Дисертації з теми "Brain encoding"

The studies in this thesis explored how pain and its relief are represented in the human brain. Pain and relief are important survival signals that motivate escape from danger and search for safety, however, they are often evaluated by subjective descriptions only. Studying how humans learn and adapt to pain and relief allows objective investigation of the information processing and neural circuitry underlying these internal experiences. My research set out to use computational learning models to provide mechanistic explanations for the behavioural and functional neuroimaging data collected in pain/relief learning experiments with independent groups of healthy human participants. With a Pavlovian acute pain conditioning task in Experiment 1, I found that 'associability' (a form of uncertainty signal) had a crucial role in controlling the learning rates of different conditioned responses, and can be used to anatomically dissociate underlying neural systems. Experiment 2 focused on relief learning of terminating a tonic pain stimulus, in which the priority for relief-seeking is in conflict with the general suppression of cognition and attention. I showed that associability during active learning not only controls the relief learning rate, but also correlates with endogenously modulated (reduced) ongoing pain. This finding was confirmed in Experiment 3 using an independent active relief learning paradigm in a complex dynamic environment. Critically, both experiments showed that associability was correlated with responses in the pregenual anterior cingulate cortex (pgACC), a brain region previously implicated in aspects of endogenous pain control related to attention and controllability. This provided a potential computational account of an information-sensitive endogenous analgesic mechanism. In Experiment 4, I explored the implications of endogenous controllability for technology-based pain therapeutics. I designed an adaptive closed-loop system that learned to control pain stimulation using decoded real-time pain representations from the brain. Subjects were shown to actively enhance the discriminability of pain only in the pgACC, and uncertainty during learning again correlated with endogenously modulated pain and were associated with pgACC responses. Together, these studies (i) show the importance of uncertainty in controlling learning during both acute and tonic pain, (ii) describe how uncertainty also flexibly modulates pain to maximise the impact of learning, (iii) illustrate a central role for the pgACC in this process, and (iv) reveal the implications for future technology-based therapeutic systems.

It has been shown that brain activity before an item can predict whether this item will later be remembered. However, the cognitive mechanisms underlying this so-called prestimulus brain activity are poorly understood. The studies in this PhD thesis addressed the role of prestimulus neural activity in long-term memory encoding and whether this activity is under voluntary control. To allow better dissociation between brain activity before and after an item, electroencephalography (EEG) was used due to its high temporal resolution. In a series of studies EEG data were analyzed in terms of Event-Related Potentials (ERPs) and oscillatory power in the theta frequency band (4-8 Hz) that plays a crucial role in memory processes. The findings demonstrate that brain activity preceding a stimulus is indeed under a person‟s control. In one experiment, ERP and frontal theta prestimulus activity before an item was only evident when participants were highly motivated to encode an upcoming item. In another experiment, ERP prestimulus activity only emerged when participants prioritized encoding over a concurrent task. These studies suggest that, at least, some prestimulus activities reflect preparatory processes that depend on the available cognitive resources. Two further experiments demonstrated that frontal prestimulus encoding-related theta power is specific to semantic encoding conditions. Finally, a series of behavioural experiments showed that memory performance does not differ depending on the opportunity to prepare during encoding. The findings of my PhD thesis suggest that (i) some prestimulus signals (i.e. frontal theta) reflect a preparatory process ahead of semantic encoding, (ii) and, most importantly, prestimulus signals (i.e. ERPs and frontal theta) reflect active preparatory processes for long-term memory formation. The results of this thesis could lead to the development of new strategies of how to improve memory, especially in clinical settings.

Acoustic regularity encoding has been associated with a decrease of the neural response to repeated stimulation underlying the representation of auditory objects in the brain. The present thesis encloses two studies that sought to assess the neural correlates of acoustic regularity encoding in the human auditory system, by means of analyzing auditory evoked potentials. Study I was conducted at the Cognitive Neuroscience Research Group, at the Faculty of Psychology of the University of Barcelona (Barcelona, Catalonia, Spain), under the direct supervision of Dr. Carles Escera. This study aimed to explore the dynamics of adaptation of the auditory evoked potentials to probabilistic stimuli embedded in a complex sequence of sounds. The main outcome of this study was the demonstration that the amplitude of auditory evoked potentials adapts to the complex history of stimulation with different time constants concurrently: it adapts faster to local and slower to global probabilities of stimulation. This study also showed that auditory evoked potential amplitudes correlate with stimulus expectancy as defined by a combination of local and global stimulus probabilities. Study II was conducted at the Institute of Child Health (ICH), at the University College of London (UCL; London, United Kingdom), under the direct supervision of Dr. Torsten Baldeweg. This study aimed to explore the influence of timing predictability in the neural adaptation to probabilistic stimuli. The main outcome of this study was the demonstration that timing predictability enhances the repetition-related modulation of the auditory evoked potentials amplitude, being essential for repetition effects at early stages of the auditory processing hierarchy.
La codificació de regularitats acústiques està associada amb la reducció de la resposta neuronal a l’estimulació repetida, essent la base de la representació dels objectes auditius al cervell. La present tesi doctoral inclou dos estudis que exploren els correlats neuronals de la codificació de regularitats acústiques al sistema auditiu humà, mitjançant l’anàlisi dels potencials evocats auditius. L’objectiu del primer estudi, realitzat al Grup de Recerca en Neurociència Cognitiva de la Facultat de Psicologia de la Universitat de Barcelona (UB) i sota la supervisió directa del Dr. Carles Escera, va ser el d’explorar les dinàmiques d’adaptació dels potencials evocats auditius a estímuls probabilístics en una complexa seqüència de sons. El resultat principal d’aquest estudi va ser la demostració de que l’amplitud dels potencials evocats auditius s’adapta a la historia complexa d’estimulació amb diferents constants temporals simultàniament: s’adapta més ràpidament a probabilitats d’estimulació locals que globals. Aquest estudi també va mostrar que l’amplitud dels potencials evocats auditius correlaciona amb l’expectància d’un estímul definida com a una combinació de probabilitats locals i globals d’estimulació. L’objectiu del segon estudi, realitzat al Institute of Child Health (ICH), de l’University College of London (UCL), sota la supervision directa del Dr. Torsten Baldeweg, va ser el d’explorar la influència de la predictabilitat temporal en l’adaptació de l’activitat neuronal a estímuls probabilístics. El resultat principal d’aquest estudi va ser la demostració que la predictabilitat temporal intensifica la modulació de l’amplitud dels potencials evocats auditius a la repetició dels estímuls, essent esencial pels efectes que la repetició exerceix en etapes primerenques de la jerarquía de processament auditiu.

El nostre sistema auditiu codifica regularitats acústiques i les compara amb els inputs sensorials de forma continuada. Els sons novedosos o canvis acústics són detectats inconscientment de forma ràpida i automàtica, permetent així la mobilització de recursos atencionals i un ajustament adequat de la nostra conducta. La tesi aquí present inclou tres estudis que utilitzen Magnetoencefalografia i la localització de camps evocats auditius generats en un paradigma oddball per tal d'investigar els correlats neuronals de la detecció de novetat i la codificació de regularitats en estadis primerencs del sistema auditiu humà. El primer estudi, portat a terme al Grup de Recerca en Neurociència Cognitiva (Universitat de Barcelona), mostra diferents generadors neuronal involucrats en la codificació de son novedodos en finestres temporals primerenques i tardanes; respectivament indicades per les Respostes de Latència Mitja (MLR) generades entre els 20 i 50 milisegons després del inici de un so, i el posterior Potencial de Disparitat (MMN), evocat entre els 100 and 250 milisegons. El segon estudi, portat a terme a l'Institut für Biomagnetismus & Biosignalanalys (Universitat de Münster), mostra que característiques acústiques novedoses formades per nivells de complexitat diferents són processades en intervals de temps diferents i en àrees neuronals separades, suggerint per tant una organització jeràrquica de la detecció de novetat i codificació de regularitats. El tercer estudi, portat a terme al Grup de Recerca en Neurociència Cognitiva emprant un paradigma roving-standard, mostra que tant la supressió com l'increment neuronal relacionats amb la repetició de un estímul estan involucrats en el procés de formació de traces de memòria, i que els generadors neuronals relacionats amb aquest procés estan localitzats tant en regions auditives com regions no auditives d'alt ordre. En resum, els resultats d'aquesta tesi doctoral suggereixen que la percepció auditiva es basa en un sistema sensorial organitzat jeràrquicament que té com a objectiu la predicció d'esdeveniments futurs partint de la predicció de regularitats prèviament codificades.
Our auditory system is continuously encoding acoustic regularities and comparing them with incoming sensory inputs. Novel sounds or acoustic changes must be detected fast in an automatic and unconscious fashion, thus allowing for the reallocation of attentional resources and the proper adjustment of our behaviour. The present thesis encloses three studies that employ Magnetoencephalography and source localization of auditory evoked fields as generated in oddball paradigms to assess the neural correlates of deviance detection and regularity encoding in early stages of the human auditory system. The first study, conducted at the Cognitive Neuroscience Research Group (University of Barcelona), shows distinct neuronal generators involved in the encoding of novel sounds in early and late time intervals; as respectively indexed by Middle Latency-Responses (MLR) evoked between 20 and 50 milliseconds after sound onset, and the later Mismatch component (MMN) generated between 100 and 250 milliseconds. The second study, conducted at the Institut fur Biomagnetismus & Biosignalanalys (University of Munster), shows that deviant acoustic features involving different levels of complexity are processed in distinct time ranges and generated in separated neuronal sources, thus suggesting a hierarchical organization of deviance detection and regularity encoding. The third study, conducted in the Cognitive Neuroscience Research Group using a roving-standard paradigm, indicates that neural repetition-related suppression and repetition enhancement underlie auditory memory trace formation, and that neural generators involved in this process are located in both auditory and non-auditory high-order regions. In sum, results from this thesis suggest that auditory perception is based on a hierarchically organized sensory system whose goal is to predict future events on the basis of previously encoded regularities.

