Skip to main content
main-content

Über dieses Buch

This book constitutes the refereed proceedings of the International Conference on Brain Informatics, BI 2018, held in Arlington, TX, USA, in December 2018. The 46 revised full papers were carefully reviewed and selected from 53 submissions. The papers are grouped thematically on cognitive and computational foundations of brain science, human information processing systems, brain big data analysis, curation and management, informatics paradigms for brain and mental health research, brain-machine intelligence and brain-inspired computing.

Inhaltsverzeichnis

Frontmatter

Cognitive and Computational Foundations of Brain Science

Frontmatter

Emotion Recognition Based on Gramian Encoding Visualization

This paper addresses the problem that emotional computing is difficult to be put into real practical fields intuitively, such as medical disease diagnosis and so on, due to poor direct understanding of physiological signals. In view of the fact that people’s ability to understand two-dimensional images is much higher than one-dimensional signals, we use Gramian Angular Fields to visualize time series signals. GAF images are represented as a Gramian matrix where each element is the trigonometric sum between different time intervals. Then we use Tiled Convolutional Neural Networks (tiled CNNs) on 3 real world datasets to learn high-level features from GAF images. The classification results of our method are better than the state-of-the-art approaches. This method makes visualization based emotion recognition become possible, which is beneficial in the real medical fields, such as making cognitive disease diagnosis more intuitively.

Jie-Lin Qiu, Xin-Yi Qiu, Kai Hu

EEG Based Brain Mapping by Using Frequency-Spatio-Temporal Constraints

In this paper an improvement of the dynamic inverse problem solution is proposed by using constraints in the space-time-frequency domain. The method is based on multi-rate filter banks for frequency selection of the EEG signals and a cost function that includes spatial and temporal constraints. As a result, an iterative method which includes Frequency-Spatio-temporal constraints is proposed. The performance of the proposed method is evaluated by using simulated and real EEG signals. It can be concluded that the enhanced IRA-L1 method with the frequency-spatio-temporal stage improves the quality of the brain reconstruction performance in terms of the Wasserstein metric, in comparison with the other methods, for both simulated and real EEG signals.

Pablo Andrés Muñoz-Gutiérrez, Juan David Martinez-Vargas, Sergio Garcia-Vega, Eduardo Giraldo, German Castellanos-Dominguez

An EEG-Based Emotion Recognition Model with Rhythm and Time Characteristics

As a senior function of human brain, emotion has a great influence on human study, work, and all aspects of life. Correctly recognizing human emotion can make artificial intelligence serve human being better. EEG-based emotion recognition (ER) has become more popular in these years, which is one of the utilizations of Brain Computer Interface (BCI). However, due to the ambiguity of human emotions and the complexity of EEG signals, the EEG-ER system which can recognize emotions with high accuracy is not easy to achieve. In this paper, based on the time scale, we choose recurrent neural network as the breakthrough point of the screening model. And according to the rhythmic characteristics and temporal memory characteristics of EEG, we propose a Rhythmic Time EEG Emotion Recognition Model (RT-ERM) based on the valance and arousal of LSTM. When using this model, the classification results of different rhythms and time scales are different. Through the results of the classification accuracy of different rhythms and different time scales, the optimal rhythm and time scale of the RT-ERM model are obtained, and the classification of emotional EEG is carried out by the best time scales corresponding to different rhythms, and we found some interesting phenomena. Finally, by comparing with other existing emotional EEG classification methods, it is found that the rhythm and time scale of the model can provide a good accuracy rate for RT-ERM.

Jianzhuo Yan, Sinuo Deng

Influence of Realistic Head Modeling on EEG Forward Problem

We study the influence of realistic head modeling on the EEG forward problem. To this end, we define a high-resolution patient-specific realistic head model from 3T-MRI and DWI data. Further, we performed a nine tissues segmentation and a white matter anisotropy estimation. Then we solve the forward problem using a state-of-the-art FDM solution that allows volumetric voxelwise anisotropic definition in a reciprocity sensors space. Finally, we compared the 9-tissues realistic head model against the commonly used 5-tissues isotropic representation. Our results show significant potential deviations due to the white matter anisotropy, and radio to tangential patterns in the outer skull regions as a direct effect of essential tissues like fat or muscle. Moreover, we analyze the dipole estimation errors in a parametric inverse setup, finding DLE’s larger than 20 mm. Additionally, we study the influence of neglecting the blood vessels, finding DLE’s larger than 4 mm in deep brain areas.

Ernesto Cuartas Morales, Yohan Ricardo Céspedes Villar, Héctor Fabio Torres Cardona, Carlos Daniel Acosta, German Castellanos Dominguez

Computational Model for Reward-Based Generation and Maintenance of Motivation

In this paper, a computational model for the motivation process is presented that takes into account the reward pathway for motivation generation and associative learning for maintaining motivation through Hebbian learning approach. The reward prediction error is used to keep motivation maintained. These aspects are backed by recent neuroscientific models and literature. Simulation experiments have been performed by creating scenarios for student learning through rewards and controlling their motivation through regulation. Mathematical analysis is provided to verify the dynamic properties of the model.

Fawad Taj, Michel C. A. Klein, Aart van Halteren

Current Design with Minimum Error in Transcranial Direct Current Stimulation

As a non-invasive brain stimulation technology, transcanial direct current stimulation (tDCS) has been recently attracting more and more attention in research and clinic applications due to its convenient implementation and modulation of the brain functionality. In this paper, we propose a novel multi-electrode tDCS current configuration model that minimizes the total error under the safety constraints. After rewriting the model as a linearly constrained minimization problem, we develop an efficient numerical algorithm based on the alternating direction method of multipliers (ADMM). Numerical experiments have shown the great potential of the proposed method in terms of accuracy and focality.

