive conducting polymer PEDOT/pTS/NGF was deposited on electrode arrays intended for implantation. PC12s were used to assess the bioactivity of NGF functionalised PEDOT when electrode size was micronised. Flexibility of the design was tested by tailoring PEDOT bioactivity for the cloned retinal ganglion cell, RGC-5, differentiated via staurasporine. It was established that PEDOT films had superior electrical and cell growth characteristics, but only PPy was able to benefit from incorporation of MWNTs. Bioactive polymers were produced through inclusion of both laminin peptides and NGF, but the optimum film constitution was found to be PEDOT doped with pTS with NGF entrapped during electrodeposition. Application of this polymer to an implant device was confirmed through positive neurite outgrowth on vision prosthesis electrode arrays. The design was shown to be flexible when tailored for RGC-5s, with differentiation occurring on both PEDOT/pTS and PEDOT/DEDEDYFQRYLI. Conducting polymers demonstrate the potential to improve electrode-cell interactions. Future work will focus on the effect of electrical stimulation and design of bioactive polymers with improved cell attachment properties.

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2011.
Cataloged from PDF version of thesis.
Includes bibliographical references (p. 179-187).
In the past few decades, direct recordings from different areas of the brain have enabled scientists to gradually understand and unlock the secrets of neural coding. This scientific advancement has shown great promise for successful development of practical brain-machine interfaces (BMIs) to restore lost body functions to patients with disorders in the central nervous system. Practical BMIs require the uses of implantable wireless neural recording systems to record and process neural signals, before transmitting neural information wirelessly to an external device, while avoiding the risk of infection due to through-skin connections. The implantability requirement poses major constraints on the size and total power consumption of the neural recording system. This thesis presents the design of an ultra-low-power implantable wireless neural recording system for use in brain-machine interfaces. The system is capable of amplifying and digitizing neural signals from 32 recording electrodes, and processing the digitized neural data before transmitting the neural information wirelessly to a receiver at a data rate of 2.5 Mbps. By combining state-of-the-art custom ASICs, a commercially-available FPGA, and discrete components, the system achieves excellent energy efficiency, while still offering design flexibility during the system development phase. The system's power consumption of 6.4 mW from a 3.6-V supply at a wireless output data rate of 2.5 Mbps makes it the most energy-efficient implantable wireless neural recording system reported to date. The system is integrated on a flexible PCB platform with dimensions of 1.8 cm x 5.6 cm and is designed to be powered by an implantable Li-ion battery. As part of this thesis, I describe the design of low-power integrated circuits (ICs) for amplification and digitization of the neural signals, including a neural amplifier and a 32-channel neural recording IC. Low-power low-noise design techniques are utilized in the design of the neural amplifier such that it achieves a noise efficiency factor (NEF) of 2.67, which is close to the theoretical limit determined by physics. The neural recording IC consists of neural amplifiers, analog multiplexers, ADCs, serial programming interfaces, and a digital processing unit. It can amplify and digitize neural signals from 32 recording electrodes, with a sampling rate of 31.25 kS/s per channel, and send the digitized data off-chip for further processing. The IC was successfully tested in an in-vivo wireless recording experiment from a behaving primate with an average power dissipation per channel of 10.1 [mu]W. Such a system is also widely useful in implantable brain-machine interfaces for the blind and paralyzed, and in cochlea implants for the deaf.
by Woradorn Wattanapanitch.
Ph.D.

Brain Machine Interfaces (BMI) provide a communication pathway between the electrical conducting units of the brain (neurons) and external devices. BMI technology may provide improved neurological and physiological functions to patients suffering from disabilities due to damaged nervous systems. Unfortunately, microelectrodes used in Intracortical Neural Interfaces (INI), a subset of the BMI device family, have yet to demonstrate long-term in vivo performance due to material, mechanical and electrical failures. Many state-of-the-art INI devices are constructed using stacks of multiple materials, such as silicon (Si), titanium (Ti), platinum (Pt), parylene C, and polyimide. Not only must each material tolerate the biological environment without exacerbating the inflammatory response, each of the materials used must physically withstand the environment as well as interact well with each other. One approach to address abiotic mechanisms has been optimizing the materials required to fabricate the INI devices. Silicon Carbide (SiC) is a physically robust, hemo and biocompatible, and chemically inert semiconductor. An ‘all-SiC’, or monolithic SiC, device may be the disruptive technology needed in the BMI field to finally achieve long-term and wide-spread use of this technology in humans. The all-SiC device concept is where SiC serves as all device layers: the base (substrate), the conducting traces (electrodes), and the surface conformal insulating layer. The monolithic SiC neural probe is realized by forming high-quality pn junctions of heavily doped SiC on a layer of the opposite polarity. Heavily doped semiconductors display semi-metallic electrical performance, which allow for efficient electrical conduction in the electrode without the deleterious effects of metal ions interacting with the neural environment. The conformal insulator is realized using amorphous-SiC (a-SiC) which can be patterned to open windows to allow electrical conduction to occur between the electrode tips and the brain. Several generations of monolithic SiC devices have been fabricated, tested and are reported in this dissertation. The devices were fabricated utilizing two polytypes of SiC (4H-SiC and 3C-SiC). The monolithic SiC microelectrodes were fabricated utilizing techniques used in the fabrication of Si based microelectrodes. Monolithic SiC devices fabricated include planar single-ended MEAs (with varying lengths and varying active recording area), 60-channel MEAs for in-vitro studies, and 16-electrode Michigan style neural probes for in-vivo studies. Electrical testing of the pn junction demonstrated that the 4H-SiC device can block a forward bias voltage of up to 2.3V and displays reverse bias leakage below 1 nArms well past -20V. Current leakage between adjacent electrodes was ~7.5 nArms over a voltage range of -50V to +50V. Furthermore, electrochemical results show that the 4H-SiC microelectrodes interact with an electrochemical environment primarily through capacitive mechanisms and has an impedance comparable to gold electrodes. Electrode impedance ranged from 675±130 kΩ (GSA = 496 µm2) to 46.5±4.80 kΩ (GSA = 500K µm2). However, the 4H-SiC devices cannot deliver charge as efficiently as other conventionally used microelectrode materials, such as iridium oxide (IrOx), but a larger water window compensates for this since larger stimulation voltages are supported compared to IrOx. All studies and data collected thus far indicate that the monolithic SiC neural device can aid in the advancement of chronic INI use in clinical settings. The all-SiC devices rely on the integration of only robust and highly compatible SiC material, they may offer a promising solution to probe delamination and biological rejection associated with the use of multiple materials used in many current INI devices. Follow-on work is planned to prove this assertion via in vivo studies.