Due to a steady increase in the number of babies born prematurely over the past 20 years, prematurity (a birth occurring before 37 weeks gestation) has emerged as an important public health concern. Even with improved survival of these infants, they remain at risk for many unfavorable health outcomes. Most of those risks include cognitive and behavioral deficits that show up later in life, highlighting the importance of studying the development of the brain, in particular. The current study investigates brain development outcomes in the first years of life using: (1) structural magnetic resonance imaging (MRI) to study brain structure, and (2) three novel cognitive assessments of visual working memory, attention, and speed of processing information. Healthy 12-month-old infants were recruited through University of Iowa’s Neonatal Admissions Registry. An MRI imaging acquisition protocol was developed in order to scan infants during their naptime without sedation. Additionally, a new automatic approach to classifying areas of the brain was developed at the University of Iowa Department of Radiology for 12-month-old brain images. These novel cognitive assessments are based on infant eye movements (including how long it takes for an infant to react to certain stimuli and the direction of their looking). Results from this study support the use of these cognitive tasks to detect specific functional changes in performance based on gestational age. Therefore, these tasks may be potential early markers of risk in preterm populations, but continued investigations are necessary to fully elucidate early brain outcomes during this critical period of development.

Understanding the mnemonic functions of the brain has been extensively facilitated by the development of functional neuroimaging techniques such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI). The present thesis aims at investigating the neural mechanisms underlying memory for personally experienced events (episodic memory), using PET. In paper I, similarities between encoding and retrieval of enacted (motor) information were explored. We observed increased retrieval activation in right premotor areas in the brain when sentences encoded by motor enactment and sentences encoded by maintenance rehearsal were contrasted. In paper II, overlap between encoding and retrieval was explicitly tested for three types of event information: spatial, item, and temporal. Using conjunction analyses, we found that encoding and retrieval of spatial information was associated with increased brain activity in bilateral inferior parietal regions. Encoding and retrieval of item information were related to increased activation in right inferior temporal cortex, and encoding and retrieval of temporal information were associated with increased activation in left inferior temporal and left inferior frontal cortex. In paper III, brain activity associated with retrieval success was examined. Conditions included three levels of retrieval success (high, medium, and low level), for two types of information (pictures and sentences). The results showed a pattern of activation that distinguished between brain regions involved in processing of sentences vs. processing of pictures. A second pattern that distinguished between brain regions involved in encoding vs. retrieval processes, irrespectively of material (sentences and pictures) and retrieval success, was also found. The manipulation of retrieval success was associated with systematic changes in the correlation between material specific regions and other areas of the brain. In study IV, changes in activation related to successful retrieval of pictures were investigated. More specifically, we expected to find decreases in infero-temporal (IT) regions of the brain that were associated with successful recognition memory. As expected, we found a region in left IT cortex that showed decreased activation related to memory for event information. This decrease in activation could be dissociated from responses related to novelty detection, and perceptual priming. The results from study I and II are discussed in relation to findings and theories regarding similarities between encoding and retrieval processes, and reactivation of modality-specific brain areas important for memory storage. The results from studies III and IV are discussed in relation to differences between encoding and retrieval processes, e.g. asymmetric frontal activation and sub-processes of episodic memory, such as retrieval mode, retrieval success, and novelty detection. Taken together, the studies show that different episodic memory processes are correlated with distinct brain areas, hence supporting the view that remembering is based on multiple component processes.
digitalisering@umu.se

Neurons in human somatosensory cortex are somatotopically organized, with sensation from the lower limbs mediated by neurons near the midline of the brain, whereas sensations from the upper body, hands and orofacial surfaces are mediated by neurons located more laterally in a sequential map. Neurons in Brodmann's area (BA) 3b are exquisitely sensitive to tactile stimulation of these skin surfaces. Moreover, the location, velocity and direction of tactile stimuli on the skin's surface are discriminable features of somatosensory processing, however their role in fine motor control and passive detection are poorly understood in health, and as a neurotherapeutic agent in sensorimotor rehabilitation. To better understand the representation and processing of dynamic saltatory tactile arrays in the human somatosensory cortex, high resolution functional magnetic resonance (fMRI) is utilized to delineate neural networks involved in processing these complex somatosensory events to the glabrous surface of the hand.

The principal goal of this dissertation is to map the relation between a dynamic saltatory pneumatic stimulus array delivered at 3 different velocities on the glabrous hand and the evoked blood-oxygen level-dependent (BOLD) brain response, hypothesized to involve a network consisting of primary and secondary somatosensory cortices (S1 and S2), insular cortex, posterior parietal cortex (PPC), and cerebellar nuclei. A random-balanced block design with fMRI will be used to record the BOLD response in healthy right-handed adults. Development of precise stimulus velocities, rapid rise-fall transitions, salient amplitude, is expected to optimize the BOLD response.