Jing Qin, Yushan Wang, Wentai Liu

Assessment of Source Connectivity for Emotional States Discrimination

In this paper a novel methodology for assessing source connectivity applied to emotional states discrimination is proposed. The method involves (i) designing the set of Regions-of-interest (ROIs) over the cortical surface, (ii) estimating the ROI time-courses using a dynamic inverse problem formulation, (iii) estimating the pairwise functional connectivity between ROIs, and (iv) feeding a Support Vector Machine Classifier with the estimated connectivity to discriminate between emotional states. The performance of the proposed methodology is evaluated over a real database where obtained results improve state-of-the-art methods that either compute connectivity between pairs of EEG channels or do not consider the non-stationary nature of the EEG data.

J. D. Martinez-Vargas, D. A. Nieto-Mora, P. A. Muñoz-Gutiérrez, Y. R. Cespedes-Villar, E. Giraldo, G. Castellanos-Dominguez

Rich Dynamics Induced by Synchronization Varieties in the Coupled Thalamocortical Circuitry Model

Epileptic disorders are typically characterized by the synchronous spike-wave discharges (SWD). However, the mechanism of SWD is not well-understood in terms of its synchronous spatio-temporal features. In this paper, based on the coupled thalamocortical (TC) neural field models we first investigate the SWD complete synchronization (CS), lag synchronization (LS) and anticipated synchronization (AS) mainly using the adaptive delayed feedback (ADF) and active control (AC). Then we explore the dynamics of 3-compartment coupled TC motifs with the interactive connectivity patterns of ADF and AC, as well as the various interactive weights. It is found that CS, LS and AS of motifs can coexist and transit between each other by changing the various interactive modes and weights. These results provide the complementary synchronization effects and conditions for the basic 3-node motifs. This may facilitate to construct the architecture based on patient EEG data and reveal the abnormal information expression of epileptic oscillatory network.

Denggui Fan, Jianzhong Su, Ariel Bowman

Human Information Processing Systems

Frontmatter

Humans Have a Distributed, Molecular Long-Term Memory

Most memory research has assumed that our long-term memories are somehow retained in our brain, usually by modified synaptic connections. This paper proposes a very different scenario, in which the basic substrate of these memories are molecules which flow within a newly discovered circulatory system similar to our lymph system. Moreover, the information bearing molecules are postulated to be cyclic protein polymers similar to the proteins found in all cell membranes.Two network algorithms are presented which convert networks to, and from, such cyclic structures and seem to mimic the psychological processes of consolidation, recall, and reconsolidation.

John L. Pfaltz

Functional Connectivity Analysis Using the Oddball Auditory Paradigm for Attention Tasks

Nowadays, cognitive stimulus processing using Electroencephalographic (EEG) recordings is accomplished by analyzing individually the time-frequency information belonging to each EEG channel. Nevertheless, several studies have characterized cognitive functions as synchronized brain networks depending on the underlying neural interactions. As a result, connectivity analysis provides essential information for improving both the interpretation and interpretability of brain functionality under specific tasks. In this research, we perform functional connectivity analysis by measuring the stability of the phase difference between EEG channels, aiming to include synchronization patterns for studying the brain reaction to cognitive stimulus. Experiments are carried out in subjects responding to an oddball paradigm. Results show statistical differences between target and non-target labels, making the proposed methodology a suitable alternative to support cognitive neurophysiological applications.

Juana Valeria Hurtado-Rincón, Francia Restrepo, Jorge Ivan Padilla, Hector Fabio Torres, German Castellanos-Dominguez

Perspective Taking vs Mental Rotation: CSP-Based Single-Trial Analysis for Cognitive Process Disambiguation

Mental Rotation (i.e. the ability to mentally rotate representations of 2D and 3D objects) and egocentric Perspective Taking (i.e. the ability to adopt an imagined spatial perspective) represent the two most well-known and used types of spatial transformation. Yet, these two spatial transformations are conceptually, visually, and mathematically equivalent. Thus, an active debate in the field is whether these two types of spatial transformations are cognitively and neurally distinct or whether they represent different manifestation of the same underlying core mental process. In this study, we utilize a machine learning approach to extract neural activity from electroencephalography (EEG) measures and identify neural differences between mental rotation and perspective taking tasks. Our results provide novel empirical evidence in support of the view that these two types of spatial transformation correspond to district cognitive processes at the neural level. Importantly, the proposed framework provides a novel approach that can facilitate the study of the neural correlates of spatial cognition.

Christoforos Christoforou, Adamantini Hatzipanayioti, Marios Avraamides

Using the Partial Directed Coherence to Understand Brain Functional Connectivity During Movement Imagery Tasks

We propose to use the partial directed coherence (PDC) to analyze the coupling between pairs of electroencephalographic (EEG) measurements during movement imagery tasks, as well as the directionality of such coupling. For this, we consider the multivariate autoregressive model of the signals from a selection of eleven EEG channels that are assumed as a fully-connected measurement network. Then, we aim to find differences in connectivity patterns between motor imagery and resting state that arise in a brain-computer interface (BCI) system with visual feedback that controls the movement of a robot. Our preliminary results show that it is possible to relate the changes in the magnitude of the PDC to different connectivity patterns in the measurement network we have considered, and those changes are in agreement with brain functional connectivity that has been reported in other studies based mainly in magnetic resonance imaging.

