On the relation between brain images and brain neural networks

Changes in EEG Brain Connectivity Caused by Short-Term BCI Neurofeedback-Rehabilitation Training: A Case Study

Background

In combined with neurofeedback, Motor Imagery (MI) based Brain-Computer Interface (BCI) has been an effective long-term treatment therapy for motor dysfunction caused by neurological injury in the brain (e.g., post-stroke hemiplegia). However, individual neurological differences have led to variability in the single sessions of rehabilitation training. Research on the impact of short training sessions on brain functioning patterns can help evaluate and standardize the short duration of rehabilitation training. In this paper, we use the electroencephalogram (EEG) signals to explore the brain patterns’ changes after a short-term rehabilitation training.

Materials and Methods

Using an EEG-BCI system, we analyzed the changes in short-term (about 1-h) MI training data with and without visual feedback, respectively. We first examined the EEG signal’s Mu band power’s attenuation caused by Event-Related Desynchronization (ERD). Then we use the EEG’s Event-Related Potentials (ERP) features to construct brain networks and evaluate the training from multiple perspectives: small-scale based on single nodes, medium-scale based on hemispheres, and large-scale based on all-brain.

Results

Results showed no significant difference in the ERD power attenuation estimation in both groups. But the neurofeedback group’s ERP brain network parameters had substantial changes and trend properties compared to the group without feedback. The neurofeedback group’s Mu band power’s attenuation increased but not significantly (fitting line slope = 0.2, t-test value p > 0.05) after the short-term MI training, while the non-feedback group occurred an insignificant decrease (fitting line slope = −0.4, t-test value p > 0.05). In the ERP-based brain network analysis, the neurofeedback group’s network parameters were attenuated in all scales significantly (t-test value: p < 0.01); while the non-feedback group’s most network parameters didn’t change significantly (t-test value: p > 0.05).

Conclusion

The MI-BCI training’s short-term effects does not show up in the ERD analysis significantly but can be detected by ERP-based network analysis significantly. Results inspire the efficient evaluation of short-term rehabilitation training and provide a useful reference for subsequent studies.

Meta-Analysis of the Structural Equation Models' Parameters for the Estimation of Brain Connectivity with fMRI

Structural Equation Models (SEM) is among of the most extensively applied statistical techniques in the study of human behavior in the fields of Neuroscience and Cognitive Neuroscience. This paper reviews the application of SEM to estimate functional and effective connectivity models in work published since 2001. The articles analyzed were compiled from Journal Citation Reports, PsycInfo, Pubmed, and Scopus, after searching with the following keywords: fMRI, SEMs, and Connectivity.

Results: A 100 papers were found, of which 25 were rejected due to a lack of sufficient data on basic aspects of the construction of SEM. The other 75 were included and contained a total of 160 models to analyze, since most papers included more than one model. The analysis of the explained variance (R²) of each model yields an effect of the type of design used, the type of population studied, the type of study, the existence of recursive effects in the model, and the number of paths defined in the model. Along with these comments, a series of recommendations are included for the use of SEM to estimate of functional and effective connectivity models.

fMRI MODELS OF DENDRITIC AND ASTROCYTIC NETWORKS

In order to elucidate the relationships between hierarchical structures within the neocortical neuropil and the information carried by an ensemble of neurons encompassing a single voxel, it is essential to predict through volume conductor modeling LFPs representing average extracellular potentials, which are expressed in terms of interstitial potentials of individual cells in networks of gap-junctionally connected astrocytes and synaptically connected neurons. These relationships have been provided and can then be used to investigate how the underlying neuronal population activity can be inferred from the measurement of the BOLD signal through electrovascular coupling mechanisms across the blood-brain barrier. The importance of both synaptic and extrasynaptic transmission as the basis of electrophysiological indices triggering vascular responses between dendritic and astrocytic networks, and sequential configurations of firing patterns in composite neural networks is emphasized. The purpose of this review is to show how fMRI data may be used to draw conclusions about the information transmitted by individual neurons in populations generating the BOLD signal.

Mapping cerebral blood flow during speech production in hereditary ataxia

Dysarthria is a significant feature of the dominantly inherited spinocerebellar ataxias (SCA), but little is known about the patterns of brain activity associated with this disorder of motor speech control. Positron emission tomography (PET) was used to study regional cerebral blood flow during speech and rest in a group of 24 subjects with hereditary ataxia with mild-to-moderate dysarthria. These data were compared to the results obtained from a group of 13 age-matched, normal speakers. In the ataxic subjects, speech rates during scanning were significantly slowed compared to normal speakers. Significant reductions in mean regional blood flow were found in the cerebellum but not in supratentorial regions in the ataxic subjects. Multiple linear regression was used to model speech rate from regional blood flow. Four regions were identified as having significant relationships with speech rate in the model: the left inferior frontal and transverse temporal regions, and the right inferior cerebellar region and caudate nucleus. The relationship between flow and rate was positive in the inferior frontal and cerebellar regions and negative in the caudate and the transverse temporal region. The ataxic model represents an elaboration of the relationship previously reported for normal speakers, likely reflecting both the effects of, and compensation for, cerebellar degeneration in motor speech control. Although the mean regional blood flow values presented a pattern of functional organization for motor speech control at odds with lesion data, the performance-based model was in agreement with clinical experience. Incorporating performance data in functional image analysis may be more revealing of system characteristics than simply examining mean blood flow values.