Neuronal networks are at the base of information processing in the brain. They are series of interconnected neurons whose activation defines a recognizable linear pathway. The main goal of studying neural ensembles is to characterize the relationship between the stimulus and the individual or general neuronal responses and the relation amongst the electrical activities of neurons within the network, also understanding how topology and connectivity relates to their function. Many techniques have been developed to study these complex systems: single-cell approaches aim to investigate single neurons and their connections with a limited number of other nerve cells; on the opposite side, low-resolution large-scale approaches, such as functional MRI (Magnetic Resonance Imaging) or electroencephalography (EEG), record signal changes in the brain that are generated by large populations of cells. More recently, multisite recording techniques have been developed to overcome the limitations of previous approaches, allowing to record simultaneously from huge neuronal ensembles with high spatial resolution and in different brain regions, i.e. by using implantable semiconductor chips. Local Field Potentials (LFPs), the part of electrophysiological signals that has frequencies below 500 Hz, capture key integrative synaptic processes that cannot be measured by analyzing the spiking activity of few neurons alone. Several studies have used LFPs to investigate cortical network mechanisms involved in sensory processing, motor planning and higher cognitive processes, like memory and perception. LFPs are also promising signals for steering neuroprosthetic devices and for monitoring neural activity even in human beings, since they are more easily and stably recorded in chronic settings than neuronal spikes. In this work, LFP profiles recorded in the rat barrel cortex through high-resolution CMOS-based needle chips are presented and compared to those obtained by means of conventional Ag/AgCl electrodes inserted into glass micropipettes, which are widely used tools in electrophysiology. The rat barrel cortex is a well-known example of topographic mapping where each of the whiskers on the snout of the animal is mapped onto a specific cortical area, called a barrel. The barrel cortex contains the somatosensory representation of the whiskers and forms an early stage of cortical processing for tactile information, along with the trigeminal ganglion and the thalamus. It is an area of great importance for understanding how the cerebral cortex works, since the cortical columns that form the basic building blocks of the neocortex can be actually seen within the barrel. Moreover, the barrel cortex has served as a test-bed system for several new methodologies, partly because of its unique and instantly identifiable form, and partly because the species that have barrels, i.e. rodents, are the most commonly used laboratory mammal. The barrel cortex, the whiskers that activate it and the intervening neural pathways have been increasingly the subject of focus by a growing number of research groups for quite some time. Nowadays, studies (such this one) are directed not only at understanding the barrel cortex itself but also at investigating issues in related fields using the barrel cortex as a base model. In this study, LFP responses were evoked in the target barrel by repeatedly deflecting the corresponding whisker in a controlled fashion, by means of a specifically designed closed-loop piezoelectric bending system triggered by a custom LabView acquisition software. Evoked LFPs generated in the barrel cortex by many consecutive whiskers' stimulations show large variability in shapes and timings. Moreover, anesthetics can deeply affect the profile of evoked responses. This work presents preliminary results on the variability and the effect of commonly used anesthetics on these signals, by comparing the distributions of evoked responses recorded from rats anesthetized with tiletamine-xylazine, which mainly blocks the excitatory NMDA receptors, and urethane, which conversely affects both the excitatory and inhibitory system, in a complex and balanced way yet preserving the synaptic plasticity. Representative signal shape characteristics (e.g., latencies and amplitude of events) extracted from evoked responses acquired from different cortical layers are analyzed and discussed. Statistical distributions of these parameters are estimated for all the different depths, in order to assess the variability of LFPs generated by individual mechanical stimulations of single whiskers along the entire cortical column. Preliminary results showed a great variability in cortical responses, which varied both in latency and amplitude across layers. We found significant difference in the latency of the first principal peak of the responses: under tiletamine-xylazine anesthetic, the responses or events of the evoked LFPs occurred later than the ones recorded while urethane was administered. Furthermore, the distributions of this parameter in all cortical layers were narrower in case of urethane. This behavior should be attributed to the different effects of these two anesthetics on specific synaptic receptors and thus on the encoding and processing of the sensory input information along the cortical pathway. The role of the ongoing basal activity on the modulation of the evoked response was also investigated. To this aim, spontaneous activity was recorded in different cortical layers of the rat barrel cortex under the two types of anesthesia and analyzed in the statistical context of neuronal avalanches. A neuronal avalanche is a cascade of bursts of activity in neural networks, whose size distribution can be approximated by a power law. The event size distribution of neuronal avalanches in cortical networks has been reported to follow a power law of the type P(s)= s^-a, with exponent a close to 1.5, which represent a reflection of long-range spatial correlations in spontaneous neuronal activity. Since negative LFP peaks (nLFPs) originates from the sum of synchronized Action Potentials (AP) from neurons within the vicinity of the recording electrode, we wondered if it were possible to model single nLFPs recorded in the basal activity traces by means of only one electrode as the result of local neuronal avalanches, and thus we analyzed the size (i.e. the amplitude in uV) distribution of these peaks so as to identify a suitable power-law distribution that could describe also these single-electrode records. Finally, the results of the first ever measurements of evoked LFPs within an entire column of the barrel cortex obtained by means of the latest generation of CMOS-based implantable needles, having 256 recording sites arranged into two different array topologies (i.e. 16 x 16 or 4 x 64, pitches in the x- and y-direction of 15 um and 33 um respectively), are presented and discussed. A propagation dynamics of the LFP can be already recognized in these first cortical profiles. In the next future, the use of these semiconductor devices will help, among other things, to understand how degenerating syndromes like Parkinson or Alzheimer evolve, by coupling detected behaviors and symptoms of the disease to neuronal features. Implantable chips could then be used as 'electroceuticals', a newly coined term that describes one of the most promising branch of bioelectronic medicine: they could help in reverting the course of neurodegenerative diseases, by constituting the basis of neural prostheses that physically supports or even functionally trains impaired neuronal ensembles. High-resolution extraction and identification of neural signals will also help to develop complex brain-machine interfaces, which can allow intelligent prostheses to be finely controlled by their wearers and to provide sophisticated feedbacks to those who have lost part of their body or brain functions.
Le reti neuronali sono alla base della codifica dell'informazione cerebrale. L'obiettivo principale dello studio delle popolazioni neuronali è quello di caratterizzare la relazione tra uno stimolo e la risposta individuale o globale dei neuroni e di studiare il rapporto tra le varie attività elettriche dei neuroni appartenenti ad una particolare rete, comprendendo anche come la topologia e la connettività della rete neuronale influiscano sulla loro funzionalità. Fino ad oggi, molte tecniche sono state sviluppate per studiare questi sistemi complessi: studi a singola cellula mirano a studiare singoli neuroni e le loro connessioni con un numero limitato di altre cellule; sul lato opposto, approcci su larga scala e a bassa risoluzione, come la risonanza magnetica funzionale o l'elettroencefalogramma, registrano segnali elettrofisiologici generati nel cervello da vaste popolazioni di cellule. Più recentemente, sono state sviluppate tecniche di registrazione multisito che mirano ad abbattere le limitazioni dei precedenti approcci, rendendo possibile la misurazione ad alta risoluzione di segnali generati da grandi ensamble neuronali e da diverse regioni del cervello simultaneamente, ad esempio mediante l'uso di chip impiantabili a semiconduttore. I potenziali di campo locali (LFP) catturano processi sinaptici chiave che non possono essere estratti dall'attività di spiking di qualche neurone isolato. Numerosi studi hanno utilizzato gli LFP per studiare i meccanismi corticali coinvolti nei processi sensoriali, motori e cognitivi, come la memoria e la percezione. Gli LFP rappresentano anche dei segnali interessanti nell'ambito delle applicazioni neuroprotesiche e per monitorare l'attività cerebrale negli esseri umani, dal momento che possono essere registrati più stabilmente e facilmente in impianti cronici rispetto agli spike neuronali. In questo studio, sono riportati dei profili LFP registrati dalla barrel cortex di ratto tramite chip ad ago ad alta risoluzione basati su tecnologia CMOS e confrontati con quelli ottenuti tramite elettrodi convenzionali in Ag/AgCl inseriti in micropipette di vetro, strumenti comunemente usati in elettrofisiologia. La barrel cortex di ratto è un esempio ben noto di mapping topografico, nel quale ogni baffo sul muso dell'animale è mappato in una specifica area corticale, chiamata barrel. La barrel cortex contiene la rappresentazione sensoriale dei baffi dell'animale e rappresenta uno dei primi stadi di elaborazione dell'informazione tattile, insieme al ganglio del trigemino e al talamo. Essa è un'area di primaria importanza per lo studio del funzionamento della corteccia cerebrale, visto che le colonne corticali che formano i blocchi di base della neocorteccia possono essere visualizzati facilmente all'interno della barrel cortex. La barrel cortex inoltre è utilizzata come sistema di test in numerose metodologie innovative, grazie alla sua struttura unica ed istantaneamente identificabile, e grazie anche al fatto che le specie dotate di barrel, i roditori, sono gli animali da laboratorio più comuni. La barrel cortex e le sue interconnessioni neuronali sono stati oggetto delle ricerche più disparate in questi ultimi decenni. Attualmente, alcuni studi (come questo) non mirano solamente a comprendere meglio la barrel cortex, ma anche ad analizzare problematiche in campi scientifici collegati, utilizzando la barrel cortex come modello base. In questo lavoro, sono stati evocati segnali LFP nella barrel cortex tramite deflessioni ripetute dei baffi dell'animale, realizzate in modo controllato tramite un sistema di deflessione piezoelettrica a closed-loop innescato da un sistema di acquisizione LabView. Le risposte evocate generate nella barrel dalla stimolazione ripetuta dei baffi presentano elevata variabilità nella forma e nelle latenze temporali. Inoltre, il tipo di anestesia utilizzata può influenzare profondamente il profilo della risposta evocata. Questo studio riporta i risultati preliminari sulla variabilità della risposta neuronale e sull'effetto di due anestetici di uso comune su questi segnali, confrontando le distribuzioni delle risposte evocate in ratti anestetizzati con tiletamina-xylazina (il quale agisce prevalentemente sui recettori eccitatori di tipo NMDA) e uretano (che agisce in modo più bilanciato e complesso su entrambi i sistemi eccitatori ed inibitori, preservando la plasticità sinaptica). Sono state analizzate e discusse alcune caratteristiche rappresentative del segnale evocato (ad esempio, le latenze temporali e l'ampiezza degli eventi), registrato a varie profondità corticali. Per tutte le prondità corticali acquisite, sono state stimate le distribuzioni statistiche di tali parametri, in modo da valutare la variabilità degli LFP evocati dalle stimolazioni meccaniche individuali delle vibrisse del ratto lungo l'intera colonna corticale. I primi risultati presentano una grande variabilità nelle risposte corticali, sia in latenza che in ampiezza. Inoltre, è stata riscontrata una differenza significativa nella latenza del primo picco principale delle risposte evocate: gli LFP evocati in animali anestetizzati con tiletamina-xylazina presentavano una latenza più lunga di quelli registrati in ratti anestetizzati con uretano. Inoltre, le distribuzioni dei parametri analizzati erano più strette e piccate in uretano, in corrispondenza di tutte le profondità corticali. Questo comportamento è sicuramente da attribuire al differente meccanismo d'azione dei due anestetici su specifici recettori sinaptici, e quindi nell'elaborazione e nella trasmissione dell'informazione sensoriale lungo tutto il percorso corticale. E' stato inoltre discusso il ruolo della attività basale nella modulazione della risposta evocata. A questo proposito, è stata registrata l'attività spontanea in corrispondenza dei vari layer corticali ed analizzata nel contesto statistico delle 'valanghe neuronali'. Una valanga neuronale è una cascata di attività elettrica in una rete neuronale, la cui distribuzione statistica dei parametri principali (dimensione e vita media) può essere approssimata da una legge di potenza. La distribuzione delle dimensioni di una valanga in una rete neuronale segue una legge di potenza del tipo P(s)=s^-a, con a=1.5. Tale esponente è un riflesso delle correlazioni spaziali a lungo raggio nell'attività neuronale spontanea. Dal momento che i picchi negativi (nLFPs) nelle tracce elettrofisiologiche originano dalla somma di potenziali d'azione sincronizzati generati da neuroni posti nelle vicinanze dell'elettrodo di registrazione, ci siamo chiesti se fosse possibile modellizare i singoli nLFP registrati nell'attività basale tramite un singolo elettrodo come il risultato di valanghe neuronali locali. Pertanto, abbiamo analizzato la distribuzione della dimensione (cioè l'ampiezza in uV) di tali picchi, in modo da identificare una distribuzione power-law appropriata, che potesse descrivere anche le registrazioni a singolo elettrodo. Infine, sono presentate e discusse le prime registrazioni in assoluto degli LFP evocati lungo un'intera colonna corticale ottenute tramite l'ultima generazione di chip impiantabili a tecnologia CMOS. Questi ultimi presentano una matrice di 256 siti di registrazione, organizzata secondo due possibili topologie, 16 x 16 o 4 x 64, e avente una distanza tra gli elettrodi pari a 15 um o 33 um rispettivamente. Una precisa dinamica di propagazione dei potenziali evocati può già essere riconosciuta in questi primissimi profili corticali. Nel prossimo futuro, l'uso di questi dispositivi a semiconduttore potrà aiutare a comprendere il decorso di sindromi neurodegerative come il Parkinson o l'Alzheimer, associando sintomi e comportamenti tipo della malattia a specifiche caratteristiche neuronali. I chip impiantabili potranno anche essere utilizzati come 'electroceuticals', ossia potranno aiutare a rallentare (o addirittura a capovolgere) il decorso delle malattie neurogenerative, costituendo le basi di protesi neuronali in grado di sostenere fisicamente o allenare funzionalmente le popolazioni neuronali danneggiate. L'identificazione e il rilevamento di segnali neuronali ad alta risoluzione aiuterà anche a sviluppare complesse interfacce cervello-macchina, che consentiranno il controllo di protesi intelligenti e che forniranno sofisticati meccanismi di feedback a chi ha perso l'uso di alcune parti del proprio corpo o determinate funzioni cerebrali.