Myriam Alanis-Espinosa, David Gutiérrez

Combining fMRI Data and Neural Networks to Quantify Contextual Effects in the Brain

Does word meaning change according to the context? Although this hypothesis has existed for a long time, only recently it has become possible to test it based on neuroimaging. Embodiment theories of knowledge representation suggest that word meaning consist of a collection of attributes defined in terms of various neural systems. This approach represents an unlimited number of objects through weighted attributes and the weights may change in context. This paper aims at quantifying such dynamic meanings using computational modeling. A neural network is trained with backpropagation to map attribute-based representations to fMRI images of subjects reading everyday sentences. Backpropagation is then extended to the features, demonstrating how they change in different sentence contexts for the same word. Indeed, statistically significant changes occurred across similar contexts and across different subjects, quantifying for the first time how attribute weightings for the same word are modified by context. Such dynamic representations of meaning could be used in future natural language processing systems, allowing them to mirror human performance more accurately.

Nora Aguirre-Celis, Risto Miikkulainen

Network Analysis of Brain Functional Connectivity in Mental Arithmetic Using Task-Evoked fMRI

Mental arithmetic is the complete use of brain functions to complete the basic arithmetic process without the aid of other tools and equipment. Neuroimaging studies of mental arithmetic have revealed some brain regions and networks associated with them. However, there are still many unsolved problems about the brain function network structure in mental arithmetic. We designed a mental arithmetic experiment consisting of four experimental conditions, and a group of 21 subjects were recruited in the experiment. The collected fMRI data was used to construct the brain functional connectivity network with atlas of Dosenbach-160. We used graph theoretic based network analysis method to calculate the small world attributes and network efficiencies of each subject’s brain functional network, and tested the statistical differences between the experimental conditions. The results show that, when the human brain performs addition or subtraction, the functional connectivity network statistically shows a significant characteristic of small world from the resting state. And experiment condition of resting has a higher clustering coefficient over a continuous graph density than number matching, showing a more significant small world property. Results of network efficiency show that resting has slightly higher network-wide information exchange productivity than number matching. Furthermore, due to the limitations of the size of the recruited subjects, the report results can only be an exploratory attempt, which needs to be verified by a larger sample of data in the future.

Xiaofei Zhang, Yang Yang, Ming-Hui Zhang, Ning Zhong

The Astrocytic Microdomain as a Generative Mechanism for Local Plasticity

Mounting experimental evidence suggests that astrocytes have an active role in synaptic modification. A central premise is that they modify the structure and the function of the neuronal network but the underlying mechanisms for doing so remain elusive. Here, we developed a biophysically constrained 2D compartmental model of an astrocytic microdomain that suggests an explanation for the recently reported functional clustering of synapses. Our model followed the typical geometrical structure of astrocytes, comprising of functionally independent microdomains, and the spatial allocation of their sub-cellular organelles giving rise to (a) fast, process-specific and (b) delayed, microdomain-wide calcium waves. These waves encoded the neuronal activity into their spatial extent and interacted with each other to impose locally restricted synaptic weight modifications constrained in the microdomain. Our results give a possible explanation for the recently reported spatially clustered functional groups in dendritic spines, advocating the astrocytic microdomain as a fundamental learning unit in the brain.

Ioannis Polykretis, Vladimir Ivanov, Konstantinos P. Michmizos

Determining the Optimal Number of MEG Trials: A Machine Learning and Speech Decoding Perspective

Advancing the knowledge about neural speech mechanisms is critical for developing next-generation, faster brain computer interface to assist in speech communication for the patients with severe neurological conditions (e.g., locked-in syndrome). Among current neuroimaging techniques, Magnetoencephalography (MEG) provides direct representation for the large-scale neural dynamics of underlying cognitive processes based on its optimal spatiotemporal resolution. However, the MEG measured neural signals are smaller in magnitude compared to the background noise and hence, MEG usually suffers from a low signal-to-noise ratio (SNR) at the single-trial level. To overcome this limitation, it is common to record many trials of the same event-task and use the time-locked average signal for analysis, which can be very time consuming. In this study, we investigated the effect of the number of MEG recording trials required for speech decoding using a machine learning algorithm. We used a wavelet filter for generating the denoised neural features to train an Artificial Neural Network (ANN) for speech decoding. We found that wavelet based denoising increased the SNR of the neural signal prior to analysis and facilitated accurate speech decoding performance using as few as 40 single-trials. This study may open up the possibility of limiting MEG trials for other task evoked studies as well.

Debadatta Dash, Paul Ferrari, Saleem Malik, Albert Montillo, Joseph A. Maldjian, Jun Wang

Brain Big Data Analytics, Curation and Management

Frontmatter

Use of Temporal Attributes in Detection of Functional Areas in Basal Ganglia

Basal ganglia are a target for deep brain stimulation (DBS) in various neurological disorders like Parkinson’s Disease (PD) or Dystonia. Due to the complex excitatory and inhibitory interactions between various components of basal ganglia, it is often as much important to stimulate certain regions as it is not to stimulate others. Such is the case in DBS surgery for PD where the goal is the stimulation of the Subthalamic Nucleus (STN) while the Substantia Nigra Pars reticulata (SNr) should not be stimulated. In this paper it is shown that use of temporal attributes extracted from microrecordings acquired during DBS procedure not only allows for better detection of the STN itself but also helps to prevent false positive identification of the SNr recordings as the STN ones.

Konrad A. Ciecierski

Influence of Time-Series Extraction on Binge Drinking Interpretability Using Functional Connectivity Analysis

Brain connectivity analysis has gained considerable importance in different cognitive tasks and the detection of pathological conditions. Despite recent advances in connectivity analysis, there are still problems to be solved, being a proper extraction of the time-series to characterize the regions of interest (ROI) one of the challenges. In this work, we examine the influence of the time-varying mean estimation on the brain connectivity analysis for control and binge drinkers subjects. The obtained results show that the performance of brain connectivity improves using the eigenvalue-based averaging since it may face better the nonstationarity behavior and inter-trial variability of MEG activity.