To develop an optimal cognitive neural prosthetic to assist patients with motor dysfunction we must have the ability to accurately decode reach goals from patient's neural activity. Implantable microelectrode arrays are often used to record neural activity. A challenge researchers typically face when using arrays is that only a small number of neurons can be isolated for recording. Moreover, these neurons tend to over-represent certain spatial locations while under-representing others, thereby outputting a bias towards only a portion of reachable directions. Several studies using brain-machine interface (BMI) experiments have shown that subject's neural activity undergoes a functional reorganization in response to constraints imposed by a BMI. In this study, we investigated whether we could use reward to not only actively induce the reorganization of neural activity, but also guide it to increase the amount of information extracted from a neural population. To address this, we recorded neural activity using 96 electrodes implanted in parietal and pre-motor areas of a Rhesus macaque monkey. We designed a brain-control task in which an adaptive Bayesian decoder used this activity to predict the intended reach goals of the monkey. A successful brain-control trial was established when the decoder predicted the instructed reach goal, for which the monkey was rewarded with juice. Real-time visual feedback of decode performance was provided by a visual cue. Decode performance also determined the size of the reward to be received. Our findings show that decoding performance improved over the course of a recording session and that this improvement coincided with both changes in neural firing activity as well as a more uniform distribution in the neuronal representation of reach space. These results suggest that reward can be used to actively induce neural changes that increase the amount of information that can be extracted from the brain to subsequently improve decoding.
Pour développer une prothèse neurale basée sur la cognition qui aiderait les patients souffrant de dysfonction moteur, nous devons être capable de décoder les intentions de mouvements des patients à partir de leur activité cérébrale. Des matrices de microélectrodes sont fréquemment employées pour échantillonner l'activité des neurones. Un défi auquel font habituellement face les chercheurs, lorsqu'ils utilisent des matrices de microélectrode, est que seul un petit ensemble de neurones peut être isolé et échantillonné adéquatement. De plus, ces neurones ont tendance à surreprésenter certaine région de l'espace et à présenter un manque dans la représentation des autres. Il y a donc un biais dans la représentation de l'espace dans lequel s'effectue normalement les mouvements. Plusieurs études réalisées sur des expériences impliquant des interfaces cerveau-machine (ICM) ont rapporté que l'activité des neurones des sujets entreprend une réorganisation fonctionnelle, suite aux contraintes imposées par l'ICM. Dans cette étude, nous avons testé l'hypothèse selon laquelle nous pourrions utiliser la récompense pour, non seulement induire activement une reconfiguration de l'activité, mais aussi pour guider la réorganisation fonctionnelle de façon à ce qu'elle résulte en une augmentation de la quantité d'information qui peut être extraite d'une population de neurones. Pour ce faire, nous avons échantillonné l'activité des neurones du cortex pariétal et de l'aire pré-moteur d'un macaque Rhésus en y implantant 96 électrodes. Nous avons conçu une tâche de contrôle par la pensé dans laquelle un décodeur Bayesien adaptatif utilise l'activité des neurones pour prédire les intentions de mouvement du bras du singe. Lors des essais fructueux, le singe a été récompensé par du jus. Au cours d'un essai, une indication visuelle délivrée en temps réel informait le singe de la performance du décodeur à déduire ses intentions. La performance du décodeur fixait la quantité de jus donnée au singe en récompense. Nos résultats démontrent que les performances du décodeur ont augmenté au cours du déroulement de l'expérience en même temps que les changements dans l'activité des neurones ont entraîné une représentation plus uniforme de l'espace. Ces résultats suggèrent que la récompense peut-être utilisée pour induire des changements dans l'activité des neurones qui améliore les performances d'un décodeur en augmentant la quantité d'information qui peut-être extraite de l'activité cérébrale.

Beginning with the introduction of electrically powered prostheses more than 65 years ago surface electromyographic (sEMG) signals recorded from residual muscles in amputated limbs have served as the primary source of upper-limb myoelectric prosthetic control. The majority of these devices use one or more neural interfaces to translate the sEMG signal amplitude into voltage control signals that drive the mechanical components of a prosthesis. In so doing, users are able to directly control the speed and direction of prosthetic actuation by varying the level of muscle activation and the associated sEMG signal amplitude. Consequently, in spite of decades of development, myoelectric prostheses are prone to highly variable functional control, leading to a relatively high-incidence of prosthetic abandonment among 23-35% of upper-limb amputees. Efforts to improve prosthetic control in recent years have led to the development and commercialization of neural interfaces that employ pattern recognition of sEMG signals recorded from multiple locations on a residual limb to map different intended movements. But while these advanced algorithms have made strident gains, there still exists substantial need for further improvement to increase the reliability of pattern recognition control solutions amongst the variability of muscle co-activation intensities. In efforts to enrich the control signals that form the basis for myoelectric control, I have been developing advanced algorithms as part of a next generation neural interface research and development, referred to as Motor Unit Drive (MU Drive), that is able to non-invasively extract the firings of individual motor units (MUs) from sEMG signals in real-time and translate the firings into smooth biomechanically informed control signals. These measurements of motor unit firing rates and recruitment naturally provide high-levels of motor control information from the peripheral nervous system for intact limbs and therefore hold the greater promise for restoring function for amputees. The goal for my doctoral work was to develop advanced algorithms for the MU Drive neural interface system, that leverage MU features to provide intuitive control of multiple degrees-of-freedom. To achieve this goal, I targeted 3 research aims: 1) Derive real-time MU-based control signals from motor unit firings, 2) Evaluate feasibility of motor unit action potential (MUAP) based discrimination of muscle intent 3) Design and evaluate MUAP-based motion Classification of motions of the arm and hand.