Cette thèse explore la synergie entre l'intelligence artificielle (IA) et la neuroscience cognitive pour faire progresser les capacités de traitement du langage. Elle s'appuie sur l'idée que les avancées en IA, telles que les réseaux neuronaux convolutionnels et des mécanismes comme le « replay d'expérience », s'inspirent souvent des découvertes neuroscientifiques. Cette interconnexion est bénéfique dans le domaine du langage, où une compréhension plus profonde des capacités cognitives humaines uniques, telles que le traitement de structures linguistiques complexes, peut ouvrir la voie à des systèmes de traitement du langage plus sophistiqués. L'émergence de riches ensembles de données neuroimagerie naturalistes (par exemple, fMRI, MEG) aux côtés de modèles de langage avancés ouvre de nouvelles voies pour aligner les modèles de langage computationnels sur l'activité cérébrale humaine. Cependant, le défi réside dans le discernement des caractéristiques du modèle qui reflètent le mieux les processus de compréhension du langage dans le cerveau, soulignant ainsi l'importance d'intégrer des mécanismes inspirés de la biologie dans les modèles computationnels.En réponse à ce défi, la thèse introduit un cadre basé sur les données qui comble le fossé entre le traitement neurolinguistique observé dans le cerveau humain et les mécanismes computationnels des systèmes de traitement automatique du langage naturel (TALN). En établissant un lien direct entre les techniques d'imagerie avancées et les processus de TALN, elle conceptualise le traitement de l'information cérébrale comme une interaction dynamique de trois composantes critiques : le « quoi », le « où » et le « quand », offrant ainsi des perspectives sur la manière dont le cerveau interprète le langage lors de l'engagement avec des récits naturalistes. L'étude fournit des preuves convaincantes que l'amélioration de l'alignement entre l'activité cérébrale et les systèmes de TALN offre des avantages mutuels aux domaines de la neurolinguistique et du TALN. La recherche montre comment ces modèles computationnels peuvent émuler les capacités de traitement du langage naturel du cerveau en exploitant les technologies de réseau neuronal de pointe dans diverses modalités - langage, vision et parole. Plus précisément, la thèse met en lumière comment les modèles de langage pré-entraînés modernes parviennent à un alignement plus étroit avec le cerveau lors de la compréhension de récits. Elle examine le traitement différentiel du langage à travers les régions cérébrales, le timing des réponses (délais HRF) et l'équilibre entre le traitement de l'information syntaxique et sémantique. En outre, elle explore comment différentes caractéristiques linguistiques s'alignent avec les réponses cérébrales MEG au fil du temps et constate que cet alignement dépend de la quantité de contexte passé, indiquant que le cerveau code les mots légèrement en retard par rapport à celui actuel, en attendant plus de contexte futur. De plus, elle met en évidence la plausibilité biologique de l'apprentissage des états de réservoir dans les réseaux à état d'écho, offrant ainsi une interprétabilité, une généralisabilité et une efficacité computationnelle dans les modèles basés sur des séquences. En fin de compte, cette recherche apporte des contributions précieuses à la neurolinguistique, à la neuroscience cognitive et au TALN
This thesis explores the synergy between artificial intelligence (AI) and cognitive neuroscience to advance language processing capabilities. It builds on the insight that breakthroughs in AI, such as convolutional neural networks and mechanisms like experience replay 1, often draw inspiration from neuroscientific findings. This interconnection is beneficial in language, where a deeper comprehension of uniquely human cognitive abilities, such as processing complex linguistic structures, can pave the way for more sophisticated language processing systems. The emergence of rich naturalistic neuroimaging datasets (e.g., fMRI, MEG) alongside advanced language models opens new pathways for aligning computational language models with human brain activity. However, the challenge lies in discerning which model features best mirror the language comprehension processes in the brain, underscoring the importance of integrating biologically inspired mechanisms into computational models. In response to this challenge, the thesis introduces a data-driven framework bridging the gap between neurolinguistic processing observed in the human brain and the computational mechanisms of natural language processing (NLP) systems. By establishing a direct link between advanced imaging techniques and NLP processes, it conceptualizes brain information processing as a dynamic interplay of three critical components: "what," "where," and "when", offering insights into how the brain interprets language during engagement with naturalistic narratives. This study provides compelling evidence that enhancing the alignment between brain activity and NLP systems offers mutual benefits to the fields of neurolinguistics and NLP. The research showcases how these computational models can emulate the brain’s natural language processing capabilities by harnessing cutting-edge neural network technologies across various modalities—language, vision, and speech. Specifically, the thesis highlights how modern pretrained language models achieve closer brain alignment during narrative comprehension. It investigates the differential processing of language across brain regions, the timing of responses (Hemodynamic Response Function (HRF) delays), and the balance between syntactic and semantic information processing. Further, the exploration of how different linguistic features align with MEG brain responses over time and find that the alignment depends on the amount of past context, indicating that the brain encodes words slightly behind the current one, awaiting more future context. Furthermore, it highlights grounded language acquisition through noisy supervision and offers a biologically plausible architecture for investigating cross-situational learning, providing interpretability, generalizability, and computational efficiency in sequence-based models. Ultimately, this research contributes valuable insights into neurolinguistics, cognitive neuroscience, and NLP

The entorhinal cortex (EC) in the medial temporal lobe plays a critical role in memory formation and is implicated in several neurological diseases including temporal lobe epilepsy and Alzheimer’s disease. Despite the known importance of this brain region, little is known about the normal bioelectrical activity patterns of the EC in awake, behaving primates. In order to develop effective therapies for diseases affecting the EC, we must first understand its normal properties. To contribute to our understanding of the EC, I monitored the activity of individual neurons and populations of neurons in the EC of rhesus macaque monkeys during free-viewing of photographs using electrophysiological techniques. The results of these experiments help to explain how primates can form memories of, and navigate through, the visual world. These experiments revealed neurons in the EC that represent visual space with triangular grid receptive fields and other neurons that prefer to fire near image borders. These properties are similar to those previously described in the rodent EC, but here the neuronal responses relate to viewing of remote space as opposed to representing the physical location of the animal. The representation of visual space may be aided by another EC neuron type that was discovered, free-viewing saccade direction cells, neurons that signaled the direction of upcoming saccades. Such a signal could be used by other cells to prepare to fire according to the future gaze location. Many of these spatially-responsive neurons also represented memory for images, suggesting that they may be useful for associating items with their locations. I also examined the neuronal circuitry of recognition memory for visual stimuli in the EC, and I found that population synchronization within the gamma-band (30-140 Hz) in superficial layers of the EC was modulated by stimulus novelty, while the strength of memory formation modulated gamma-band synchronization in the deep layers and in layer III. Furthermore, the strength of connectivity in the gamma-band between different layers was correlated with the strength of memory formation, with deep to superficial power transfer being correlated with stronger memory formation and superficial to deep transfer correlated with weaker memory formation. These findings support several previous investigations of hippocampal-entorhinal connectivity in the rodent and advance our understanding of the functional circuitry of the medial temporal lobe memory system. Finally, I explored the design of a device that could be used to investigate properties of brain tissue in vitro, potentially aiding in the development of treatments for disorders of the EC and other brain structures. We designed, fabricated, and validated a novel device for long-term maintenance of thick brain slices and 3-dimensional dissociated cell cultures on a perforated multi-electrode array. To date, most electrical recordings of thick tissue preparations have been performed by manually inserting electrode arrays. This work demonstrates a simple and effective solution to this problem by building a culture perfusion chamber around a planar perforated multi-electrode array. By making use of interstitial perfusion, the device maintained the thickness of tissue constructs and improved cellular survival as demonstrated by increased firing rates of perfused slices and 3-D cultures, compared to unperfused controls. To the best of our knowledge, this is the first thick tissue culture device to combine forced interstitial perfusion for long-term tissue maintenance and an integrated multi-electrode array for electrical recording and stimulation.

