Sparse Bayesian Classification of EEG for Brain-Computer Interface

Multivariate classification of earthquake survivors with post-traumatic stress disorder based on large-scale brain networks

OBJECTIVE

The identification of post-traumatic stress disorder (PTSD) among natural disaster survivors is remarkably challenging, and there are no reliable objective signatures that can be used to assist clinical diagnosis and optimize treatment. The current study aimed to establish a neurobiological signature of PTSD from the connectivity of large-scale brain networks and clarify the brain network mechanisms of PTSD.

METHODS

We examined fifty-seven unmedicated survivors with chronic PTSD and 59 matched trauma-exposed healthy controls (TEHCs) using resting-state functional magnetic resonance imaging (rs-fMRI). We extracted the node-to-network connectivity and obtained a feature vector with a dimensionality of 864 (108 nodes × 8 networks) to represent each subject's functional connectivity (FC) profile. Multivariate pattern analysis with a relevance vector machine was then used to distinguish PTSD patients from TEHCs.

RESULTS

We achieved a promising diagnostic accuracy of 89.2% in distinguishing PTSD patients from TEHCs. The most heavily weighted connections for PTSD classification were among the default mode network (DMN), visual network (VIS), somatomotor network, limbic network, and dorsal attention network (DAN). The strength of the anticorrelation of FC between the ventral medial prefrontal cortex (vMPFC) in DMN and the VIS and DAN was associated with the severity of PTSD.

CONCLUSIONS

This study achieved relatively high accuracy in classifying PTSD patients vs. TEHCs at the individual level. This performance demonstrates that rs-fMRI-derived multivariate classification based on large-scale brain networks can provide potential signatures both to facilitate clinical diagnosis and to clarify the underlying brain network mechanisms of PTSD caused by natural disasters.

Monitoring Depth of Anesthesia Based on Hybrid Features and Recurrent Neural Network

Electroencephalogram (EEG) signals contain valuable information about the different physiological states of the brain, with a variety of linear and nonlinear features that can be used to investigate brain activity. Monitoring the depth of anesthesia (DoA) with EEG is an ongoing challenge in anesthesia research. In this paper, we propose a novel method based on Long Short-Term Memory (LSTM) and a sparse denoising autoencoder (SDAE) to combine the hybrid features of EEG to monitor the DoA. The EEG signals were preprocessed using filtering, etc., and then more than ten features including sample entropy, permutation entropy, spectra, and alpha-ratio were extracted from the EEG signal. We then integrated the optional features such as permutation entropy and alpha-ratio to extract the essential structure and learn the most efficient temporal model for monitoring the DoA. Compared with using a single feature, the proposed model could accurately estimate the depth of anesthesia with higher prediction probability (P_k). Experimental results evaluated on the datasets demonstrated that our proposed method provided better performance than the methods using permutation entropy, alpha-ratio, LSTM, and other traditional indices.

Detection and Prediction of Microsleeps from EEG using Spatio-Temporal Patterns

A microsleep is a brief lapse in performance due to an involuntary sleep-related loss of consciousness. These episodes are of particular importance in occupations requiring extended unimpaired visuomotor performance, such as driving. Detection and even prediction of microsleeps has the potential to prevent catastrophic events and fatal accidents. In this study, we examined detection and prediction of microsleeps using EEG data of 8 subjects who performed two 1-h sessions of continuous 1-D tracking. A regularized spatio-temporal filtering and classification (RSTFC) method was used to extract features from 5-s EEG segments. These features were then used to train three different linear classifiers: linear discriminant analysis (LDA), sparse Bayesian learning (SBL), and variational Bayesian logistic regression (VBLR). The performance of microsleep state detection and prediction was evaluated using leave-one-subject-out cross-validation. The detection performance measures were AUC_ROC 0.96, AUC_PR 0.52, and phi 0.47. As expected, prediction of microsleep states with a 0.25-s ahead prediction time resulted in slightly lower performances compared to the detection. Prediction performance measures were substantially higher than those achieved with log-power spectral features, i.e., AUC_ROC 0.95 (cf. 0.90), AUC_PR 0.50 (cf. 0.36), and phi 0.46 (cf. 0.34).