El descobriment experimental del grafè el 2004 va suposar l’aparició d’un nou camp de recerca basat en materials bidimensionals, que investiga les seves propietats per a aplicacions en electrònica, fotònica i optoelectrònica i, recentment, també en tecnologia biomèdica. La neurotecnologia en particular, és un tema que podria beneficiar-se fortament d’aquests nous materials, ja que la seva naturalesa mecànica i química els permet formar una interfície estable i adaptable a la topologia del cervell. El transistor de grafè de porta liquida (gSGFET) és un dels diversos dispositius de detecció emergents que utilitzen materials prims i ha demostrat un gran potencial per a les interfícies cervell-màquina (BMIs), ja que és capaç de registrar l’activitat neuronal amb alta precisió. Històricament s’han utilitzat els elèctrodes per a aquesta aplicació, ja que són fàcils de fabricar i els senyals enregistrats es poden adquirir mitjançant tècniques de lectura senzilles. En els darrers anys, però, s’ha trobat que l’ús de transistors es beneficiós per a aplicacions específiques, ja que també son capaços de registrar l’activitat oscil·latòria lenta, que encara està en la seva majoria inexplorada. A més, els transistors permeten construir matrius de sensors amb un gran nombre de llocs de gravació, ja que, al contrari que els elèctrodes, no cal combinar-los amb components electrònics addicionals per permetre esquemes d’adreçament sofisticats, facilitant així la complexitat tecnològica de la fabricació d’aquestes matrius. L’elevat soroll intrínsec del sensor continua sent un problema crític en electrofisiologia a causa de la petita amplitud dels senyals neuronals a la superfície de l’escorça cerebral. La naturalesa geomètrica del grafè i d’altres materials bidimensionals els exposa a influències degradants del seu entorn i fa difícil crear interfícies d’alta qualitat amb aquests materials. Això condueix, per exemple, a un augment del soroll de baixa freqüència del gSGFET, que contamina la banda de freqüències d’interès per als enregistraments neuronals. Per tant, com a primer pas cap al desenvolupament d’un conjunt de sensors neuronals d’alta qualitat, en el capítol 3 s’exploren les millores tecnològiques que permeten reduir aquest soroll intrínsec del dispositiu. A continuació es descriu la fabricació i caracterització de interfícies neuronals de alta densitat, fins a 1000 gSGFET sensors sobre substrat flexible. Com que la mida del connector que augmenta directament amb el nombre de sensors, aquest imposa restriccions en el numero i la densitat de llocs de gravació assolibles a la matriu. D’aquesta forma, els esquemes de lectura multiplexada compatibles amb la tecnologia gSGFET són de vital importància. La multiplexació permet la combinació de múltiples fluxos d’informació en un sol senyal i, per tant, permetria superar les limitacions de connectivitat de les BMIs actuals. Més concretament, en al capítol 4 es presenta la compatibilitat de la tecnologia gSGFET amb els dos esquemes més comuns de lectura de dades multiplexades, la divisió de freqüències i la divisió de temps. En última instància, en el capítol 5 s’explora la integració monolítica de circuits de adreçament a la matriu de sensors flexibles, ja que això evitaria l’acoblament entre llocs que pot degradar greument la fidelitat del senyal enregistrat en grans matrius de sensors multiplexats. Per a aquest propòsit, es suggereix l’ús dels materials de “transition-metall-dichalcogenides” de dues dimensions (per exemple, MoS2), ja que combinen la flexibilitat mecànica amb una banda prohibida àmplia, necessària per obtenir transistors d’efecte de camp amb altes relacions encès / apagat. Un conjunt híbrid de sensors gSGFET / MoS2-FET amb funcionalitat de lectura multiplexada es presenta com a primer prototip cap a una nova generació de BMIs i es presenta un full de ruta per a l’ampliació de la tecnologia.
El descubrimiento experimental del grafeno en 2004 marcó el nacimiento de un nuevo campo de investigación basado en materiales bidimensionales, investigando sus propiedades para aplicaciones en electrónica, fotónica y optoelectrónica y, recientemente, también en tecnología biomédica. El transistor de puerta liquida basado en grafeno (gSGFET) es uno de los dispositivos de detección emergentes que utilizan materiales delgados, y que ha demostrado un gran potencial para las interfaces cerebro-máquina (BMIs), ya que es capaz de registrar con alta precisión la actividad neuronal. Históricamente, se han utilizado electrodos para tal aplicación, ya que son fáciles de fabricar y las señales grabadas se pueden adquirir usando técnicas simples de lectura. En los últimos años, sin embargo, se descubrió que el uso de sensores con un diseño de transistor es beneficioso para aplicaciones específicas, ya que también son capaces de registrar la actividad oscilatoria lenta, que todavía está en su mayoría inexplorada. Además, los transistores permiten construir matrices de sensores con una gran cantidad de sitios de grabación, ya que, a diferencia de los electrodos, no es necesario combinarlos con componentes electrónicos adicionales para permitir esquemas de direccionamiento sofisticados, lo que alivia en gran medida la complejidad tecnológica de fabricar estas matrices. El alto ruido intrínseco del sensor sigue siendo un problema crítico en electrofisiología debido a la pequeña amplitud de las señales neuronales en la superficie de la corteza. La naturaleza geométrica del grafeno y otros materiales bidimensionales los expone a influencias degradantes de su entorno y dificulta la creación de interfaces de alta calidad con dichos materiales. Esto conduce, por ejemplo, a un aumento de ruido de baja frecuencia en el gSGFET que contamina la banda de frecuencia de interés para los registros neuronales. Por lo tanto, como un primer paso hacia el desarrollo de una matriz de sensores neurales de alta calidad, en el capítulo 3 se exploran las mejoras tecnológicas que permiten reducir dicho ruido intrínseco del dispositivo. A continuación, se describe la fabricación y caracterización en sustrato flexible de matrices de sensores neurales de alta densidad, hasta 1000 gSGFET sensores. Dado que el tamaño del conector en aumenta directamente con el número de sensores, este impone restricciones en el recuento y en la densidad de los sitios de grabación que se pueden lograr en la matriz. De esta forma, los esquemas de lectura multiplexados compatibles con la tecnología gSGFET son de vital importancia. La multiplexación permite la combinación de múltiples flujos de información en una sola señal y, por lo tanto, permitiría superar las limitaciones de conectividad de las BMIs actuales. Más concretamente, en el capítulo 4 se presenta la compatibilidad de la tecnología gSGFET con los dos esquemas más comunes de lectura de datos multiplexados, la multiplexación por división de frecuencia y por división de tiempo. En última instancia, en el capítulo 5 se explora la integración monolítica de circuitos de direccionamiento en la matriz de sensores flexibles, ya que esto evitaría el acoplamiento entre sitios que puede degradar severamente la fidelidad de la señal grabada en grandes conjuntos de sensores multiplexados. Para este propósito, se sugiere el uso los materiales de “transition-metal-dichalcogenides” de dos dimensiones (por ejemplo, MoS2), ya que combinan la flexibilidad mecánica con una banda prohibida amplia, necesaria para obtener transistores de efecto de campo con altas relaciones encendido/apagado. Un conjunto de sensores híbridos gSGFET / MoS2-FET con funcionalidad de lectura multiplexada se presenta como un primer prototipo hacia una nueva generación de BMIs y se presenta una hoja de ruta para la ampliación de la tecnología.
The experimental discovery of graphene in 2004 marked the advent of a new research field based on two-dimensional materials, investigating their properties for applications in electronics, photonics and optoelectronics and, recently, also biomedical technology. Neurotechnology in particular, is a subject which could strongly benefit from these new materials, as their mechanical and chemical nature allow them to form a stable, conformable interface with the brain. The graphene solution-gated field-effect transistor (gSGFET) is one of several emerging sensing devices utilizing thin materials, and has shown great potential for brain-machine interfaces (BMIs), as it is able to record neural activity with high accuracy. Historically electrodes have been favored for such application, as they are easy to fabricate and the recorded signals can be acquired using simple read-out techniques. In recent years, however, the use of sensors with a transistor design was found to be beneficial for specific applications as they also unveil slow oscillatory activity, which is yet mostly unexplored. Moreover, transistors allow to construct sensor arrays with a large number of recording sites, since in contrary to electrodes they do not need to be combined with additional electronic components to enable sophisticated addressing schemes, thus strongly easing the technologic complexity of fabricating these arrays. High intrinsic sensor noise remains a critical issue in electrophysiology due to the small amplitude of neural signals on the surface of the cortex. The geometric nature of graphene and other two-dimensional materials exposes them to degrading influences from their surroundings and makes it challenging to create high-quality interfaces with such materials. This leads to, for example, augmented low-frequency noise in the gSGFET which contaminates the frequency band of interest for neural recordings. Thus, as a first step towards the development of a high quality neural sensor array, technological improvements which allow to lower such intrinsic device noise are explored in chapter 3. Next, the fabrication and characterization of high-density neural sensor arrays of above 1000 gSGFET sensors on flexible substrate is described. As a rapidly increasing size of the connector footprint with the number of sensors poses restrictions on the count and the density of recording sites achievable on the array, the compatibility of the gSGFET technology with multiplexed readout schemes is of critical importance. Multiplexing enable the combination of multiples streams of information into a single signal and would thereby allow to overcome connectivity limitations of the state-of-the-art BMIs. More concretely, the compatibility of the gSGFET technology with the two most common schemes of multiplexed data readout, namely frequency-division and time-division multiplexing, is presented in chapter 4. Ultimately, a monolithic integration of addressing circuitry into the flexible sensor array is explored in chapter 5, as this would prevent inter-site crosstalk which can severely degrade the fidelity of the recorded signal in large multiplexed sensor arrays. Two-dimension transition-metal-dichalcogenides (e.g. MoS2) are suggested for this purpose, as they combine mechanical flexibility with a wide bandgap necessary for building field-effect transistors with high on/off-ratios. A hybrid gSGFET/MoS2-FET sensor array with multiplexed readout functionality is showcased as a first prototype towards a new generation of BMIs and a roadmap for the technology's scale-up is presented.
Universitat Autònoma de Barcelona. Programa de Doctorat en Enginyeria Electrònica i de Telecomunicació