J. I. Padilla-Buriticá, H. F. Torres, E. Pereda, A. Correa, G. Castellanos-Domínguez

Regularized State Observers for Source Activity Estimation

The brain is a complex system and the activity inside can describe non-linear behaviors where the signals of the EEG which are taken from the scalp represent the mixture of the activity in each distributed source inside the brain. This activity can be represented by non-linear models and the inverse problem for source activity estimation can consider these models in the solutions. This paper presents the design of linear and nonlinear regularized observers for neural activity estimation, where the solutions involve a discrete physiologically-based non-linear model as spatio-temporal constraints. Furthermore, this document presents the estimation of the regularization hyper-parameters based on the application of a genetic algorithm over the Generalized Cross Validation cost function, which reduced the computational load. The aforementioned methods are compared with Multiple Sparse Priors (MSP) method of the state-of-the-art by using a simulated and real EEG signals.

Andrés Felipe Soler, Pablo Andrés Muñoz-Gutiérrez, Eduardo Giraldo

Distributional Representation for Resting-State Functional Brain Connectivity Analysis

Most analyses on functional brain connectivity across a group of brains are under the assumption that the positions of the voxels are aligned into a common space. However, the alignment errors are inevitable. To address such issue, a distributional representation for resting-state functional brain connectivity is proposed here. Unlike other relevant connectivity analyses that only consider connections with higher correlation values between voxels, the distributional approach takes the whole picture. The spatial structure of connectivity is captured by the distance between voxels so that the relative position information is preserved. The distributional representation can be visualized to find outliers in a large dataset. The centroid of a group of brains is discovered. The experimental results show that resting-state brains are distributed on the ‘orbit’ around their categorical centroid. In contrast to the main-stream representation such as selected network properties for disease classification, the proposed representation is task-free, which provides a promising foundation for further analysis on functional brain connectivity in various ends.

Jiating Zhu, Jiannong Cao

Deep Neural Networks for Automatic Classification of Anesthetic-Induced Unconsciousness

Despite the common use of anesthetics to modulate consciousness in the clinic, brain-based monitoring of consciousness is uncommon. We combined electroencephalographic measurement of brain activity with deep neural networks to automatically discriminate anesthetic states induced by propofol. Our results with leave-one-participant-out-cross-validation show that convolutional neural networks significantly outperform multilayer perceptrons in discrimination accuracy when working with raw time series. Perceptrons achieved comparable accuracy when provided with power spectral densities. These findings highlight the potential of deep convolutional networks for completely automatic extraction of useful spatio-temporo-spectral features from human EEG.

Konstantinos Patlatzoglou, Srivas Chennu, Mélanie Boly, Quentin Noirhomme, Vincent Bonhomme, Jean-Francois Brichant, Olivia Gosseries, Steven Laureys

Efficient and Automatic Subspace Relevance Determination via Multiple Kernel Learning for High-Dimensional Neuroimaging Data

Alzheimer’s disease is a major cause of dementia. Its pathology induces complex spatial patterns of brain atrophy that evolve as the disease progresses. The diagnosis requires accurate biomarkers that are sensitive to disease stages. Probabilistic biomarkers naturally support the interpretation of decisions and evaluation of uncertainty associated with them. We obtain probabilistic biomarkers via Gaussian Processes, which also offer flexible means to accomplish Multiple Kernel Learning. Exploiting this flexibility, we propose a novel solution, Multiple Kernel Learning for Automatic Subspace Relevance Determination, to tackle the challenges of working with high-dimensional neuroimaging data. The proposed Gaussian Process models are competitive with or better than the well-known Support Vector Machine in terms of classification performance even in the cases of single kernel learning. Also, our method improves the capability of the Gaussian Process models and their interpretability in terms of the known anatomical correlates of the disease.

Murat Seçkin Ayhan, Vijay Raghavan

Improving SNR and Reducing Training Time of Classifiers in Large Datasets via Kernel Averaging

Kernel methods are of growing importance in neuroscience research. As an elegant extension of linear methods, they are able to model complex non-linear relationships. However, since the kernel matrix grows with data size, the training of classifiers is computationally demanding in large datasets. Here, a technique developed for linear classifiers is extended to kernel methods: In linearly separable data, replacing sets of instances by their averages improves signal-to-noise ratio (SNR) and reduces data size. In kernel methods, data is linearly non-separable in input space, but linearly separable in the high-dimensional feature space that kernel methods implicitly operate in. It is shown that a classifier can be efficiently trained on instances averaged in feature space by averaging entries in the kernel matrix. Using artificial and publicly available data, it is shown that kernel averaging improves classification performance substantially and reduces training time, even in non-linearly separable data.

Matthias S. Treder

Automatic Recognition of Resting State fMRI Networks with Dictionary Learning

Resting state functional magnetic resonance imaging (rs-fMRI) is a functional neuroimaging technique that investigates the spatially remote yet functionally linked neuronal coactivation patterns of the brain at rest. Non-invasiveness and task-free characteristics of rs-fMRI make it particularly suitable for aging, pediatric and clinical population. Researchers typically follow a source separation strategy to efficiently reconstruct the concurrent interacting resting state networks (RSN) from a myriad of whole brain fMRI signals. RSNs are currently identified by visual inspection with prior knowledge of spatial clustering of RSNs, as the variability and spatial overlapping nature of RSNs combined with presence of various sources of noise make automatic identification of RSNs a challenging task. In this study, we have developed an automated recognition algorithm to classify all the distinct RSNs. First, in contrast to traditional single level decomposition, a multi-level deep sparse matrix factorization-based dictionary leaning strategy was used to extract hierarchical features from the data at each level. Then we used maximum likelihood estimates of these spatial features using Kullback-Leibler divergence to perform the recognition of RSNs. Experimental results confirmed the effectiveness of our proposed approach in accurately classifying all the RSNs.