Decoding Movement-Related Cortical Potentials Based on Subject-Dependent and Section-Wise Spectral Filtering

An important challenge in developing a movement-related cortical potential (MRCP)-based brain-machine interface (BMI) is an accurate decoding of the user intention for real-world environments. However, the performance remains insufficient for real-time decoding owing to the endogenous signal characteristics compared to other BMI paradigms. This study aims to enhance the MRCP decoding performance from the perspective of preprocessing techniques (i.e., spectral filtering). To the best of our knowledge, existing MRCP studies have used spectral filters with a fixed frequency bandwidth for all subjects. Hence, we propose a subject-dependent and section-wise spectral filtering (SSSF) method that considers the subjects' individual MRCP characteristics for two different temporal sections. In this study, MRCP data were acquired under a powered exoskeleton environments in which the subjects conducted self-initiated walking. We evaluated our method using both our experimental data and a public dataset (BNCI Horizon 2020). The decoding performance using the SSSF was 0.86 (±0.09), and the performance on the public dataset was 0.73 (±0.06) across all subjects. The experimental results showed a statistically significant enhancement ( ) compared with the fixed frequency bands used in previous methods on both datasets. In addition, we presented successful decoding results from a pseudo-online analysis. Therefore, we demonstrated that the proposed SSSF method can involve more meaningful MRCP information than conventional methods.

The Spatio-Temporal Equalization for Evoked or Event-Related Potential Detection in Multichannel EEG Data

GOAL

Evoked or Event-Related Potential (EP/ERP) detection is a major area of interest within the domain of EEG (electroencephalography) signal processing. While traditional methods of EEG processing have mostly focused on enhancing signal components, few have explored background noise suppression techniques. Optimizing the suppression of background noise can play a critical role in improving the performance of EP/ERP detection.

METHODS

In this study, a spatio-temporal equalization (STE) method was proposed based on the Multivariate Autoregressive (MVAR) model, which has been shown to suppress the spatio-temporal correlation of background noise and improve the EEG signal detection performance.

RESULTS

For practical applications, two optimization schemes based on the spatio-temporal equalization method were designed to solve two common challenges in EEG signal detection: P300 and steady state visual evoked potentials. Our results demonstrated that the STE method effectively improves recognition performance of evoked or event-related potential detection. Additionally, the STE method offers fewer parameters, lower computational complexity, and easier implementation.

CONCLUSION

These attributes allow the STE approach to be extended as a preprocessing method which can be used in combination with existing techniques.

A strategy combining intrinsic time-scale decomposition and a feedforward neural network for automatic seizure detection

Epilepsy is a common neurological disorder which can occur in people of all ages globally. For the clinical treatment of epileptic patients, the detection of epileptic seizures is of great significance.

OBJECTIVE

Electroencephalography (EEG) is an essential component in the diagnosis of epileptic seizures, from which brain surgeons can detect important pathological information about patient epileptiform discharges. This paper focuses on adaptive seizure detection from EEG recordings. We propose a new feature extraction model based on an adaptive decomposition method, named intrinsic time-scale decomposition (ITD), which is suitable for analyzing non-linear and non-stationary data.

APPROACH

Firstly, using the ITD technique, every EEG recording is decomposed into several proper rotation components (PRCs). Secondly, the instantaneous amplitudes and frequencies of these PRCs can be calculated and then we extract their statistical indices. Furthermore, we combine all these statistical indices of the corresponding five PRCs as the feature vector of each EEG signal. Finally, these feature vectors are fed into a feedforward neural network (FNN) classifier for EEG classification. The whole process of feature extraction proposed in this paper only involves one parameter and the role of the ITD method is based on a piecewise linear function, which makes the computation of the model simple and fast. More useful information for classification can be obtained since we take advantage of both instantaneous amplitude and instantaneous frequency for feature extraction.

MAIN RESULTS

We consider the 17 classification problems which contain normal versus epileptic, non-seizure versus seizure and normal versus interictal versus ictal using a FNN classifier which only contains one hidden layer. Experimental results show that the proposed method can catch the discriminative features of EEG signals and obtain comparable results when compared with state-of-the-art detection methods.

SIGNIFICANCE

Therefore, the proposed system has a great potential in real-time seizure detection and provides physicians with a real-time diagnostic aid in their practice.

Brain regulation of emotional conflict predicts antidepressant treatment response for depression

The efficacy of antidepressant treatment for depression is controversial due to the only modest superiority demonstrated over placebo. However, neurobiological heterogeneity within depression may limit overall antidepressant efficacy. We sought to identify a neurobiological phenotype responsive to antidepressant treatment by testing pretreatment brain activation during response to, and regulation of, emotional conflict as a moderator of the clinical benefit of the antidepressant sertraline versus placebo. Using neuroimaging data from a large randomized controlled trial, we found widespread moderation of clinical benefits by brain activity during regulation of emotional conflict, in which greater downregulation of conflict-responsive regions predicted better sertraline outcomes. Treatment-predictive machine learning using brain metrics outperformed a model trained on clinical and demographic variables. Our findings demonstrate that antidepressant response is predicted by brain activity underlying a key self-regulatory emotional capacity. Leveraging brain-based measures in psychiatry will forge a path toward better treatment personalization, refined mechanistic insights and improved outcomes.