This research consists of probabilistic modeling of speech audio signals and deep-brain neurological signals in brain-computer interfaces. A significant portion of this research consists of a collaborative effort with Neural Signals Inc., Duluth, GA, and Boston University to develop an intracortical neural prosthetic system for speech restoration in a human subject living with Locked-In Syndrome, i.e., he is paralyzed and unable to speak. The work is carried out in three major phases. We first use kernel-based classifiers to detect evidence of articulation gestures and phonological attributes speech audio signals. We demonstrate that articulatory information can be used to decode speech content in speech audio signals. In the second phase of the research, we use neurological signals collected from a human subject with Locked-In Syndrome to predict intended speech content. The neural data were collected with a microwire electrode surgically implanted in speech motor cortex of the subject's brain, with the implant location chosen to capture extracellular electric potentials related to speech motor activity. The data include extracellular traces, and firing occurrence times for neural clusters in the vicinity of the electrode identified by an expert. We compute continuous firing rate estimates for the ensemble of neural clusters using several rate estimation methods and apply statistical classifiers to the rate estimates to predict intended speech content. We use Gaussian mixture models to classify short frames of data into 5 vowel classes and to discriminate intended speech activity in the data from non-speech. We then perform a series of data collection experiments with the subject designed to test explicitly for several speech articulation gestures, and decode the data offline. Finally, in the third phase of the research we develop an original probabilistic method for the task of spike-sorting in intracortical brain-computer interfaces, i.e., identifying and distinguishing action potential waveforms in extracellular traces. Our method uses both action potential waveforms and their occurrence times to cluster the data. We apply the method to semi-artificial data and partially labeled real data. We then classify neural spike waveforms, modeled with single multivariate Gaussians, using the method of minimum classification error for parameter estimation. Finally, we apply our joint waveforms and occurrence times spike-sorting method to neurological data in the context of a neural prosthesis for speech.

Dans ce projet nous nous intéresserons à la création d’un nouveau modèle permettant de simuler des enregistrements extracellulaires de l’activité électrique neurale dans le système nerveux périphérique. Ce modèle fut développé pour permettre une meilleure compréhension de l’impact des différentes propriétés des interfaces sur la qualité des signaux recueillis. Ce projet fut en particulier conduit pour répondre au contexte actuel qui voit le développement de nombreuses applications dans le domaine des neuro-prothèses et autres interfaces neurales à but biomédical. Nos intentions étaient de fournir un nouvel outil permettant de mieux comprendre les particularités des interfaces existantes ou d’aider à leur amélioration et à la planification de futures innovations. Ce modèle est construit comme la synthèse de la compréhension actuelle des différents rouages biophysiques impactant les enregistrements. Sa structure peut être perçue comme l’assemblage de différents sous-systèmes interconnectés et représentant chacun une dimension du processus. Il s’avère particulièrement efficace pour l’analyse comparative des performances entre diffèrent types/géométries d’électrodes invasives. Dans ce document, nous nous efforcerons d’expliquer en détail la structure et les paramètres de notre modèle. Nous décrirons ensuite les différents tests que nous avons entrepris pour sa validation expérimentale, ainsi que les différentes voies d’applications que nous avons commencé à explorer. Nous finirons par décrire les améliorations qui nous sont apparues comme nécessaires ou possibles et par une discussion sur les ouvertures futures offerte à ce domaine de recherche
Neural interfaces are becoming a newly dynamic and promising field especially thanks to the numerous applications they could have in the biomedical domain. A great deal of these applications requires a monitoring of targeted neural activity. Among the different technologies available for such recording practice, chronic electrodes implanted in the peripheral nervous system offer a good compromise on the resolution versus invasiveness technological constraint. A large array of electrodes has been developed in this intention but there is still only a limited comprehension of their recording principles and weakness. This makes difficult any targeted improvement of the electrodes and led this field to be mainly dominated by a trial and error empirical approach simultaneously costly in funds, animal lives and time. In particular, intrafascicular electrodes, while providing exiting results for stimulation, have often failed in recordings. These electrodes typically show interesting recording performance right after implantation but have rapid decline of their efficacy up to the points that they often become useless after a few weeks. Such performance proves problematic as they drastically limit the transfer of experimental results to human applications. The extent of our work has been the development of a theoretical framework for the study of implantable electrodes. Our goal here has been to construct a model that could be used as a platform to better understand implanted electrode and compare their performance and possible improvement. We focused our work on intrafascicular electrode for the peripheral nervous system. However, our procedure could easily be applied to other type of interface. During this project we first constructed a detailed model of the recording biophysical process happening at the peripheral nerve electrical interface. This model encompasses all the mechanism known to influence the quality and shape of neural activity recordings. We have then recreated within our model specific controlled experiments and by comparing the properties of the simulated recording with their experimental counterparts demonstrated the potency of our approach to produce bio-plausible signals. This validated our model as an in silico alternative to compare and test electrodes. We then further developed this model to also simulate some of the changes happening in the nerve post implantation. In particular, we found that the growth of the fibrotic scar could already explain a large part of the signal degradation happening in the first weeks. Then to demonstrate the adaptability of this model we used it to compare the performance of the main type of electrodes implanted nowadays peripherally. Finally, as the main weakness of our model relied in its relative complexity and the related long computing time, we started to analyze how this model could be simplified without losing the precision necessary for the intended applications. In conclusion, this project led to the creation of a model which in its current form can be used as an in silico platform to test and compare electrodes. This will facilitate the planning and development of future peripheral neural interface by proving both more economical and informative that current strategies. Conjointly, we opened the way to future improvement of our model, leading to more practicality

Esta dissertação de mestrado tem por objetivo analisar os conjuntos-K, uma hierarquia de redes neurais biologicamente mais plausíveis, e aplicá-los ao problema de classificação de imagética motora através do eletroencefalograma (EEG). A imagética motora consiste no ato de processar um movimento motor da memória humana de longo tempo para a memória de curto prazo. A imagética motora deixa um rastro no sinal do EEG que torna possível a identificação e classificação dos diferentes movimentos motores. A tarefa de classificação de imagética motora através do EEG é reconhecida como complexa devido à não linearidade e quantidade de ruído da série temporal do EEG e da pequena quantidade de dados disponíveis para aprendizagem. Os conjuntos-K são um modelo conexionista que simula o comportamento dinâmico e caótico de populações de neurônios do cérebro e foram modelados com base em observações do sistema olfatório feitas por Walter Freeman. Os conjuntos-K já foram aplicados em diversos domínios de classificação diferentes, incluindo EEG, tendo demonstrado bons resultados. Devido às características da classificação de imagética motora, levantou-se a hipótese de que a aplicação dos conjuntos-K na tarefa pudesse prover bons resultados. Um simulador para os conjuntos-K foi construído para a realização dos experimentos. Não foi possível validar a hipótese levantada no trabalho, dado que os resultados dos experimentos realizados com conjuntos-K e imagética motora não apresentaram melhorias significativas para a tarefa nas comparações realizadas.
This dissertation aims to examine the K-sets, a hierarchy of biologically plausible neural networks, and apply them to the problem of motor imagery classification through electroencephalogram (EEG). Motor imagery is the act of processing a motor movement from long-term to short-term memory. Motor imagery leaves a trail in the EEG signal, which makes possible the identification and classification of different motor movements. Motor imagery classification is a complex problem due to non-linearity of the EEG time series, low signal-to-noise ratio, and the small amount of data typically available for learning. K-sets are a connectionist model that simulates the dynamic and chaotic behavior of populations of neurons in the brain, modeled based on observations of the olfactory system by Walter Freeman. K-sets have already been used in several different classification domains, including EEG, showing good results. Due to the characteristics of motor imagery classification, a hypothesis that the application of K-sets in the task could provide good results was raised. A simulator for K-sets was created for the experiments. Unfortunately, the hypothesis could not be validated, as the results of the conducted experiments with K-sets and motor imagery showed no significant improvements in comparison in the task performed.