An ambitious long-term goal for neuroevolution, which studies how artificial evolutionary processes can be driven to produce brain-like structures, is to evolve neurocontrollers with a high density of neurons and connections that can adapt and learn from past experience. Yet while neuroevolution has produced successful results in a variety of domains, the scale of natural brains remains far beyond reach. In this dissertation two extensions to the recently introduced Hypercube-based NeuroEvolution of Augmenting Topologies (HyperNEAT) approach are presented that are a step towards more brain-like artificial neural networks (ANNs). First, HyperNEAT is extended to evolve plastic ANNs that can learn from past experience. This new approach, called adaptive HyperNEAT, allows not only patterns of weights across the connectivity of an ANN to be generated by a function of its geometry, but also patterns of arbitrary local learning rules. Second, evolvable-substrate HyperNEAT (ES-HyperNEAT) is introduced, which relieves the user from deciding where the hidden nodes should be placed in a geometry that is potentially infinitely dense. This approach not only can evolve the location of every neuron in the network, but also can represent regions of varying density, which means resolution can increase holistically over evolution. The combined approach, adaptive ES-HyperNEAT, unifies for the first time in neuroevolution the abilities to indirectly encode connectivity through geometry, generate patterns of heterogeneous plasticity, and simultaneously encode the density and placement of nodes in space. The dissertation culminates in a major application domain that takes a step towards the general goal of adaptive neurocontrollers for legged locomotion.
ID: 031001435; System requirements: World Wide Web browser and PDF reader.; Mode of access: World Wide Web.; Adviser: Kenneth O. Stanley.; Title from PDF title page (viewed June 24, 2013).; Thesis (Ph.D.)--University of Central Florida, 2012.; Includes bibliographical references (p. 165-178).
Ph.D.
Doctorate
Computer Science
Engineering and Computer Science
Computer Science

La génétique inverse est devenue une méthode clé pour la production de virus à ARN génétiquement modifiés et pour comprendre les propriétés cellulaires et biologiques des virus. Cependant les méthodes les plus fréquemment utilisées, basées sur le clonage de génomes viraux complets dans des plasmides, sont laborieuses et imprévisibles. La première partie de cette thèse présente des études sur la mise au point d'un nouveau système de génétique inverse, appelé méthode ISA (Amplicons-Sous génomique-Infectieux), qui permet la génération, en quelques jours, de virus infectieux sauvages et génétiquement modifiés appartenant à trois familles différentes de virus à ARN simple brin de polarité positive, avec une grande maîtrise des séquences virales. Dans la deuxième partie de cette thèse, nous avons appliqué pour la première fois à un arbovirus (CHIKV), le ré-encodage des codons - une méthode développée récemment et très excitante pour le développement de vaccins vivants atténués. En utilisant une approche aléatoire de ré-encodage des codons qui attribue au hasard des codons sur la base de la séquence en acides aminés correspondante, nous avons mis en évidence des pertes importantes de fitness réplicatif sur des cellules de primates et d'arthropodes. La diminution du fitness réplicatif est en corrélation avec le degré de ré-encodage, une observation qui peut aider à la modulation de l'atténuation virale. En utilisant l'expérience acquise avec le CHIKV, nous avons transposé avec succès ce mécanisme d'atténuation au JEV et amélioré notre maîtrise du processus d'atténuation en utilisant une combinaison de la synthèse de novo et de la méthode ISA
Reverse genetics has become a key methodology for producing genetically modified RNA viruses and deciphering cellular and viral biological properties, but the most commonly used methods, based on the preparation of plasmid-based complete viral genomes, are laborious and unpredictable. The first part of this thesis presents studies relating to the development of a new reverse genetics system, designated the ISA (Infectious-Subgenomic-Amplicons) method, which enabled the generation of both wild-type and genetically modified infectious viruses belonging to three different families of positive, single stranded RNA viruses within days with great control of the viral sequences. In the second part of this thesis, we applied for the first time to an arbovirus (CHIKV), codon re-encoding - a recently developed and very exciting method for the development of live attenuated vaccines. Using a random codon re-encoding approach which randomly attributed nucleotide codons based on their corresponding amino acid sequence, we identified major fitness losses of CHIKV in both primate and arthropod cells. The decrease of replicative fitness correlated with the extent of re-encoding, an observation that may assist in the modulation of viral attenuation. Detailed analysis of these observed replicative fitness losses indicated that they are the consequence of several independent re-encoding induced events. Using the experience acquired on the CHIKV, we successfully transposed this attenuation mechanism to JEV and improved our control of the attenuation process by using a combination of de novo synthesis and the ISA method

The desire to enhance and make ourselves better is not a new one and it has continued to intrigue throughout the ages. Individuals have continued to seek ways to improve and enhance their well-being for example through nutrition, physical exercise, education and so on. Crucial to this improvement of their well-being is improving their ability to remember. Hence, people interested in improving their well-being, are often interested in memory as well. The rationale being that memory is crucial to our well-being. The desire to improve one’s memory then is almost certainly as old as the desire to improve one’s well-being. Traditionally, people have used different means in an attempt to enhance their memories: for example in learning through storytelling, studying, and apprenticeship. In remembering through practices like mnemonics, repetition, singing, and drumming. In retaining, storing and consolidating memories through nutrition and stimulants like coffee to help keep awake; and by external aids like notepads and computers. In forgetting through rituals and rites. Recent scientific advances in biotechnology, nanotechnology, molecular biology, neuroscience, and information technologies, present a wide variety of technologies to enhance many different aspects of human functioning. Thus, some commentators have identified human enhancement as central and one of the most fascinating subject in bioethics in the last two decades. Within, this period, most of the commentators have addressed the Ethical, Social, Legal and Policy (ESLP) issues in human enhancements as a whole as opposed to specific enhancements. However, this is problematic and recently various commentators have found this to be deficient and called for a contextualized case-by-case analysis to human enhancements for example genetic enhancement, moral enhancement, and in my case memory enhancement (ME). The rationale being that the reasons for accepting/rejecting a particular enhancement vary depending on the enhancement itself. Given this enormous variation, moral and legal generalizations about all enhancement processes and technologies are unwise and they should instead be evaluated individually. Taking this as a point of departure, this research will focus specifically on making a case for ME and in doing so assessing the ESLP implications arising from ME. My analysis will draw on the already existing literature for and against enhancement, especially in part two of this thesis; but it will be novel in providing a much more in-depth analysis of ME. From this perspective, I will contribute to the ME debate through two reviews that address the question how we enhance the memory, and through four original papers discussed in part three of this thesis, where I examine and evaluate critically specific ESLP issues that arise with the use of ME. In the conclusion, I will amalgamate all my contribution to the ME debate and suggest the future direction for the ME debate.

Neuroeconomists investigate how the human brain analyzes and makes decisions about financial situations. They use functional magnetic resonance imaging (fMRI) of subjects who participate in economic games. Here we present three such experiments.

In the first experiment, we investigate how the brain recombines expected reward (ER) and risk. Recent fMRI results show that the brain decomposes a gamble in terms of these two metrics. However, economic theory predicts that the brain must recombine them in order to obtain an effective evaluation of the gamble. It was not clear what biological mechanism directs such recombination. Here we show that the brain uses the correlation of noise to recombine signals. We implement a new technique based on canonical correlation analysis and we show that ER is added to risk to form a metric that activates the medial prefrontal cortex.

In the second experiment, we investigate how the brain encodes two gambles instead of one. The brain is likely to encode the utility of each gamble in a common area but in separate groups of neurons. However, it is unknown how the brain indexes the gambles. Indeed, which group of neuron encodes which gamble can be decided in many ways. We hypothesized that the brain would use either the physical position of the gambles or an idiosyncratic parameter, such as ER or risk. Here we introduce a new analysis technique based on Hotelling T-squared statistics and we show that the brain uses risk as an index.

In the third experiment, we investigate a much more complex situation: a stock market. Contrary to what standard finance theory predicts, we hypothesize that the brain does not use mathematical models but instead heuristically uses a social cognition approach. Specifically, we posit that humans understand stock markets by using Theory of Mind (ToM), the ability to attribute to others mental states different from one's own. Here we show that humans engage brain structures related to ToM (paracingulate cortex, anterior cingulate cortex, insula, and amygdala). Subsequent behavioral tests show that ToM, rather than mathematical, abilities are better predictors of success in forecasting stock markets.