Debadatta Dash, Bharat Biswal, Anil Kumar Sao, Jun Wang

Simultaneous EEG Analysis and Feature Extraction Selection Based on Unsupervised Learning

Time-series EEG signals in a raw form are challenging to analyze, train, and compute. Several feature extraction methods, such as fast Fourier transform, wavelet transform, and time-frequency distributions, are commonly employed for this purpose. However, when applied to different datasets, the alignment between the method and machine learning algorithms varies significantly. Through an EEG experiment, we test a simultaneous analysis and unsupervised learning application that can effectively determine what feature extraction method will potentially lead to a higher prediction precision when the ground truth is provided by the participants at a later stage.

Badar Almarri, Chun-Hsi Huang

Construction of Sparse Weighted Directed Network (SWDN) from the Multivariate Time-Series

There are many studies focusing on network detection in multivariate (MV) time-series data. A great deal of focus have been on estimation of brain networks using functional Magnetic Resonance Imaging (fMRI), functional Near-Infrared Spectroscopy (fNIRS) and electroencephalogram (EEG). We present a sparse weighted directed network (SWDN) estimation approach which can detect the underlying minimum spanning network with maximum likelihood and estimated weights based on linear Gaussian conditional relationship in the MV time-series. Considering the brain neuro-imaging signals as the multivariate data, we evaluated the performance of the proposed approach using the publicly available fMRI data-set and the results of the similar study which had evaluated popular network estimation approaches on the simulated fMRI data.

Rahilsadat Hosseini, Feng Liu, Shouyi Wang

Preference Prediction Based on Eye Movement Using Multi-layer Combinatorial Fusion

Face image preference is influenced by many factors and can be detected by analyzing eye movement data. When comparing two face images, our gaze shifts within and between the faces. Eye tracking data can give us insights into the cognitive processes involved in forming a preference. In this paper, a gaze tracking dataset is analyzed using three machine learning algorithms (MLA): AdaBoost, Random Forest, and Mixed Group Ranks (MGR) as well as a newly developed machine learning framework called Multi-Layer Combinatorial Fusion (MCF) to predict a subject’s face image preference. Attributes constructed from the dataset are treated as input scoring systems. MCF involves a series of layers that consist of expansion and reduction processes. The expansion process involves performing exhaustive score and rank combinations, while the reduction process uses performance and diversity to select a subset of systems that will be passed onto the next layer of analysis. Performance and cognitive diversity are used in weighted scoring system combinations and system selection. The results outperform the Mixed Group Ranks algorithm, as well as our previous work using pairwise scoring system combinations.

Christina Schweikert, Louis Gobin, Shuxiao Xie, Shinsuke Shimojo, D. Frank Hsu

A Multilayer Network Approach for Studying Creative Ideation from EEG

The neural mechanisms underlying creative ideation are not clearly understood owing to the widespread cognitive processes involved in the brain. Current research states alpha band’s relation to creative ideation, as the most consistent finding. However, creative ideation appear at the signal level within multiple frequency bands and cross-frequency coupling phenomenon. To address this issue, we analyzed both within band and cross-frequency functional connectivity in a single framework using multilayer network. To further investigate the time evolution of creative thinking, we performed the analysis for three phases (early, middle and later). The experimental design used in this study consists of divergent thinking as an indicator of creativity where the subjects were instructed to give alternative uses of an object. As a control task, convergent thinking was used where the subjects were asked to list typical characteristics of an object. We evaluated global and nodal metrics (i.e., clustering coefficient, local efficiency, and nodal degree) for the three phases. Each metric was calculated separately for within band (intra layer) and cross-frequency (inter layer) connectivity. Paired t-test results showed significant difference in the later phase for both inter layer clustering coefficient and inter layer local efficiency. In nodal metrics, significant difference was observed in the later phase for intra layer degree and in all the phases for inter layer degree. The results from this study demonstrate that both the cross-frequency coupling and within-band connectivity can reveal more information regarding the neural processes related to creative ideation.

Rohit Bose, Kumar Ashutosh, Junhua Li, Andrei Dragomir, Nitish Thakor, Anastasios Bezerianos

Estimating Latent Brain Sources with Low-Rank Representation and Graph Regularization

To infer latent brain source activation patterns under different cognitive tasks is an integral step to understand how our brain works. Traditional electroencephalogram (EEG) Source Imaging (ESI) methods usually do not distinguish task-related and spurious non-task-related sources that jointly generate EEG signals, which inevitably yield misleading reconstructed activation patterns. In this research, we assume that the task-related source signal intrinsically has a low-rank property, which is exploited to infer the true task-related EEG sources location. Although the true task-related source signal is sparse and low-rank, the contribution of spurious sources scattering over the source space with intermittent activation patterns makes the actual source space lose the low-rank property. To reconstruct a low-rank true source, we propose a novel ESI model that involves a spatial low-rank representation and a temporal Laplacian graph regularization, the latter of which guarantees the temporal smoothness of the source signal and eliminate the spurious ones. To solve the proposed model, an augmented Lagrangian objective function is formulated and an algorithm in the framework of alternating direction method of multipliers (ADMM) is proposed. Numerical results illustrate the effectivenesks of the proposed method in terms of reconstruction accuracy with high efficiency.