Um dos desafios propostos por pesquisadores em neurociência aos engenheiros biomédicos é a interação cérebro-máquina. O sistema nervoso comunica-se interpretando sinais eletroquímicos, e circuitos implantáveis podem tomar decisões de modo a interagir com o meio biológico. Sabe-se também que a doença de Parkinson está relacionada a um déficit do neurotransmissor dopamina. Para controlar a concentração de dopamina diferentes técnicas tem sido empregadas como estimuladores elétricos, magnéticos e drogas. Neste trabalho obteve-se o controle da concentração do neurotransmissor de maneira automática uma vez que atualmente isto não é realizado. Para tanto, projetou-se e desenvolveu-se quatro sistemas: a estimulação cerebral profunda ou deep brain stimulation (DBS), a estimulação transmagnética ou transmagnetic stimulation (TMS), um controle de bomba de infusão ou infusion pump control (IPC) para a entrega de drogas e um sistema de voltametria cíclica de varredura rápida ou fast scan ciclic voltammetry (FSCV) (circuito que detecta variações de concentração de neurotransmissores como a dopamina - DA). Também foi necessário o desenvolvimento de softwares para a visualização de dados e análises em sincronia com acontecimentos ou experimentos correntes, facilitando a utilização destes dispositivos quando emprega-se bombas de infusão e a sua flexibilidade é tal que a DBS ou a TMS podem ser utilizadas de maneira manual ou automática além de outras técnicas de estimulação como luzes, sons, etc. O sistema desenvolvido permite controlar de forma automática a concentração da DA. A resolução do sistema é de 0.4 µmol/L podendo-se ajustar o tempo para correção da concentração entre 1 e 90 segundos. O sistema permite controlar concentrações entre 1 e 10 µmol/L, com um erro de cerca de +/- 0,8 µmol/L. Embora desenhado para o controle da concentração de dopamina o sistema pode ser utilizado para controlar outros neurotransmissores. Propõe-se continuar o desenvolvimento em malha fechada empregando FSCV e DBS (ou TMS, ou infusão), utilizando modelos animais parkinsonianos.
One of the challenges to biomedical engineers proposed by researchers in neuroscience is brain machine interaction. The nervous system communicates by interpreting electrochemical signals, and implantable circuits make decisions in order to interact with the biological environment. It is well known that Parkinson’s disease is related to a deficit of dopamine (DA). Different methods has been employed to control dopamine concentration like magnetic or electrical stimulators or drugs. In this work was automatically controlled the neurotransmitter concentration since this is not currently employed. To do that, four systems were designed and developed: deep brain stimulation (DBS), transmagnetic stimulation (TMS), Infusion Pump Control (IPC) for drug delivery, and fast scan cyclic voltammetry (FSCV) (sensing circuits which detect varying concentrations of neurotransmitters like dopamine caused by these stimulations). Some softwares also were developed for data display and analysis in synchronously with current events in the experiments. This allowed the use of infusion pumps and their flexibility is such that DBS or TMS can be used in single mode and other stimulation techniques and combinations like lights, sounds, etc. The developed system allows to control automatically the concentration of DA. The resolution of the system is around 0.4 µmol/L with time correction of concentration adjustable between 1 and 90 seconds. The system allows controlling DA concentrations between 1 and 10 µmol/L, with an error about +/- 0.8 µmol/L. Although designed to control DA concentration, the system can be used to control, the concentration of other substances. It is proposed to continue the closed loop development with FSCV and DBS (or TMS, or infusion) using parkinsonian animals models.

Electrical Engineering
Ph.D.
Decoding algorithms for brain machine interfaces map neural firing times to the underlying biological output signal through dynamic tuning functions. In order to maintain an accurate estimate of the biological signal, the state of the tuning function parameters must be tracked simultaneously. The evolution of this system state is often estimated by an adaptive filter. Recent work demonstrates that the Bayesian auxiliary particle filter (BAPF) offers improved estimates of the system state and underlying output signal over existing techniques. Performance of the BAPF is evaluated under both ideal conditions and commonly encountered spike detection errors such as missed and false detections and missorted spikes. However, this increase in neuronal signal decoding accuracy is at the expense of an increase in computational complexity. Real-time execution of the BAPF algorithm for neural signals using a sequential processor becomes prohibitive as the number of particles and neurons in the obs
Temple University--Theses

This thesis analyzes the relationship between users&rsquo
characteristics and users&rsquo
interface preferences. An online survey is developed for this study. This survey composed of two types of questions: (1) users&rsquo
personal information such as age, gender, country, cognitive structure, and also computer experience and (2) user interface elements. More than 2,500 participants from 120 different countries throughout the world completed our survey. Results were analyzed using cross tables. Our findings show that there is a relationship between users&rsquo
characteristics and users&rsquo
interface preferences. In the presence of this relationship, an artificial neural network model is developed for the estimation of the interface preferences based on the user characteristics.