The wake human brain constantly samples perceptual information from the environment and integrates them into existing neuronal networks. Neuronal oscillations have been ascribed a key role in the formation of novel memories. The theta rhythm (3-8 Hz) reflects a central executive mechanism, which integrates novel information, reflected in theta-coupled gamma oscillations (> 30 Hz). Alpha oscillations (8-14 Hz) reflect an attentional gating mechanism, which inhibit task irrelevant neuronal processes. In my dissertation I further scrutinized the oscillatory dynamics of memory formation. Study 1 demonstrated that theta-gamma coupling reflects a specific mechanism for associative memory formation. In study 2, I experimentally entrained memory encoding by visual evoked theta-gamma coupling processes, to underline its functional relevance. In two developmental studies, I found that the theta rhythm indexes explicit learning processes in adults and young children (study 3), and that visually entrained theta oscillations are sensitive to the encoding of novel, unexpected events, already in the first year of life (study 4). Throughout these studies alpha oscillations were not sensitive to memory formation processes, but indicated perceptual (study 1) and semantic (study 3) processes. I propose an integrative framework, suggesting that the alpha rhythm reflects activated semantic representations in the neocortex, while theta-gamma coupling reflects an explicit mnemonic control mechanism, which selects, elaborates and integrates activated representations. Specifically, by squeezing real time events onto a faster, neuronal time scale, theta-gamma coding facilitates neuronal plasticity in medio-temporal networks and advances neuronal processes ahead of real time to emulate and guide future behavior.

Sensory object recognition is the most fundamental of operations performed by the brain. A key computational difficulty of object recognition is that it requires both selectivity to particular objects (e.g., exact odor mixture identification) and generalization across objects (identifying particular features or components common to different odors). Although previous results (1) suggest that odor identity and intensity are represented in the activity of both PNs and KCs, it is not clear how these representations generalize across complex odor mixtures. In particular, it is not clear what types of information are available in KC population (or if its even possible to decode across KC populations?) and how is this information represented? Using the locust olfactory system as a model system, we found that Kenyon cells (KCs), the principal neurons of the mushroom body, an area required for associative learning can identify the presence of components in mixtures and thus enable odor segmentation. As a population, small groups of KCs can both identify and categorize odors with high accuracy. We identified and tested simple circuit requirements for this computation, and propose that odor representations in mushroom bodies are optimized for odor memorization, identification and generalization. These rules may be relevant for pattern classifying circuits in general.

碩士
大葉大學
資訊工程學系碩士班
100
The present literatures almost discuss about context-aware applications by the elements of Global Positioning System (GPS) which can detect a geographical position, and Radio Frequency Identification (RFID) which discriminates from a status. It performs that the mobile learning of the conditions which in outside a student can be perceived. Mostly, the functions of the related sensors detect the conditions which outside a student. Basically, the traditional learning assessment mechanism is a passive and negative assessment mechanism, which cannot provide an real-time learning warning mechanism for teachers or students to find out problems as early as possible (including such learning conditions as “absence of mind” resulting from poor learning stage or physical or psychological factor), and the post-assessment mechanism also cannot assess the learning effectiveness provided by the online learning system. From a viewpoint of cognitive neuroscience, this research proposes the technique of digital encoding of brain-wave characteristic frequency bands to discriminate from the feature of brain wave when performing a student's online learning or computer game. The experiments proceeded by catching brain wave signals of human vision while sensing the test interfaces of graphics and words representing learning or computer games by brain wave sensor. To compile the related samples of energies from brain wave frequency band and times of appearances, then establish the characteristic frequency bands of brain wave and its digital coding modes which stand for the statuses of learning, deep sleeping, playing computer games and taking a break. The proposed digital encoding technique not only can discriminate from whether for a student to concentrate, and for it to learn or to perform a computer game, but it can recognize from performing the heterogeneous games of which kind of character. It can turn out that a teacher and the partner of study also bring about the cause of a student's learning disability, and the system not only can provide a student with the system of the early warning of instant study, but can offer suitable consideration and encouragement.

Diffusion MRI (dMRI) is a non-invasive technique sensitive to microstructural changes that cannot be resolved by other conventional imaging techniques. Diffusion anisotropy measures from conventional dMRI techniques, such as Diffusion Tensor Imaging, do not only depend on tissue microstructural properties but are also confounded by tissue orientational dispersion. As an attempt to suppress this confounding effect, more advanced dMRI imaging techniques based on non-conventional diffusion MRI had been recently developed to quantify the microscopic diffusion fractional anisotropy (𝜇FA) without any prior assumptions of the underlying tissue. Measuring 𝜇FA allows the assessment of microstructural alterations related to tissue maturation, degeneration, and pathology independently of changes in tissue organization. However, non-conventional dMRI sequences are not easily accessible on current clinical MRI scanners. Therefore, more recent studies had suggested the use of microstructural models to quantify 𝜇FA from data acquired from conventional dMRI sequences. In this dissertation, the first reference open-source implementations of different microstructural models to estimate 𝜇FA are provided, in which by computing the spherical mean of the signal acquired, the orientation’s influence from the dMRI data is withdrawn. These models included the one and two-compartmental spherical mean techniques (SMT1 and SMT2) and a novel adaption of the fiber ball imaging model for 𝜇FA estimation. The implementation of these models is evaluated based on numerical simulations, tested on open-source in vivo data of a healthy human brain, and finally compared to the gold standard reference estimated from non-conventional dMRI sequences in a pre-clinical setting. Results show that though their parameter estimates are independent to tissue orientation effect, the SMT1 and SMT2 models provide over and underestimated values of 𝜇FA, which were shown to be a consequence of their imposed assumptions. Although the adapted version of the FBI model did not show to provide robust 𝜇FA estimates, its’ axonal water fraction estimates showed a high correlation to the gold standard 𝜇FA estimates. This finding supports that axonal water fraction is a determinant factor of 𝜇FA in heathy neural tissues. Therefore, in future studies, the further development of alternative techniques to estimate 𝜇FA based on the information captured by FBI's axonal water estimates could be of interest.
A ressonância magnética por difusão (dMRI, do inglês diffusion magnetic resonance imaging) é uma técnica não invasiva, sensível a alterações microestruturais que podem não ser resolvidas por outras técnicas de imagem estruturais convencionais. As medidas de anisotropia, a partir de técnicas convencionais de dMRI, não dependem apenas das propriedades microestruturais do tecido, mas também da dispersão na orientação do tecido. Na tentativa de suprimir esse efeito, foram recentemente desenvolvidas técnicas mais avançadas de imagem de ressonância magnética por difusão baseadas em sequências de difusão não convencionais de forma a quantificar a anisotropia microscópica fracionária de difusão (𝜇FA, do inglês microscopic fractional anisotropy) sem qualquer suposição prévia sobre o tecido subjacente. A medição da anisotropia microscópica de difusão permite a avaliação das alterações microestruturais relacionadas à maturação, degeneração e patologia de um tecido, independentemente das alterações na organização do mesmo. No entanto, as sequências de dMRI não convencionais não são facilmente acessíveis nos aparelhos de ressonância magnética atuais. Posto isto, estudos mais recentes sugeriram o uso de modelos microestruturais para quantificar 𝜇FA a partir de dados adquiridos através de sequências de dMRI convencionais. No decorrer desta dissertação, são fornecidas as primeiras implementações de diferentes modelos microestruturais de referência em open-source para estimar 𝜇FA. Modelos estes que através do cálculo da média esférica de um sinal adquirido, retiram a influência da orientação nos dados de dMRI. Esses modelos incluem as técnicas baseadas em médias esféricas de um sinal dMRI (SMT, do inglês Spherical Mean Techniques), de um e dois compartimentos, (SMT1 e SMT2) e uma nova adaptação do modelo da imagem por fibras em bola (FBI, do inglês Fiber Ball Imaging) para estimar o 𝜇FA. A implementação dos modelos é avaliada tendo como base simulações numéricas, testadas em dados de cérebros humanos saudáveis, adquiridos in vivo. Finalmente, o 𝜇FA é então comparado com a referência padrão estimada a partir de sequências dMRI não convencionais, num ambiente pré-clínico. Os resultados mostram que, embora as estimativas dos parâmetros sejam independentes do efeito da orientação do tecido, os modelos SMT1 e SMT2 fornecem valores de 𝜇FA acima e abaixo da referência, o que se mostra ser uma consequência das suposições impostas. Embora a versão adaptada do modelo FBI não tenha resultado em estimativas robustas de 𝜇FA, as suas estimativas da fração de água axonal (AWF, do inglês axonal water fraction) evidenciaram uma alta correlação com as estimativas padrão de 𝜇FA. Assim, estes resultados demonstram que a AWF é, efetivamente, um fator determinante de 𝜇FA em tecidos neuronais saudáveis. No que concerne a estudos futuros, pode revelar-se interessante analisar o desenvolvimento de técnicas alternativas para estimar 𝜇FA com base em informações obtidas pelas estimativas de AWF do modelo FBI.