Feng Liu, Shouyi Wang, Jing Qin, Yifei Lou, Jay Rosenberger

Informatics Paradigms for Brain and Mental Health Research

Frontmatter

Analysis of Epileptic Activity Based on Brain Mapping of EEG Adaptive Time-Frequency Decomposition

The applications of Empirical Mode Decomposition (EMD) in Biomedical Signal analysis have increased and is common now to find publications that use EMD to identify behaviors in the brain or heart. EMD has shown excellent results in the identification of behaviours from the use of electroencephalogram (EEG) signals. In addition, some advances in the computer area have made it possible to improve their performance. In this paper, we presented a method that, using an entropy analysis, can automatically choose the relevant Intrinsic Mode Functions (IMFs) from EEG signals. The idea is to choose the minimum number of IMFs to reconstruct the brain activity. The EEG signals were processed by EMD and the IMFs were ordered according to the entropy cost function. The IMFs with more relevant information are selected for the brain mapping. To validate the results, a relative error measure was used.

Maximiliano Bueno-López, Pablo A. Muñoz-Gutiérrez, Eduardo Giraldo, Marta Molinas

Resting State EEG Based Depression Recognition Research Using Deep Learning Method

Deep learning has obtained state-of-the-art performance in many fields with its powerful ability of representation learning. However, unlike other data, EEG signals have temporal, spatial and frequency characteristics. For the EEG based depression detection, how to preserve these features when EEG signals are fed into neural networks and select appropriate network structure to extract the corresponding inherent patterns is a problem that needs to be solved. Inspired by the application of deep learning in image processing, we used the distance-based projection method and the non-distance projection method to construct EEG signals as inputs of neural networks. Four different networks were used to extract inherent structure from constructed data. As a result, CNN outperformed other networks, with the highest classification accuracy of 77.20% using the non-distance projection method and 76.14% using the distance-based projection method. The results demonstrate that application of deep learning methods in the research of depression is feasible.

Wandeng Mao, Jing Zhu, Xiaowei Li, Xin Zhang, Shuting Sun

Tensor Decomposition for Neurodevelopmental Disorder Prediction

Functional Magnetic Resonance Imaging (fMRI) has been successfully used by the neuroscientists for diagnosis and analysis of neurological and neurodevelopmental disorders. After transforming fMRI data into functional networks, graph classification algorithms have been applied for distinguishing healthy controls from impaired subjects. Recently, classification followed by tensor decomposition has been used as an alternative, since the sparsity of the functional networks is still an open question. In this work, we present five tensor models of fMRI data, considering the time series of the brain regions as the raw form. After decomposing the tensor using CANDECOMP/PARAFAC (CP) and Tucker decomposition, we compared nearest neighbor classification accuracy on the resulting subject factor matrix. We show experimental results using an fMRI dataset from adult subjects with neurodevelopmental reading disabilities and normal controls.

Shah Muhammad Hamdi, Yubao Wu, Soukaina Filali Boubrahimi, Rafal Angryk, Lisa Crystal Krishnamurthy, Robin Morris

Feature Selection and Imbalanced Data Handling for Depression Detection

Major Depressive Disorder (MDD) is the most common disorder worldwide. Accurate detection of depression is a challenging problem. Machine learning-based automated depression detection provides useful assistance to the clinicians for effective depression diagnosis. One of the most fundamental steps in any automated detection is feature selection and investigation of the most relevant features. Studies show that regional volumes of the brain are affected in response to depression. Regional volumes are considered as features. The gray matter volumes’ correlation with depression and the most effective gray volumes for depression detection is investigated in this study. Various feature selection techniques are studied, along with the investigation on the importance of resampling to handle imbalanced data, which is typically the case for depression detection, as the number of depressed instances is commonly a fraction of the entire data size. Experimental results using Random Forests (RF) and support vector machines (SVM) with a Gaussian kernel (RBF) as classifiers show that feature selection followed by data resampling gives superior performance measured by Area Under the ROC Curve (AUC) as well as prediction accuracy, and RF outperforms SVM for the depression detection task.

Marzieh Mousavian, Jianhua Chen, Steven Greening

Deep Convolutional Neural Networks for Automated Diagnosis of Alzheimer’s Disease and Mild Cognitive Impairment Using 3D Brain MRI

We consider the automated diagnosis of Alzheimer’s Disease (AD) and Mild Cognitive Impairment (MCI) in 3D structural MRI brain scans. We develop an efficient deep convolutional neural network (CNN) based classifier by analyzing 3D brain MRI. The proposed model extracts features from the MRI scans and learns significant information related to Alzheimer’s Disease (AD) and Mild Cognitive Impairment (MCI). We perform motion correction, non-uniform intensity normalization, Talairach transform, intensity normalization, and skull-stripping in the raw MRI scans. After that several 2D slices are generated, and center patch is cropped from the slices before passing them to the CNN classifier. Besides, we demonstrate ways to improve the performance of a CNN classifier for AD and MCI diagnosis. We conduct experiments using Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset for classification of the AD, MCI and CN (normal/healthy controls) to evaluate the proposed model. The proposed model achieves 94.97% accuracy for AD/CN classification and 91.98% accuracy for AD/MCI classification outperforming baseline models and several competing methods from other studies.

Jyoti Islam, Yanqing Zhang

Evaluating Mental Health Encounters in mTBI: Identifying Patient Subgroups and Recommending Personalized Treatments

Mild Traumatic Brain Injuries (mTBIs) are “poorly understood” [6] and often associated with psychiatric conditions [21]. While machine learning techniques have explored these comorbidities, the utilization of psychiatric Electronic Health Records (EHRs) poses unique challenges, but provides great promise in the understanding of the brain and the effect of an mTBI [3, 14]. Therefore, in an effort to assist clinical practice in the field of mTBI, we present our work on utilizing EHR in which we apply machine learning models to identify and compare patient subgroups and explore algorithms to recommend patient catered treatment plans. Through this work, we aim to highlight effective techniques for handling the complexities of EHR and psychiatric-specific data.