O registro de sinais neurais através de matrizes de microeletrodos implantáveis no meio extracelular do córtex cerebral tem-se tornado um paradigma experimental para a neurociência. Por outro lado, a pesquisa recente sobre neuropróteses motoras tem mostrado que é possível decodificar comandos motores a partir dos sinais registrados no meio extracelular do córtex cerebral. Em ambos os contextos, neurociência experimental e desenvolvimento de neuropróteses motoras, um dos aspectos encontrados no estado da arte ´e a utilização de circuitos integrados (chips) implantados no cérebro. Nesses chips, os sinais neurais medidos com os microeletrodos são amplificados, filtrados, processados e transmitidos a um computador externo mediante fios que atravessam o crânio. Existe o interesse em desenvolver chips implantáveis que transmitam os sinais ao computador externo sem a necessidade de fios que atravessem o crânio. Na pesquisa do estado da arte tem-se encontrado a utilização de tais chips implantáveis sem fio em ratos e macacos, porém até a data da elaboração deste texto não foram encontrados relatos da aplicação em humanos. Um dos aspectos que deve se levar em consideração no desenvolvimento de interfaces neurais implantáveis sem fio é a largura de banda do canal de comunicação. Quanto maior a quantidade de dados a serem transmitidos, maior a largura de banda necessária e maior o aquecimento do chip devido à dissipação de potência. Esta tese aborda sistemas de processamento de sinais neurais extracelulares que tem como objetivo reduzir a quantidade de dados a serem transmitidos e assim viabilizar a transmissão sem fio. Para poder ser integrados dentro do chip implantável, esses sistemas de processamento devem estar otimizados em termos de área e consumo de potência. Dois processamentos encontrados na pesquisa de interfaces neurais implantáveis são a detecção de impulsos elétricos e a separação de impulsos elétricos (Spike Sorting). Nesta tese apresentam-se soluções para esses tipos de processamentos visando a implementação mediante tecnologia CMOS (Complementary Metal Oxide Semiconductor). Para o caso da detecção de impulsos elétricos (spikes), nesta tese apresenta-se uma alternativa de implementação em hardware de um operador matemático conhecido como operador não linear de energia (NEO do inglês Nonlinear Energy Operator) ou operador Teager. Através da aplicação desse operador a um sinal neural evidencia-se a presença de spikes e atenua-se o ruído. Uma das características inovadoras da implementação apresentada nesta tese é a utilização de um circuito elevador ao quadrado que consiste de apenas três transistores, como bloco funcional básico para a realização da operação NEO. O circuito NEO desenvolvido consome 300 pJ no processamento de um spike e foi caracterizado por simulação até em 30 kHz, frequência que é compatível com as taxas de amostragem encontradas na literatura. O outro processamento abordado nesta tese, conhecido como separação de impulsos elétricos ou Spike Sorting, consiste no agrupamento dos impulsos elétricos registrados por um eletrodo em categorias, de maneira que em uma categoria estejam os impulsos gerados por um mesmo neurônio. Em outras palavras, o objetivo é reconhecer quais dos impulsos elétricos medidos pelo eletrodo pertencem a um mesmo neurônio, sendo possível que vários neurônios influenciem na medida realizada por um único eletrodo. Uma solução para a separação de impulsos, apropriada no contexto de sistemas implantáveis, é o template matching. Essa técnica baseia-se na geração de modelos (templates) durante uma fase inicial ao final da qual o número de templates gerados corresponde ao número de neurônios identificados pelo eletrodo. Numa fase seguinte, o sistema associa cada impulso elétrico detectado a um dos modelos inicialmente gerados. Nesta tese propõe-se um sistema de classificação que executa essa segunda fase do processo de spike sorting. Nesta tese apresenta-se o projeto de um sistema de classificação de spikes baseado na t écnica template matching, implementado com tecnologia CMOS. A implementação proposta nesta tese baseia-se na representação de amostras analógicas mediante o tempo. Esse tipo de representação de sinais analógicos mediante atrasos de pulsos digitais está sendo muito utilizado como alternativa à representação no domínio da tensão, da corrente ou da carga elétrica. A vantagem desse tipo de representação é que não se vê severamente afetada pela redução da tensão de alimentação dos circuitos integrados fabricados em tecnologias submicrométricas. A taxa de acerto na classificação do sistema desenvolvido é maior que 99% inclusive considerando um offset de até 20mV no comparador de saída. Os circuitos apresentados neste trabalho foram projetados considerando dispositivos da tecnologia TSMC de 90nm.
Neural signals recording through implantable microelectrode arrays in cortex extracellular medium has become an experimental paradigm for neuroscience. Moreover, recent research about motor neuroprostheses has shown that it is possible to decode motor commands from the signals recorded in the cerebral cortex extracellular medium. In both situations, experimental neuroscience and motor neuroprostheses development, one of the issues encountered in the state-of-the-art is the use of integrated circuits (chips) implanted in the brain. In these chips, neural signals measured with microelectrodes are amplified, filtered, processed, and transmitted to an external computer through wires that run through the skull. There is interest in developing implantable chips that transmit signals to the external computer without the need for wires passing through the skull. In the survey of the state-of-the-art it has found the use of such implantable wireless chips in rats and monkeys, but until the date of this writing we have not found reports of application in humans. One of the aspects that must be taken into account in the development of wireless implantable neural interfaces is the bandwidth of the communication channel. The greater the amount of data to be transmitted, the greater the bandwidth required and higher chip heating due to power dissipation. This thesis deals with extracellular neural signals processing systems that aim to reduce the amount of data to be transmitted and in this way to enable wireless transmission. In order to integrate them into an implantable chip, those processing systems must be optimized in terms of area and power consumption. Two processes found in the research of implantable neural interfaces are spike detection and spike sorting. In this thesis solutions for these types of processing are presented considering their implementation by CMOS (Complementary Metal Oxide Semiconductor). For the case of spike detection in this thesis it is presented an alternative for the hardware implementation of a mathematical operator known as NEO (Nonlinear Energy Operator). Through the application of this operator to a neural signal the presence of spikes becomes evident and the noise is attenuated. One of the innovative characteristics of the implementation presented in this thesis is the use of a squarer circuit which consists of only three transistors, as a basic function block for performing operation of NEO. NEO circuit consumes 300 pJ in processing a spike, and was characterized by simulation up to 30 kHz, frequency which is compatible with sampling rates found in the literature. The other processing discussed in this thesis, known as Spike Sorting, is the grouping of electrical impulses recorded by an electrode into categories so that the spikes belonging to the same category were generated by a single neuron. In other words, the goal is to recognize which of the spikes measured by the electrode belong to the same neuron, given that it is possible that several neurons influence the measure performed by a single electrode. A solution for the Spike Sorting suitable in the context of implantable systems, is the template matching. This technique is based on generating templates during an initial phase at the end of which the number of generated templates corresponds to the number of neurons identified by the electrode. In the next phase, the system associates each detected spike to one of the templates generated initially. In this thesis it is proposed a classification systems which performs that second phase of the spike sorting process. This thesis presents the design of a spike classification system based on template matching technique, implemented in CMOS technology. The processing proposed in this work is based on the time-based representation of the analog samples. This kind of representation of analog signals by delays of digital pulses is being widely used as an alternative to the classical representation of samples by voltage, current or electric charge. The advantage of this time-mode representation is that it is not severely affected by reduced supply voltage of integrated circuits manufactured in sub-micrometer technologies. The classification hit rate of the developed system is greater than 99% even when an offset of 20 mV is assumed for the output comparator. All the circuits presented in this work were designed using devices from TSMC 90nm technology.

Gesture retrieval can be defined as the process of retrieving the correct meaning of the hand movement from a pre-assembled gesture dataset. The purpose of the research discussed here is to design and implement a gesture interface system that facilitates retrieval for an American Sign Language gesture set using a mobile device. The principal challenge discussed here will be the normalization of 2D gestures generated from the mobile device interface and the 3D gestures captured from video samples into a common data structure that can be utilized by deep learning networks. This thesis covers convolutional neural networks and auto encoders which are used to transform 2D gestures into the correct form, before being classified by a convolutional neural network. The architecture and implementation of the front-end and back-end systems and each of their respective responsibilities are discussed. Lastly, this thesis covers the results of the experiment and breakdown the final classification accuracy of 83% and how this work could be further improved by using depth based videos for the 3D data.

Numerous applications in neuroscience research and neural prosthetics, such as retinal prostheses, spinal-cord surface stimulation for prosthetics, electrocorticogram (ECoG) recording for epilepsy detection, etc., involve electrical interaction with soft excitable tissues using a surface stimulation and/or recording approach. These applications require an interface that is able to set up electrical communications with a high throughput between electronics and the excitable tissue and that can dynamically conform to the shape of the soft tissue. Being a compliant and biocompatible material with mechanical impedance close to that of soft tissues, polydimethylsiloxane (PDMS) offers excellent potential as the substrate material for such neural interfaces. However, fabrication of electrical functionalities on PDMS has long been very challenging. This thesis work has successfully overcome many challenges associated with PDMS-based microfabrication and achieved an integrated technology platform for PDMS-based stretchable microelectrode arrays (sMEAs). This platform features a set of technological advances: (1) we have fabricated uniform current density profile microelectrodes as small as 10 microns in diameter; (2) we have patterned high-resolution (feature as small as 10 microns), high-density (pitch as small as 20 microns) thin-film gold interconnects on PDMS substrate; (3) we have developed a multilayer wiring interconnect technology within the PDMS substrate to further boost the achievable integration density of such sMEA; and (4) we have invented a bonding technology---via-bonding---to facilitate high-resolution, high-density integration of the sMEA with integrated circuits (ICs) to form a compact implant. Taken together, this platform provides a high-resolution, high-density integrated system solution for neural and muscular surface interfacing. sMEAs of example designs are evaluated through in vitro and in vivo experimentations on their biocompatibility, surface conformability, and surface recording/stimulation capabilities, with a focus on epimysial (i.e. on the surface of muscle) applications. Finally, as an example medical application, we investigate a prosthesis for unilateral vocal cord paralysis (UVCP) based on simultaneous multichannel epimysial recording and stimulation.