Theories and Methods for Labeling Cognitive Workload: Classification and Transfer Learning

There are a number of key data-centric questions that must be answered when developing classifiers for operator functional states. "Should a supervised or unsupervised learning approach be used? What degree of labeling and transformation must be performed on the data? What are the trade-offs between algorithm flexibility and model interpretability, as generally these features are at odds?" Here, we focus exclusively on the labeling of cognitive load data for supervised learning. We explored three methods of labeling cognitive states for three-state classification. The first method labels states derived from a tertiary split of trial difficulty during a spatial memory task. The second method was more adaptive; it employed a mixed-effects stress-strain curve and estimated an individual's performance asymptotes with respect to the same spatial memory task. The final method was similar to the second approach; however, it employed a mixed-effects Rasch model to estimate individual capacity limits within the context of item response theory for the spatial memory task. To assess the strength of each of these labeling approaches, we compared the area under the curve (AUC) for receiver operating curves (ROCs) from elastic net and random forest classifiers. We chose these classifiers based on a combination of interpretability, flexibility, and past modeling success. We tested these techniques across two groups of individuals and two tasks to test the effects of different labeling techniques on cross-person and cross-task transfer. Overall, we observed that the Rasch model labeling paired with a random forest classifier led to the best model fits and showed evidence of both cross-person and cross-task transfer.

Temporally Constrained Sparse Group Spatial Patterns for Motor Imagery BCI

Common spatial pattern (CSP)-based spatial filtering has been most popularly applied to electroencephalogram (EEG) feature extraction for motor imagery (MI) classification in brain-computer interface (BCI) application. The effectiveness of CSP is highly affected by the frequency band and time window of EEG segments. Although numerous algorithms have been designed to optimize the spectral bands of CSP, most of them selected the time window in a heuristic way. This is likely to result in a suboptimal feature extraction since the time period when the brain responses to the mental tasks occurs may not be accurately detected. In this paper, we propose a novel algorithm, namely temporally constrained sparse group spatial pattern (TSGSP), for the simultaneous optimization of filter bands and time window within CSP to further boost classification accuracy of MI EEG. Specifically, spectrum-specific signals are first derived by bandpass filtering from raw EEG data at a set of overlapping filter bands. Each of the spectrum-specific signals is further segmented into multiple subseries using sliding window approach. We then devise a joint sparse optimization of filter bands and time windows with temporal smoothness constraint to extract robust CSP features under a multitask learning framework. A linear support vector machine classifier is trained on the optimized EEG features to accurately identify the MI tasks. An experimental study is implemented on three public EEG datasets (BCI Competition III dataset IIIa, BCI Competition IV datasets IIa, and BCI Competition IV dataset IIb) to validate the effectiveness of TSGSP in comparison to several other competing methods. Superior classification performance (averaged accuracies are 88.5%, 83.3%, and 84.3% for the three datasets, respectively) based on the experimental results confirms that the proposed algorithm is a promising candidate for performance improvement of MI-based BCIs.

A semi-blind online dictionary learning approach for fMRI data

BACKGROUND

Online dictionary learning (ODL) has been applied to extract brain networks from functional magnetic resonance imaging (fMRI) data in recent year. Moreover, the supervised dictionary learning (SDL) that fixed the task stimulus curves as predefined atoms was proposed to improve ODL for functional networks separation. However, SDL cannot estimate the real time courses underlying the brain networks and cannot be applied to the inter-network connectivity analysis. This study aimed at investigating how to add the temporal prior information to ODL to extract the accurate task-related brain networks and the corresponding time courses.

NEW METHOD

To improve the performance of ODL, we propose a semi-blind ODL (semi-ODL) method that incorporates temporal prior information of the task paradigm into the dictionary updating process and optimizes the direction of one or more specific atoms "close" to the task time courses.

RESULTS

Results of the simulated and real fMRI experiment revealed that semi-ODL extracted more accurate task-related component and time courses than ODL and SDL. For one-task fMRI data, semi-ODL and Infomax-ICA showed similar detection power in most cases.

COMPARISON WITH EXISTING METHODS

The semi-ODL outperformed ODL, SDL in robustness to noise, spatial detection power and time course estimation. Moreover, semi-ODL showed comparable performance to Infomax-ICA for one-task fMRI data and outperformed Infomax-ICA in extracting the components related to each task from multi-task fMRI data.