Filip Dabek, Peter Hoover, Jesus Caban

DeepDx: A Deep Learning Approach for Predicting the Likelihood and Severity of Symptoms Post Concussion

In the United States alone, an estimated 1.7 million traumatic brain injuries (TBIs) occur each year, leading to more than 1.3 million TBI-related emergency room visits and hospitalizations, as well as thousands of deaths [9]. Following a mild TBI (mTBI) or concussion, patients may experience physical, psychological, and cognitive deficits lasting a period of hours, days, or for an extended period of time. Although most people recover within days to weeks of injury, some report persistent symptoms. Early detection, prognosis, and forecast of symptoms have the ability to improve the overall outcome of a patient, reduce the cost associated with treatment, and provide important insights into poorly understood brain injuries. This paper presents a deep learning approach for predicting the onset of a new diagnosis and its severity up to a year post concussion. Through the evaluation of our model we show that with thirty TBI-related symptoms, we are able to correctly predict the onset of a new symptom 93.49(±6.92)% of the time, and when predicting the entire 12-months trajectory of a patient’s symptoms we are able to exceed the expected 13.27% by more than doubling it to 33.53%. In addition, we introduce the concept of a deep recurrent neural network generating sample patients which can be used to derive the different types of patients that exist within the population.

Filip Dabek, Peter Hoover, Jesus Caban

Automatic Detection of Epileptic Spike in EEGs of Children Using Matched Filter

The Electroencephalogram (EEG) is one of the most used tools for diagnosing Epilepsy. Analyzing EEG, neurologists can identify alterations in brain activity associated with Epilepsy. However, this task is not always easy to perform, because of the duration of the EEGs or the subjectivity of the specialist in detecting alterations. Aim: To present an epileptic spike detector based on matched filter for supporting diagnosis of Epilepsy through a tool able to automatically detect spikes in EEG of children. Results: The results of the evaluation showed that the developed detector achieved a sensitivity of 89.28% which is within the range of what has been reported in the literature (82.68% and 94.4%), and a specificity of 99.96%, the later improving the specificity of the best reviewed work. Conclusions: Considering the results obtained in the evaluation, the solution becomes an alternative to support the automatic identification of epileptic spikes by neurologists.

Maritza Mera, Diego M. López, Rubiel Vargas, María Miño

fASSERT: A Fuzzy Assistive System for Children with Autism Using Internet of Things

This work presents an assistive system for child with autism spectrum disorder (C-ASD). The main objective of this system is to reduce dependency on the caregiver and parent and thereby assisting them to make independent. Fuzzy logic based expert system is designed for the assisting system which will help in intervention strategies. The system collects data from four different sensors, such as GPS, heart beat, accelerometer and sound, and generates required notification for the parent, caregiver and C-ASD. The wearables-specifically smart watches- can be used to implement such system. A case study shows the proposed expert system is able to help the C-ASD to restore dysfunction.

Anjum Ismail Sumi, Most. Fatematuz Zohora, Maliha Mahjabeen, Tasnova Jahan Faria, Mufti Mahmud, M. Shamim Kaiser

ADEPTNESS: Alzheimer’s Disease Patient Management System Using Pervasive Sensors - Early Prototype and Preliminary Results

Alzheimer’s is a catastrophic neuro-degenerative state in the elderly which reduces thinking skills and thereby hamper daily activity. Thus the management may be helpful for people with such condition. This work presents sensor based management system for Alzheimer’s patient. The main objective of this work is to report an early prototype of an eventual wearable system that can assist in managing the health of such patients and notify the caregivers in case of necessity. A brief case study is presented which showed that the proposed prototype can detect agitated and clam states of patients. As the ultimately developed assistive system will be packaged as a wearable device, the case study also investigated the usability of wearable devices on different age groups of Alzheimer’s patients. In addition, electro dermal activity for 4 patient of age group 55–60 and 60-7s years were also explored to assess the health condition of the patients.

Tajim Md. Niamat Ullah Akhund, Md. Julkar Nayeen Mahi, A. N. M. Hasnat Tanvir, Mufti Mahmud, M. Shamim Kaiser

Uncovering Dynamic Functional Connectivity of Parkinson’s Disease Using Topological Features and Sparse Group Lasso

Neuro-degenerative diseases such as Parkinson’s Disease (PD) are clinically found to cause alternations and failures in brain connectivity. In this work, a new classification framework using dynamic functional connectivity and topological features is proposed, and it is shown that such framework can give better insights over discriminative difference of the disease itself. After utilizing sparse group lasso with anatomically labeled resting-state fMRI signal, both discriminating brain regions and voxels within can be identified easily. To give an overview of the effectiveness of such framework, the classification performance with the network features extracted on dynamic functional network is quantitatively evaluated. Experimental results show that either single feature of clustering coefficient or combined feature group of characteristic path length, diameter, eccentricity and radius perform well in classifying PD, and as a result the identified feature can lead to better interpretation for clinical purposes.

Kin Ming Puk, Wei Xiang, Shouyi Wang, Cao (Danica) Xiao, W. A. Chaovalitwongse, Tara Madhyastha, Thomas Grabowski

Brain-Machine Intelligence and Brain-Inspired Computing

Frontmatter

The Effect of Culture and Social Orientation on Player’s Performances in Tacit Coordination Games

Social Value Orientation (SVO) is one of the main factors affecting strategic decision making. This study explores the effects of different cultural background on players’ SVO as well as on their ability to coordinate in tacit coordination games. Tacit coordination games are coordination games in which communication between the players is not allowed or not possible. Our results showed that the two cultural backgrounds (Israelis and Chinese players) differ in the distribution of the SVO angle (a measure of the social orientation of the player), which is useful for predicting the cultural background of the player. Next, we explored the effects of the SVO value on players’ strategies in tacit coordination games and demonstrated that players with different cultural backgrounds are endowed with different coordination abilities (as measured by a coordination index).