Orientador: Prof. Dr. Ricardo Suyama
Dissertação (mestrado) - Universidade Federal do ABC, Programa de Pós-Graduação em Engenharia da Informação, Santo André, 2018.
Uma interface cerebro-computador (BCI) e projetada para que se consiga, de modo efetivo, fornecer uma via alternativa de comunicacao entre o cerebro do usuario e o computador. Sinais captados por meio de eletrodos, tipicamente posicionados no escalpo do individuo, sao previamente processados para que haja eliminacao de ruidos externos. A partir dai, diversas tecnicas para processamento de sinais sao utilizadas para posteriormente classificar os sinais registrados e realizar a traducao do estado mental do usuario em um comando especifico a ser executado pelo computador. No presente trabalho sao utilizadas tecnicas de representacao esparsa dos sinais para a extracao de caracteristicas relevantes para classificacao dos mesmos, com intuito de aumentar a robustez e melhorar o desempenho do sistema. Para a extracao de sinais esparsos, foram utilizados algoritmos de criacao de dicionarios, a partir dos quais e possivel obter uma representacao esparsa para todo o subespaco de sinal. No trabalho foram utilizados 5 diferentes algoritmos de criacao de dicionario: Metodo de direcoes otimas (MOD), K-SVD, RLS-DLA, LS-DLA e Aprendizado de dicionario Online (ODL). A classificacao dos sinais foi realizada com o metodo de .. vizinhos mais proximos (k - NN). Os resultados obtidos com a abordagem de representacao esparsa foram comparados com os resultados do BCI Competition IV dataset 2a. Para o primeiro colocado da competicao foi obtido, em termos do coeficiente kappa, uma acuracia de 0.57 enquanto que no trabalho utilizando os metodos esparsos, obteve-se, em coeficiente kappa, uma acuracia de 0.90. Em comparacao obteve-se um ganho de 0.33 de acuracia, onde se deduz que o uso de sinais esparsos pode ser benefico para o dificil problema de se projetar uma interface cerebro computador.
A brain computer interface (BCI) is designed to effectively translate commands thought by human individuals into commands that a computer can effectively understand. Electrical impulses generated from the brain sculp are recorded from a device called an electroencephalograph and are preprocessed for elimination of external noise. From there, several techniques for signal processing are used to later classify the signals obtained by the electroencephalograph. In this work, techniques for sparse representation of signals are used for feature extraction, in order to increase robustness and system performance. For the extraction of sparse signals, five different dictionary learning algorithms were used, being able to produce a basis capable of represensing the entire signal subspace. In this work, 5 different dictionary learning algorithms were used: Method of Optimal Directions (MOD), K-SVD, Recursive Least Square Dictionary Learning (RLS-DLA), Least Square Dictionary Learning (LS-DLA) and Online Dictionary Learning (ODL). For the classification task, the k-NN method was used. The simulation results obtained with this approach were compared with the best BCI Competition IV dataset 2a results. For the first place in the competition, an accuracy of 0.57 was obtained, in terms of the kappa coefficient, whereas in the work using the sparse methods, a kappa coefficient of 0.90 was obtainned, improving accuracy in 0.33 accuracy was obtained, which indicates that the use of sparse signals may be beneficial to the difficult problem of designing a brain computer interface.

This paper proposes a new model for studying the new product development process in an artificial environment. We show how connectionist models can be used to simulate the adaptive nature of agents' learning exhibiting similar behavior as practically experienced learning curves. We study the impact of incentive schemes (local, hybrid and global) on the new product development process for different types of organizations. Sequential organizational structures are compared to two different types of team-based organizations, incorporating methods of Quality Function Deployment such as the House of Quality. A key finding of this analysis is that the firms' organizational structure and agents' incentive system significantly interact. We show that the House of Quality is less affected by the incentive scheme than firms using a Trial & Error approach. This becomes an important factor for new product success when the agents' performance measures are conflicting. (author's abstract)
Series: Report Series SFB "Adaptive Information Systems and Modelling in Economics and Management Science"

Thesis (MEng)--Stellenbosch University, 2014.
ENGLISH ABSTRACT: Muscular dystrophy, spinal cord injury, or amyotrophic lateral sclerosis (ALS) are injuries and disorders that disrupts the neuromuscular channels of the human body thus prohibiting the brain from controlling the body. Brain computer interface (BCI) allows individuals to bypass the neuromuscular channels and interact with the environment using the brain. The system relies on the user manipulating his neural activity in order to control an external device. Electroencephalography (EEG) is a cheap, non-invasive, real time acquisition device used in BCI applications to record neural activity. However, noise, known as artifacts, can contaminate the recording, thus distorting the true neural activity. Eye blinks are a common source of artifacts present in EEG recordings. Due to its large amplitude it greatly distorts the EEG data making it difficult to interpret data for BCI applications. This study proposes a new combination of techniques to detect and correct eye blink artifacts to improve the quality of EEG for BCI applications. Independent component analysis (ICA) is used to separate the EEG signals into independent source components. The source component containing eye blink artifacts are corrected by detecting each eye blink within the source component and using a trained wavelet neural network (WNN) to correct only a segment of the source component containing the eye blink artifact. Afterwards, the EEG is reconstructed without distorting or removing the source component. The results show a 91.1% detection rate and a 97.9% correction rate for all detected eye blinks. Furthermore for channels located over the frontal lobe, eye blink artifacts are corrected preserving the neural activity. The novel combination overall reduces EEG information lost, when compared to existing literature, and is a step towards improving EEG pre-processing in order to provide cleaner EEG data for BCI applications.
AFRIKAANSE OPSOMMING: Spierdistrofie, ’n rugmurgbesering, of amiotrofiese laterale sklerose (ALS) is beserings en steurnisse wat die neuromuskulêre kanale van die menslike liggaam ontwrig en dus verhoed dat die brein die liggaam beheer. ’n Breinrekenaarkoppelvlak laat toe dat die neuromuskulêre kanale omlei word en op die omgewing reageer deur die brein. Die BCI-stelsel vertrou op die gebruiker wat sy eie senuwee-aktiwiteit manipuleer om sodoende ’n eksterne toestel te beheer. Elektro-enkefalografie (EEG) is ’n goedkoop, nie-indringende, intydse dataverkrygingstoestel wat gebruik word in BCI toepassings. Nie net senuwee aktiwiteit nie, maar ook geraas , bekend as artefakte word opgeneem, wat dus die ware senuwee aktiwiteit versteur. Oogknip artefakte is een van die algemene artefakte wat teenwoordig is in EEG opnames. Die groot omvang van hierdie artefakte verwring die EEG data wat dit moeilik maak om die data te ontleed vir BCI toepassings. Die studie stel ’n nuwe kombinasie tegnieke voor wat oogknip artefakte waarneem en regstel om sodoende die kwaliteit van ’n EEG vir BCI toepassings te verbeter. Onafhanklike onderdeel analise (Independent component analysis (ICA)) word gebruik om die EEG seine te skei na onafhanklike bron-komponente. Die bronkomponent wat oogknip artefakte bevat word reggestel binne die komponent en gebruik ’n ervare/geoefende golfsenuwee-netwerk om slegs ’n deel van die komponent wat die oogknip artefak bevat reg te stel. Daarna word die EEG hervorm sonder verwringing of om die bron-komponent te verwyder. Die resultate toon ’n 91.1% opsporingskoers en ’n 97.9% regstellingskoers vir alle waarneembare oogknippe. Oogknip artefakte in kanale op die voorste lob word reggestel en behou die senuwee aktiwiteit wat die oorhoofse EEG kwaliteit vir BCI toepassings verhoog.