Systems level modeling of a neuronal network subserving intrinsic alertness

Cognitive control of alertness in unwarned situations (intrinsic alertness) relies on a predominantly right hemisphere cortical and subcortical network. In a previous functional activation study, we have demonstrated that this network comprises the anterior cingulate gyrus, the dorsolateral and polar frontal as well as the inferior parietal cortex, the thalamus and ponto-mesencephalic parts of the brain stem. The aim of this study was to study effective connectivity of this network by employing structural equation modeling. Fifteen right-handed male subjects participated in the PET study. The functional network showed stronger connectivity in the right hemisphere. Furthermore, there were strong effective connections between thalamus and brainstem on the one hand and between thalamus and anterior cingulate on the other. Our results suggest that the anterior cingulate functions as the central coordinating structure for the right hemispheric neural network of intrinsic alertness and that the anterior cingulate gyrus is modulated mainly by prefrontal and parietal cortex.

Fine Modulation in Network Activation during Motor Execution and Motor Imagery

Motor imagery, the 'mental rehearsal of motor acts without overt movements', involves either a visual representation (visual imagery, VI) or mental simulation of movement, associated with a kinesthetic feeling (kinetic imagery, KI). Previous brain imaging work suggests that patterns of brain activation differ when comparing execution (E) with either type of imagery but the functional connectivity of the participating networks has not been studied. Using functional magnetic resonance imaging (fMRI) and structural equation modeling, this study elucidates the inter-relationships among the relevant areas for each of the three motor behaviors. Our results suggest that networks underlying these behaviors are not identical, despite the extensive overlap between E and KI. Inputs to M1, which are facilitatory during E, have the opposite effect during KI, suggesting a physiological mechanism whereby the system prevents overt movements. Finally, this study highlights the role of the connection of superior parietal lobule to the supplementary motor area in both types of motor imagery.

Cerebral modeling and dynamic Bayesian networks

The understanding and the prediction of the clinical outcomes of focal or degenerative cerebral lesions, as well as the assessment of rehabilitation procedures, necessitate knowing the cerebral substratum of cognitive or sensorimotor functions. This is achieved by activation studies, where subjects are asked to perform a specific task while data of their brain functioning are obtained through functional neuroimaging techniques. Such studies, as well as animal experiments, have shown that sensorimotor or cognitive functions are the offspring of the activity of large-scale networks of anatomically connected cerebral regions. However, no one-to-one correspondence between activated networks adn functions can be found. Our research aims at understanding how the activation of large-scale networks derives from cerebral information processing mechanisms, which can only explain apparently conflicting activation data. Our work falls at the crossroads of neuroimaging interpretation techniques and computational neuroscience. Since knowledge in cognitive neuroscience is permanently evolving, our research aims more precisely at defining a new modeling formalism and at building a flexible simulator, allowing a quick implementation of the models, for a better interpretation of cerebral functional images. It also aims at providing plausible models, at eht level of large-scale networks, of cerebral information processing mechanisms in humans. In this paper, we propose a formalism, based on dynamic Bayesian networks (DBNs), that respects the following constraints: an oriented, networks architecture, whose nodes (the cerebral structures) can all be different, the implementation of causality--the activation of the structure is caused by upstream nodes' activation--the explicit representation of different time scales (from 1 ms for the cerebral activity to many seconds for a PET scan image acquisition), the representation of cerebral information at the integrated level of neuronal populations, the imprecision of functional neuroimaging data, the nonlinearity and the uncertainty in cerebral mechanisms, and brain's plasticity (learning, reorganization, modulation). One of the main problems, nonlinearity, has been tackled thanks to new extensions of the Kalman filter. The capabilities of the formalism's current version are illustrated by the modeling of a phoneme categorization process, explaining the different cerebral activations in normal and dyslexic subjects.

Feature-based attention in human visual cortex: simulation of fMRI data

We use a computational neuroscience approach to study the role of feature-based attention in visual perception. This model is used to numerically simulate a visual attention experiment. The neurodynamical system consists of many interconnected modules that can be related to the dorsal and ventral paths of the visual cortex. The biased competition hypothesis is taken into account within the model. From the experimental point of view, measurements exist, which confirm that feature-based attention influences visual cortical responses to stimuli outside the attended location. These measurements show that attention to a given stimulus attribute (in this case "color red") increases the response of cortical visual areas to a spatially distant, ignored stimulus that shares the same attribute. Our neurodynamical model is used to numerically compute the neural activity of area V4 corresponding to such ignored stimulus, giving a good description of the experimental data.