博士
國立交通大學
資訊科學與工程研究所
104
Visual encoding and decoding are crucial aspects in investigating the neural representation of visual information in the human brain. Once a decoding and encoding model is constructed, it can be used to decode the low-level visual patterns or higher-level visual stimuli such as facial images. Such decoding requires the transformation from a neural representation to a perceptual representation, which lays essential foundation to hierarchical face processing. Previous studies have proposed models for low-level visual processing. However, how the brain process the high-level perception remains an unsolved question. There are three studies in the thesis and three corresponding experiments were conducted for each of the study in the thesis. In the first study, we proposed a bidirectional model for decoding and encoding of visual stimulus based on manifold representation of the temporal and spatial information extracted from magnetoencephalographic (MEG) data. In the proposed decoding process, principal component analysis is applied to extract temporal principal components (TPCs) from the visual cortical activity estimated by a beamforming method. The spatial distribution of each TPC is in a high-dimensional space and can be mapped to the corresponding spatiotemporal component (STC) on a low-dimensional manifold. Once the linear mapping between the STC and the wavelet coefficients of the stimulus image is determined, the decoding process can synthesize an image resembling the stimulus image. The encoding process is performed by reversing the mapping or transformation in the decoding model and can predict the spatiotemporal brain activity from a stimulus image. In our first experiments using visual stimuli containing eleven combinations of checkerboard patches, the information of spatial layout in the stimulus image was revealed in the embedded manifold. The correlation between the reconstructed and original images was 0.71 and the correlation map between the predicted and original brain activity was highly correlated to the map between the original brain activity for different stimuli (r=0.89). In the second study, we applied a decoding method based on the decoding and encoding model to face representations. A low-dimensional neural manifold was constructed using a set of single-trial brain activity data evoked by stimuli with basic face viewpoints and gaze directions. As a perceptual representation with synthesis property, this manifold was able to predict composite viewpoints and directions from brain activity. In the second experiments, when facial images with varying viewpoints and gaze-directions were used as the experimental stimuli, the M170 component in occipital face area and the right superior temporal sulcus gave accurate prediction for face viewpoints and gaze directions, respectively. In the third study, we proposed a supervised locally linear embedding method to construct the embedded manifold from brain activity, taking into account similarities between corresponding stimuli. In our experiments, photographic portraits were used as visual stimuli and brain activity was calculated from MEG data using a source localization method. The results of 10×10-fold cross-validation revealed a strong correlation between manifolds of brain activity and the orientation of faces in the presented images, suggesting that high-level information related to image content can be revealed in the brain responses represented in the manifold. These results suggest that the temporal component is important in visual processing and manifolds can well represent the information related to visual perception. In addition, the proposed neural manifold method can be used to construct an effective perceptual representation for face processing and is applicable to investigation into the inherent patterns of brain activity.

Tese de mestrado integrado em Engenharia Biomédica e Biofísica, apresentada à Universidade de Lisboa, através da Faculdade de Ciências, 2016
A interação com o mundo é feita através de movimento - desde a locomoção até à comunicação verbal - tornando o controlo de movimento um dos aspectos fundamentais de maior interesse em neurociência. O controlo de movimento tem sido alvo de observação desde cedo em estudos comportamentais e neurofisiológicos, e sabemos hoje que os movimentos voluntários resultam de padrões de impulsos elétricos gerados no sistema nervoso. Contudo, não conhecemos ainda os aspetos mais precisos da geração de padrões de movimento nem a sua relação com parâmetros como direção, velocidade, etc. Uma característica importante do controlo de movimento é a existência de tuning direcional - que consiste numa resposta neuronal preferencial a uma direção de movimento. Ao executar movimentos numa direção preferida, alguns neurónios despolarizam a uma frequência máxima, e a mesma diminui gradualmente à medida que o movimento se afasta da direção preferida. Este fenómeno foi caracterizado em 1982 em áreas motoras (córtex motor primário) ao serem executados movimentos direcionais do braço contralateral. Contudo, estudos recentes mostram a existência de tuning direcional não só para o membro contralateral, mas também para o membro ipsilateral. Estas representações direcionais foram encontradas com medições electrofisiológicas ao nível celular, e também com técnicas modernas de imagiologia que medem sinal proveniente de volumes da ordem de mm3, como ressonância magnética funcional. Com recurso a ambos os tipos de técnicas foram encontradas representações ipsilaterais bem estruturadas para movimentos ao nível do braço bem como dos dedos. Desta forma, ambos os hemisférios cerebrais codificam movimentos direccionais de ambas as mãos. Sabemos também, por experiência quotidiana, que os movimentos bimanuais são bem coordenados, o que sugere que os mesmos são gerados tomando em conta informação de ambas as mãos. No entanto, a relação entre os padrões neuronais de movimentos bimanuais e unimanuais ainda não é clara. Nesta dissertação pretende-se localizar e caracterizar tuning direcional durante movimentos unimanuais e bimanuais no cérebro humano. Desta forma temos como objectivo procurar quais as regiões corticais que codificam movimentos direcionais da mão contralateral, da mão ipsilateral, bem como as representações de movimentos bimanuais e a sua relação com movimentos unimanuais. Para tal, foi desenhada uma experiência motora para testar movimentos direcionais, que foi executada em simultâneo com a aquisição de imagens de ressonância magnética funcional. Foi desenvolvido um dispositivo para monitorizar de forma precisa movimentos da mão. De forma a assegurar compatibilidade com o ambiente em ressonância magnética, foram construídos dois manípulos ergonómicos com recurso a impressão 3D em nylon. Os manípulos foram equipados com sensores de rotação resistivos, e foram montados numa mesa de suporte desenvolvida para o efeito. Afim de treinar os participantes e controlar a experiência, foi desenvolvido um protocolo motor organizado de forma semelhante a um jogo de alvos. Os participantes controlaram a posição de cursores num ecrã utilizando movimentos das mãos, monitorizados pelo dispositivo. O objectivo do protocolo motor foi atingir 6 alvos radiais com os cursores e voltar à posição central, com movimentos de cada uma das mãos, ou as duas (para todas as combinações de 6 alvos para cada mão). No total, a experiência consistiu em 48 condições de movimento – 6 movimentos radiais para a mão esquerda, 6 para a mão direita e 36 combinações bimanuais. A experiência motora foi executada por 7 sujeitos destros saudáveis. Após uma sessão de treino, a experiência decorreu num scanner de ressonância magnética funcional Siemens Trio 3T, onde foram adquiridas imagens funcionais durante 10 repetições da experiência para cada sujeito. Adicionalmente, foram adquiridos dados de cinemática para as duas mãos durante as sessões de treino e de teste. A análise de dados de cinemática consistiu na observação de tempos de reação e de movimento em cada condição. Comparámos condições unimanuais e bimanuais, testámos efeitos de direção, e ainda combinações bimanuais (movimentos simétricos, paralelos ou não relacionados). Para cada uma destas hipóteses foram usados os testes estatísticos aplicáveis. Não foram observados efeitos significativos nos tempos de reação, de forma consistente, para qualquer das condições em estudo. Pelo contrário, os tempos de movimento foram consistentemente sensíveis aos efeitos estudados. As imagens por ressonância magnética funcional foram analisadas numa primeira fase conforme o procedimento tradicional. Este consiste no pré-processamento - envolvendo correções espaciais de efeitos de campo magnético, filtragem temporal, alinhamento com a imagem anatómica e segmentação. De seguida foi aplicado um modelo linear de forma de forma de independente para cada voxel (unidade discreta de volume) nas imagens. O modelo consistiu em 48 variáveis categóricas, correspondentes às condições de movimento em estudo, e 10 variáveis categóricas correspondentes às sessões de repetição da experiência. O objetivo deste modelo é a estimação dos pesos (β) da regressão linear, i.e., para cada condição é estimada a influência da mesma no sinal em cada voxel. De seguida é possível fazer inferência sobre os valores β - sob a hipótese nula de que são, em média, zero. Procedendo desta forma, foi aplicado um teste t aos regressores β associados a movimentos da mão esquerda, direita, e movimentos bimanuais para as regiões: área sensorial somática I (S1), córtex motor primário (M1), córtex pré-motor ventral e dorsal (PMv, PMd), área motora suplementar (AMS), lóbulo parietal superior, anterior e posterior (LPSa, LPSp) e córtex visual (V12). Foram encontrados valores de activação predominantemente associados com movimentos contralaterais e bimanuais, e activação menor em movimentos ipsilaterais. Os resultados coincidem fortemente com a perspetiva clássica de que cada hemisfério está associado a controlo da mão contralateral. Contudo, os métodos univariados testam o quanto os voxels (ou regiões) variam a sua resposta com condições individuais, tornando a comparação entre condições de movimento difícil. Adicionalmente, estes métodos são indicados para o mapeamento de activação perante estímulos, mas não para avaliar a estrutura da representação de condições, i.e., caracterizar respostas neuronais associadas conjunto de estímulos - como é o caso de tuning direcional. Desta forma, foi aplicado um modelo de análise representacional no qual se pressupõe que os estímulos podem ser caracterizados por padrões de activação - neste caso correspondentes aos valores beta para cada voxel quando é executada uma condição. Neste modelo é calculada uma medida de (dis)similaridade entre todos os pares de condições. Neste projecto foi utilizada a distância Euclidiana, sendo que as comparações entre pares das 48 condições foram organizadas em matrizes de distância. Os resultados revelam, qualitativamente, a presença duma estrutura de tuning direcional bem definida para movimentos contralaterais, bem como ipsilaterais. Também os movimentos bimanuais apresentaram uma estrutura de tuning bem definida e diferenciada entre regiões. De forma a quantificar e inferir acerca da presença de codificação direcional, os valores de distância correspondentes às condições contralaterais, ipsilaterais e bimanuais foram testados estatisticamente. Este teste assenta no pressuposto de que, perante a inexistência de codificação, as distâncias são zero (este pressuposto foi confirmado). Os resultados indicam uma forte codificação direcional de movimentos contralaterais para todas as regiões testadas. Este resultado é coincidente com estudos anteriores que encontram tuning direcional contralateral em todas as regiões em que o mesmo foi investigado. Contudo, encontrámos também uma forte codificação de movimentos ipislaterais, excepto na AMS e LPS anterior no hemisfério direito (não dominante). Estes resultados são coerentes com estudos recentes que mostram uma forte presença de codificação de movimentos ipsilaterais. Os movimentos bimanuais estão também caracterizados por uma forte representação. Contudo, existe a hipótese de que estes estejam presentes apenas como consequência da codificação direcional de movimentos da mão contralateral (ou ipsilateral), e não directamente associados à codificação especializada de movimentos bimanuais. Esta hipótese é, contudo, de elevado interesse, já que uma codificação bimanual especializada pode explicar o mecanismo da coordenação bimanual. Desta forma, as matrizes de distância foram reorganizadas em termos de movimentos da mão esquerda e da mão direita. Os mapas resultantes foram comparados qualitativamente com simulações, revelando uma codificação bimanual maioritariamente associada com movimentos contralaterais. Contudo, a AMS e o córtex premotor ventral aparentam codificar movimentos bimanuais de forma não-linear, que poderá indicar alguma especialização em movimentos bimanuais que poderá ser útil para coordenação. Trabalho futuro envolverá avaliar quantitativamente estes mapas de forma a perceber quanta codificação bimanual é gerada de forma especializada. Os resultados deste estudo coincidem com estudos recentes de codificação ipsilateral, e revisitam questões acerca da codificação bimanual. No futuro pretende-se decompor a codificação bimanual, avaliar de forma extensa e continua a superfície cortical, cerebelo e núcleos da base. Adicionalmente, esperamos executar futuras aquisições em novos participantes. Este tipo de estudo pretende responder a questões no âmbito do controlo neural de movimento, que poderão ser úteis futuramente no contexto da reabilitação e controlo robótico. Consideramos também que os métodos de procura de codificação poderão ser utilizados para caracterização do sistema motor de sujeitos saudáveis em comparação com casos patológicos como acidente vascular cerebral, fornecendo um meio de avaliação dos mesmos.
We interact with the world by moving our body: legs for locomotion, hands for dexterous tasks, and articulatory muscles to communicate. It is known that these movements result from patterns of electrical impulses in the nervous system. However, it is not yet known how the brain controls the fine aspects of movement. One important characteristic of movement control in the brain is directional tuning - a preferential neuronal response to an executed direction. In this work, we examine where and how the brain encodes movement directions in unimanual and bimanual movements in humans. In order to address this question, we designed a motor experiment for directional movements. A hand device was developed in order to precisely monitor hand movements while 7 right-handed healthy participants executed a motor task. The task was built similarly to a game in which participants reached radial targets using wrist movements of one or both hands. After training, subjects executed the motor task in a magnetic resonance scanner. Functional imaging data were acquired and analysed using novel multivoxel pattern analysis, in which we calculate pairwise dissimilarities of patterns of fMRI voxel activity across movement conditions. We tested for encoding of unimanual (contralateral and ipsilateral) and bimanual movements in cortical regions of interest. Kinematics data were also analysed to test for performance effects of direction and hand combination. We found significant encoding of contralateral and bimanual movements in all tested regions. Ipsilateral movements were strongly represented in both hemispheres, except for right supplementary motor area and anterior-superior parietal lobule. Furthermore, the right (non-dominant) hemisphere encoded contralateral movements more preferentially than ipsilateral ones, when compared with the left hemisphere. These results are in line with recent findings of well-defined ipsilateral movement representations. Future work will involve decomposing bimanual tuning functions in order to find a quantitative relationship between bimanual and unimanual encoding.