Dor Mizrahi, Ilan Laufer, Inon Zuckerman, Tielin Zhang

Self-programming Robots Boosted by Neural Agents

This paper deals with Brain-Inspired robot controllers, based on a special kind of artificial neural structures that burn “dark” energy to promote the self-motivated initiation of behaviors. We exploit this ambient to train a virtual multi-joint robot, with many moving parts, muscles and sensors distributed through the robot body, interacting with elements that satisfy Newtonian laws. The robot faces a logical-mechanical challenge where a heavy, slippery ball, pressed against a wall has to be pushed up by means of coordinate muscles activation, where energy, timing and balancing conditions add noticeable technical complications. As in living brains our robots contains self-motivating neural agents that consumes energy and function by themselves even without external stimulus. Networks that handle sensory and timing information are combined with agents to construct our controller. We prove that by using appropriate learning algorithms, the self-motivating capacity of agents provides the robot with powerful self-programming aptitudes, capable of solving the ball lifting problem in a quick, efficient way.

Oscar Chang

EEG-Based Subjects Identification Based on Biometrics of Imagined Speech Using EMD

When brain activity ions, the potential for human capacities augmentation is promising. In this paper, EMD is used to decompose EEG signals during Imagined Speech in order to use it as a biometric marker for creating a Biometric Recognition System. For each EEG channel, the most relevant Intrinsic Mode Functions (IMFs) are decided based on the Minkowski distance, and for each IMF 4 features are computed: Instantaneous and Teager energy distribution and Higuchi and Petrosian Fractal Dimension. To test the proposed method, a dataset with 20 Subjects who imagined 30 repetitions of 5 words in Spanish, is used. Four classifiers are used for this task - random forest, SVM, naive Bayes, and k-NN - and their performances are compared. The accuracy obtained (up to 0.92 using Linear SVM) after 10-folds cross-validation suggest that the proposed method based on EMD can be valuable for creating EEG-based biometrics of imagined speech for Subject identification.

Luis Alfredo Moctezuma, Marta Molinas

Simulating Phishing Email Processing with Instance-Based Learning and Cognitive Chunk Activation

We present preliminary steps applying computational cognitive modeling to research decision-making of cybersecurity users. Building from a recent empirical study, we adapt Instance-Based Learning Theory and ACT-R’s description of memory chunk activation in a cognitive model representing the mental process of users processing emails. In this model, a user classifies emails as phishing or legitimate by counting the number of suspicious-seeming cues in each email; these cues are themselves classified by examining similar, past classifications in long-term memory. When the sum of suspicious cues passes a threshold value, that email is classified as phishing. In a simulation, we manipulate three parameters (suspicion threshold; maximum number of cues processed; weight of similarity term) and examine their effects on accuracy, false positive/negative rates, and email processing time.

Matthew Shonman, Xiangyang Li, Haoruo Zhang, Anton Dahbura

Sparse Sampling and Fully-3D Fast Total Variation Based Imaging Reconstruction for Chemical Shift Imaging in Magnetic Resonance Spectroscopy

We propose a 3-dimensional sparse sampling reconstruction method, aiming for chemical shift imaging in magnetic resonance spectroscopy. The method is a Compressed Sensing (CS) method based on the interior point optimization technique that can substantially reduce the number of sampling points required, and the method has been tested successfully in hyperpolarized 13C experimental data using two different sampling strategies.

Zigen Song, Melinda Baxter, Mingwu Jin, Jian-Xiong Wang, Ren-Cang Li, Talon Johnson, Jianzhong Su

An Auto TCD Probe Design and Visualization

Transcranial Doppler ultrasound (TCD) is a non-invasive ultrasound method used to examine blood circulation within the brain. During TCD, ultrasound waves are transmitted through the tissues including skull. These sound waves reflect off blood cells moving within the blood vessels, allowing the radiologist to interpret their speed and direction. In this paper, an auto TCD probe is developed to control the 2D deflection angles of the probe. The techniques of Magnetic Resonance Angiography (MRA) and Magnetic Resource Imagine (MRI) have been used to build the 3D human head model and generate the structure of cerebral arteries. The K-Nearest Neighbors (KNN) algorithm as a non-parametric method has been used for signal classification and regression of corresponding arteries. Finally, a global search and local search algorithms are used to locate the ultrasound focal zone and obtain a stronger signal efficient and more accurate result.

Yi Huang, Peng Wen, Bo Song, Yan Li

Backmatter

Weitere Informationen

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.

Whitepaper

- ANZEIGE -

Globales Erdungssystem in urbanen Kabelnetzen

Bedingt durch die Altersstruktur vieler Kabelverteilnetze mit der damit verbundenen verminderten Isolationsfestigkeit oder durch fortschreitenden Kabelausbau ist es immer häufiger erforderlich, anstelle der Resonanz-Sternpunktserdung alternative Konzepte für die Sternpunktsbehandlung umzusetzen. Die damit verbundenen Fehlerortungskonzepte bzw. die Erhöhung der Restströme im Erdschlussfall führen jedoch aufgrund der hohen Fehlerströme zu neuen Anforderungen an die Erdungs- und Fehlerstromrückleitungs-Systeme. Lesen Sie hier über die Auswirkung von leitfähigen Strukturen auf die Stromaufteilung sowie die Potentialverhältnisse in urbanen Kabelnetzen bei stromstarken Erdschlüssen. Jetzt gratis downloaden!

Bildnachweise