Predicting performance from functional imaging data: methods matter

In the standard approach to functional imaging studies, brain-behavior relationships are studied by contrasting data obtained during different behavioral states. It is generally assumed that relative change yields meaningful data about relevant brain processes, and that the magnitude of the change reflects the extent of a region's involvement in the behavior being studied. The present study takes a different approach by asking the question, Can functional imaging data predict performance? Regional cerebral blood flow was measured using positron emission tomography in a group of 13 right-handed, normal volunteers during speech production and quiet baseline. A number of methodological assumptions were addressed by examining the relationships between different imaging measures derived from the same raw data and performance on the speech task. The results demonstrate that several common assumptions are not necessarily true. First, although measures based on "activated" scans alone had predictive value with respect to speech rate, measures based on contrasts between "baseline" and "activated" states did not. This was true regardless of whether the contrast was based on subtraction or covariance analyses. Second, while many regions demonstrated large signal increases during speech, speech rate could be predicted by a linear combination of data from two regions, neither of which had the highest "activation" peak, and one of which had a negative relationship with performance. The results demonstrate that contrasting experimental conditions do not necessarily isolate or enhance brain activity related to performance, and that the current assumptions about activation in functional imaging need to be reconsidered.

How well do we understand the neural origins of the fMRI BOLD signal?

The successful use of functional magnetic resonance imaging (fMRI) as a way of visualizing cortical function depends largely on the important relationships between the signal observed and the underlying neuronal activity that it is believed to represent. Currently, a relatively direct correlation seems to be favoured between fMRI signals and population synaptic activity (including inhibitory and excitatory activity), with a secondary and potentially more variable correlation with cellular action potentials.

Interpreting PET and fMRI measures of functional neural activity: the effects of synaptic inhibition on cortical activation in human imaging studies

Human brain imaging methods such as postiron emission tomography and functional magnetic resonance imaging have recently achieved widespread use in the study of both normal cognitive processes and neurological disorders. While many of these studies have begun to yield important insights into human brain function, the relationship between these measurements and the underlying neuronal activity is still not well understood. One open question is how neuronal inhibition is reflected in these imaging results. In this paper, we describe how large-scale modeling can be used to address this question. Specifically, we identify three factors that may play a role in how inhibition affects imaging results: (1) local connectivity; (2) context; and (3) type of inhibitory connection. Simulation results are presented that show how the interaction among these three factors can explain seemingly contradictory experimental results. The modeling suggests that neuronal inhibition can raise brain imaging measures if there is either low local excitatory recurrence or if the region is not otherwise being driven by excitation. Conversely, with high recurrence or actively driven excitation, inhibition can lower observed values.

A neural model of working memory processes in normal subjects, Parkinson's disease and schizophrenia for fMRI design and predictions

A computational model was previously developed to investigate the role of parallel basal ganglia-thalamocortical loops in solving tasks that rely on working memory. Different lesions are applied to the model in order to investigate the working memory deficits observed in Parkinson's disease and schizophrenia. The simulations predict that the working memory deficits observed in Parkinson's disease result from a local dysfunction within the brain due to a problem in the disinhibitory process arising from the basal ganglia. They also predict that the working memory deficits observed in schizophrenia involve many cortical and subcortical areas and result from a problem in selecting items in working memory which are stored in basal ganglia-thalamocortical loops. The simulations predict the temporal unfolding of neuronal activity in different brain regions, both in the normal case and in the two disease states. A specific event-related functional magnetic resonance imaging study was elaborated to test some of those predictions.