碩士
國立陽明大學
生物醫學工程學系
106
The encoder in closed-loop brain-machine interfaces (BMIs) plays an important role in establishing a direct communication link between the brain and the external world. There are two methods to build up an encoder, the psychometric equivalence approach and neurophysiological approach. The psychometric equivalence function (PEF) is established by assessing the same performance of detection toward both different parameter of intracortical microstimulation (ICMS) and the mechanical stimulation. However, it’s hard to observe the quality and the quantity of the sensation evoked by ICMS. In the recent research, scientists found out that ICMS could elicit the naturalistic cortical response. Besides, somatosensory cortex, whether in neural firing rate or local field potentials (LFPs), is sensitive to the different velocity of tactile stimulus. As the result, in our research, we propose a stimulus evoked potential (SEP)-based encoder of sensory cortical system which was built up by the concept of PEF. In the past research, compare with firing rate, the LFPs based decoding model is more robust in stimulus decoding for its comprehensive information. For establishing a stable and precise sensory SEP-based encoder of sensory cortical system for the real-time closed-loop BMI model, LFPs would be more suitable. In our study, we’re going to build up a SEP-based encoder in behavioral rat by recording the evoked potential from acceleration stimulus of lever-pressing and the ICMS. By extracting the features from LFP, we could find the stimulus-correlated features for the SEP-based encoder. The SEP-based encoder be established by the linear regression models, logistic regression model, and exponential regression model. Furthermore, we would discuss the result of our SEP-based encoder, and compare the stability and precision between spike-based and SEP-based encoder.

Dans les systèmes cognitifs, le rôle de la mémoire de travail est crucial pour le raisonnement visuel et la prise de décision. D’énormes progrès ont été réalisés dans la compréhension des mécanismes de la mémoire de travail humain/animal, ainsi que dans la formulation de différents cadres de réseaux de neurones artificiels à mémoire augmentée. L’objectif global de notre projet est de former des modèles de réseaux de neurones artificiels capables de consolider la mémoire sur une courte période de temps pour résoudre une tâche de mémoire et les relier à l’activité cérébrale des humains qui ont résolu la même tâche. Le projet est de nature interdisciplinaire en essayant de relier les aspects de l’intelligence artificielle (apprentissage profond) et des neurosciences. La tâche cognitive utilisée est la tâche N-back, très populaire en neurosciences cognitives dans laquelle les sujets sont présentés avec une séquence d’images, dont chacune doit être identifiée pour savoir si elle a déjà été vue ou non. L’ensemble de données d’imagerie fonctionnelle (IRMf) utilisé a été collecté dans le cadre du projet Courtois Neurmod. Nous étudions plusieurs variantes de modèles de réseaux neuronaux récurrents qui apprennent à résoudre la tâche de mémoire de travail N-back en les entraînant avec des séquences d’images. Ces réseaux de neurones entraînés optimisés pour la tâche de mémoire sont finalement utilisés pour générer des représentations de caractéristiques pour les images de stimuli vues par les sujets humains pendant leurs enregistrements tout en résolvant la tâche. Les représentations dérivées de ces réseaux de neurones servent ensuite à créer un modèle de codage pour prédire l’activité IRMf BOLD des sujets. On comprend alors la relation entre le modèle de réseau neuronal et l’activité cérébrale en analysant cette capacité prédictive du modèle dans différentes zones du cerveau impliquées dans la mémoire de travail. Ce travail présente une manière d’utiliser des réseaux de neurones artificiels pour modéliser le comportement et le traitement de l’information de la mémoire de travail du cerveau et d’utiliser les données d’imagerie cérébrale capturées sur des sujets humains lors de la tâche N-back pour potentiellement comprendre certains mécanismes de mémoire du cerveau en relation avec ces modèles de réseaux de neurones artificiels.
In cognitive systems, the role of working memory is crucial for visual reasoning and decision making. Tremendous progress has been made in understanding the mechanisms of the human/animal working memory, as well as in formulating different frameworks of memory augmented artificial neural networks. The overall objective of our project is to train artificial neural network models that are capable of consolidating memory over a short period of time to solve a memory task and relate them to the brain activity of humans who solved the same task. The project is of interdisciplinary nature in trying to bridge aspects of Artificial Intelligence (deep learning) and Neuroscience. The cognitive task used is the N-back task, a very popular one in Cognitive Neuroscience in which the subjects are presented with a sequence of images, each of which needs to be identified as to whether it was already seen or not. The functional imaging (fMRI) dataset used has been collected as a part of the Courtois Neurmod Project. We study multiple variants of recurrent neural network models that learn to remember input images across timesteps. These trained neural networks optimized for the memory task are ultimately used to generate feature representations for the stimuli images seen by the human subjects during their recordings while solving the task. The representations derived from these neural networks are then to create an encoding model to predict the fMRI BOLD activity of the subjects. We then understand the relationship between the neural network model and brain activity by analyzing this predictive ability of the model in different areas of the brain that are involved in working memory. This work presents a way of using artificial neural networks to model the behavior and information processing of the working memory of the brain and to use brain imaging data captured from human subjects during the N-back task to potentially understand some memory mechanisms of the brain in relation to these artificial neural network models.