This thesis presents a framework for an intelligent interface for the access of robotsupported remote laboratories through the Internet. The framework is composed of the student assessment and coaching system, the experimentation scenario, and the associated graphical user interface. Student assessment and coaching system is the main feature of a successful intelligent interface for use during remote experimentation with a robot-supported laboratory setup. The system has a modular structure employing artificial neural networks and a fuzzy-rule based decision process to model the student behaviour, to evaluate the performance and to coach him or her towards a better achievement of the tasks to be done during the experimentation. With an experimentation scenario designed and a graphical user interface, the system is applied to a robotic system that is connected to the Internet for the evaluation of the proposed framework. Illustrative examples for the operation of the each module in the system in the context of the application are given and sensitivity analysis of the system to the change in parameters is also done. The framework is then applied to a mobile robot control laboratory. The user interface and the experimentation scenario is developed for the application, and necessary modifications are made to the student assessment and coaching system in order to support the experiment.

La tecnologia d’implants neuronals en medicina té com a objectiu restaurar la funcionalitat del sistema nerviós en casos de degeneració o dany greu registrant o estimulant l’activitat elèctrica del teixit nerviós. Els implants neuronals disponibles actualment ofereixen una eficàcia clínica modesta, en part a causa de les limitacions que tenen els metalls utilitzats en la interfície elèctrica amb el teixit. Aquests materials comprometen la resolució de la interfície i, per tant, la restauració funcional amb el rendiment i l’estabilitat. En aquest treball presento uns implants neuronals flexibles basats en una pel·lícula prima de grafè porós nanoestructurat i biocompatible que proporciona una interfície neural bidireccional estable i d’alt rendiment. En comparació amb els dispositius de microelectrodos de platí estàndard, elèctrodes de 25 μm de diàmetre basats en grafè ofereixen una impedància significativament menor i poden injectar de manera segura 200 vegades més càrrega durant més de 100 milions de polsos. N’evaluo les seves capacitats in vivo registrant activitat epicortical amb alta fidelitat i alta resolució, estimulant subconjunts d’axons dins del nervi ciàtic amb llindars de corrent baixos i alta selectivitat i modulant l’activitat de la retina amb alta precisió. La tecnologia de pel·lícula fina de grafè aquí descrita té el potencial de convertir-se en el nou punt de referència per la pròxima generació de tecnologia d’implants neuronals.
La tecnología de implantes neuronales en medicina tiene como objetivo restaurar la funcionalidad del sistema nervioso en casos de degeneración o daño grave registrando o estimulando la actividad eléctrica del tejido nervioso. Los implantes neurales disponibles actualmente ofrecen una eficacia clínica modesta, en parte debido a las limitaciones que plantean los metales utilizados en la interfaz eléctrica con el tejido. Dichos materiales comprometen la resolución de la interfaz y, por lo tanto, la restauración funcional con el rendimiento y la estabilidad. En este trabajo presento unos implantes neuronales flexibles basados en una película delgada de grafeno poroso nanoestructurado y biocompatible que proporciona una interfaz neural bidireccional estable y de alto rendimiento. En comparación con los dispositivos de microelectrodos de platino estándar, electrodos de 25 μm de diámetro basados en grafeno ofrecen una impedancia significativamente menor y pueden inyectar de forma segura 200 veces más carga durante más de 100 millones de pulsos. Aquí evaluo sus capacidades in vivo registrando actividad epicortical con alta fidelidad y alta resolución, estimulando subconjuntos de axones dentro del nervio ciático con umbrales de corriente bajos y alta selectividad y modulando la actividad de la retina con alta precisión. La tecnología de película fina de grafeno aquí descrita tiene el potencial de convertirse en el nuevo punto de referencia para la próxima generación de tecnología de implantes neuronales.
Neural implants technology in medicine aims to restore nervous system functionality in cases of severe degeneration or damage by recording or stimulating the electrical activity of the nervous tissue. Currently available neural implants offer a modest clinical efficacy partly due to the limitations posed by the metals used at the electrical interface with the tissue. Such materials compromise interfacing resolution, and therefore functional restoration, with performance and stability. In this work, I present flexible neural implants based on a biocompatible nanostructured porous graphene thin film that provides a stable and high performance bidirectional neural interface. Compared to standard platinum microelectrode devices, the graphene-based electrodes of 25 μm diameter offer significantly lower impedance and can safely inject 200 times more charge for more than 100 million pulses. I assessed their performance in vivo by recording high fidelity and high resolution epicortical activity, by stimulating subsets of axons within the sciatic nerve with low thresholds and high selectivity and by modulating the retinal activity with high precision. The graphene thin film technology I describe here has the potential to become the new performance benchmark for the next generation of neural implant technology.
Universitat Autònoma de Barcelona. Programa de Doctorat en Enginyeria Electrònica i de Telecomunicació

This work proposes a series of driver assistance technologies that enable blind persons to safely and independently operate an automobile on standard public roads. Such technology could additionally benefit sighted drivers by augmenting vision with suggestive cues during normal and low-visibility driving conditions. This work presents a non-visual human-computer interface system with passive and adaptive controlling software to realize this type of driver assistance technology. The research and development behind this work was made possible through the Blind Driver Challenge® initiative taken by the National Federation of the Blind. The instructional technologies proposed in this work enable blind drivers to operate an automobile through the provision of steering wheel angle and speed cues to the driver in a non-visual method. This paradigm imposes four principal functionality requirements: Perception, Motion Planning, Reference Transformations, and Communication. The Reference Transformation and Communication requirements are the focus of this work and convert motion planning trajectories into a series of non-visual stimuli that can be communicated to the human driver. This work proposes two separate algorithms to perform the necessary reference transformations described above. The first algorithm, called the Passive Non-Visual Interface Driver, converts the planned trajectory data into a form that can be understood and reliably interacted with by the blind driver. This passive algorithm performs the transformations through a method that is independent of the driver. The second algorithm, called the Adaptive Non-Visual Interface Driver, performs similar trajectory data conversions through methods that adapt to each particular driver. This algorithm uses Model Predictive Control supplemented with Artificial Neural Network driver models to generate non-visual stimuli that are predicted to induce optimal performance from the driver. The driver models are trained online and in real-time with a rapid training approach to continually adapt to changes in the driver's dynamics over time. The communication of calculated non-visual stimuli is subsequently performed through a Non-Visual Interface System proposed by this work. This system is comprised of two non-visual human computer interfaces that communicate driving information through haptic stimuli. The DriveGrip interface is pair of vibro-tactile gloves that communicate steering information through the driverâ s hands and fingers. The SpeedStrip interface is a vibro-tactile cushion fitted on the driverâ s seat that communicates speed information through the driver's legs and back. The two interfaces work simultaneously to provide a continuous stream of directions to the driver as he or she navigates the vehicle.
Ph. D.