Imaging and neural modelling in episodic and working memory processes

Neuroimaging studies using positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) have revealed the involvement of distributed brain regions in memory processes mainly by the use of subtraction strategy based data analyses. Covariance analysis based data analysis strategies have been introduced more recently which allow functional interactions between brain regions of a neuronal network to be assessed. This contribution focuses on studies aiming to (1) establish the functional topography of episodic and working memory processes in young and old normal volunteers, (2) to assess functional interactions between modules of networks of brain regions by means of covariance based analyses and systems level modelling, (3) to characterise the temporal dynamics by the use of magnetoencephalography (MEG) and (4) to relate neuroimaging data to the underpinning neural networks. Male normal young and old volunteers without neurological or psychiatric illness participated in neuroimaging studies (PET, fMRI, MEG). Studies were approved by the ethical committee and federal authorities. Our results in young volunteers show distributed brain areas that are involved in memory processes (episodic and working memory) and show much of an overlap with respect to the network components. Systems level modelling analyses support the hypothesis of bihemispheric, asymmetric networks subserving memory processes and revealed both similarities in general and differences in the interactions between brain regions during episodic encoding and retrieval as well as working memory. Changes in memory function with ageing are evident from functional topographic studies in old volunteers activating more brain regions as compared to young volunteers. There are more and stronger influences of prefrontal regions in elderly volunteers comparing the functional models between old and young subjects. We discuss the way that the systems level models of the PET and fMRI results have implications for the underlying neural network functioning of the brain. This is done by developing simplifying assumptions, which lead from the equations describing the activities of the coupled neural modules to the systems level model equations. The resulting implications for the neural interactions are then discussed, in terms of a set of synaptically coupled neural modules. Finally, we consider how a similar analysis could be extended from the spatial to the temporal domain thus including the EEG and MEG results. The implication of preliminary MEG results presented here for the temporality arising in the interaction between the coupled neural modules in a working memory paradigm is discussed in terms of the previously developed neural network models arising from the PET and fMRI data.

Decomposing memory: functional assignments and brain traffic in paired word associate learning

The recent covariance structural equation model for word-pair associate encoding and retrieval (Krause, Horwitz, Taylor, Schmidt, Mottaghy, Halsband et al., 1998; Krause, Horwitz, Taylor, Schmidt, Mottaghy, Herzog et al., 1999) is analysed to deduce possible functional assignments of the various brain modules used by subjects in solving the task. Specific processing aspects are considered, in particular, that of long-term working memory sites and how they are coupled to buffer working memory sites to enable deposition and manipulation of remembered associates. The new concept of 'brain traffic' is introduced as an aid to the assessment of how important are various brain modules. A set of functional assignments is produced for the relevant modules.

Neural modeling and functional brain imaging: an overview

This article gives an overview of the different functional brain imaging methods, the kinds of questions these methods try to address and some of the questions associated with functional neuroimaging data for which neural modeling must be employed to provide reasonable answers.

Synthetic brain imaging: grasping, mirror neurons and imitation

The article contributes to the quest to relate global data on brain and behavior (e.g. from PET, Positron Emission Tomography, and fMRI. functional Magnetic Resonance Imaging) to the underpinning neural networks. Models tied to human brain imaging data often focus on a few "boxes" based on brain regions associated with exceptionally high blood flow, rather than analyzing the cooperative computation of multiple brain regions. For analysis directly at the level of such data, a schema-based model may be most appropriate. To further address neurophysiological data, the Synthetic PET imaging method uses computational models of biological neural circuitry based on animal data to predict and analyze the results of human PET studies. This technique makes use of the hypothesis that rCBF (regional cerebral blood flow) is correlated with the integrated synaptic activity in a localized brain region. We also describe the possible extension of the Synthetic PET method to fMRI. The second half of the paper then exemplifies this general research program with two case studies, one on visuo-motor processing for control of grasping (Section 3 in which the focus is on Synthetic PET) and the imitation of motor skills (Sections 4 and 5, with a focus on Synthetic fMRI). Our discussion of imitation pays particular attention to data on the mirror system in monkey (neural circuitry which allows the brain to recognize actions as well as execute them). Finally, Section 6 outlines the immense challenges in integrating models of different portions of the nervous system which address detailed neurophysiological data from studies of primates and other species; summarizes key issues for developing the methodology of Synthetic Brain Imaging; and shows how comparative neuroscience and evolutionary arguments will allow us to extend Synthetic Brain Imaging even to language and other cognitive functions for which few or no animal data are available.

Connectivity and complexity: the relationship between neuroanatomy and brain dynamics

Nervous systems facing complex environments have to balance two seemingly opposing requirements. First, there is a need quickly and reliably to extract important features from sensory inputs. This is accomplished by functionally segregated (specialized) sets of neurons, e.g. those found in different cortical areas. Second, there is a need to generate coherent perceptual and cognitive states allowing an organism to respond to objects and events, which represent conjunctions of numerous individual features. This need is accomplished by functional integration of the activity of specialized neurons through their dynamic interactions. These interactions produce patterns of temporal correlations or functional connectivity involving distributed neuronal populations, both within and across cortical areas. Empirical and computational studies suggest that changes in functional connectivity may underlie specific perceptual and cognitive states and involve the integration of information across specialized areas of the brain. The interplay between functional segregation and integration can be quantitatively captured using concepts from statistical information theory, in particular by defining a measure of neural complexity. Complexity measures the extent to which a pattern of functional connectivity produced by units or areas within a neural system combines the dual requirements of functional segregation and integration. We find that specific neuroanatomical motifs are uniquely associated with high levels of complexity and that such motifs are embedded in the pattern of long-range cortico-cortical pathways linking segregated areas of the mammalian cerebral cortex. Our theoretical findings offer new insight into the intricate relationship between connectivity and complexity in the nervous system.