CONCLUSIONS

The semi-ODL method is potentially useful to reveal brain networks underlying various cognitive tasks and the interactions between task-related brain networks.

Brain wave classification using long short-term memory network based OPTICAL predictor

Brain-computer interface (BCI) systems having the ability to classify brain waves with greater accuracy are highly desirable. To this end, a number of techniques have been proposed aiming to be able to classify brain waves with high accuracy. However, the ability to classify brain waves and its implementation in real-time is still limited. In this study, we introduce a novel scheme for classifying motor imagery (MI) tasks using electroencephalography (EEG) signal that can be implemented in real-time having high classification accuracy between different MI tasks. We propose a new predictor, OPTICAL, that uses a combination of common spatial pattern (CSP) and long short-term memory (LSTM) network for obtaining improved MI EEG signal classification. A sliding window approach is proposed to obtain the time-series input from the spatially filtered data, which becomes input to the LSTM network. Moreover, instead of using LSTM directly for classification, we use regression based output of the LSTM network as one of the features for classification. On the other hand, linear discriminant analysis (LDA) is used to reduce the dimensionality of the CSP variance based features. The features in the reduced dimensional plane after performing LDA are used as input to the support vector machine (SVM) classifier together with the regression based feature obtained from the LSTM network. The regression based feature further boosts the performance of the proposed OPTICAL predictor. OPTICAL showed significant improvement in the ability to accurately classify left and right-hand MI tasks on two publically available datasets. The improvements in the average misclassification rates are 3.09% and 2.07% for BCI Competition IV Dataset I and GigaDB dataset, respectively. The Matlab code is available at https://github.com/ShiuKumar/OPTICAL .

STRAINet: Spatially Varying sTochastic Residual AdversarIal Networks for MRI Pelvic Organ Segmentation

Accurate segmentation of pelvic organs is important for prostate radiation therapy. Modern radiation therapy starts to use a magnetic resonance image (MRI) as an alternative to computed tomography image because of its superior soft tissue contrast and also free of risk from radiation exposure. However, segmentation of pelvic organs from MRI is a challenging problem due to inconsistent organ appearance across patients and also large intrapatient anatomical variations across treatment days. To address such challenges, we propose a novel deep network architecture, called "Spatially varying sTochastic Residual AdversarIal Network" (STRAINet), to delineate pelvic organs from MRI in an end-to-end fashion. Compared to the traditional fully convolutional networks (FCN), the proposed architecture has two main contributions: 1) inspired by the recent success of residual learning, we propose an evolutionary version of the residual unit, i.e., stochastic residual unit, and use it to the plain convolutional layers in the FCN. We further propose long-range stochastic residual connections to pass features from shallow layers to deep layers; and 2) we propose to integrate three previously proposed network strategies to form a new network for better medical image segmentation: a) we apply dilated convolution in the smallest resolution feature maps, so that we can gain a larger receptive field without overly losing spatial information; b) we propose a spatially varying convolutional layer that adapts convolutional filters to different regions of interest; and c) an adversarial network is proposed to further correct the segmented organ structures. Finally, STRAINet is used to iteratively refine the segmentation probability maps in an autocontext manner. Experimental results show that our STRAINet achieved the state-of-the-art segmentation accuracy. Further analysis also indicates that our proposed network components contribute most to the performance.

Wireless Computer-Supported Cooperative Work: A Pilot Experiment on Art and Brain⁻Computer Interfaces

The present case study looked into the feasibility of using brain–computer interface (BCI) technology combined with computer-supported cooperative work (CSCW) in a wireless network. We had two objectives; first, to test the wireless BCI-based configuration and the practical use of this idea we assessed workload perception in participants located several kilometers apart taking part in the same drawing task. Second, we studied the cortical activation patterns of participants performing the drawing task with and without the BCI technology. Results showed higher mental workload perception and broader cortical activation (frontal-temporal-occipital) under BCI experimental conditions. This idea shows a possible application of BCI research in the social field, where two or more users could engage in a computer networking task using BCI technology over the internet. New research avenues for CSCW are discussed and possibilities for future research are given.

Strength and Similarity Guided Group-level Brain Functional Network Construction for MCI Diagnosis

Sparse representation-based brain functional network modeling often results in large inter-subject variability in the network structure. This could reduce the statistical power in group comparison, or even deteriorate the generalization capability of the individualized diagnosis of brain diseases. Although group sparse representation (GSR) can alleviate such a limitation by increasing network similarity across subjects, it could, in turn, fail in providing satisfactory separability between the subjects from different groups (e.g., patients vs. controls). In this study, we propose to integrate individual functional connectivity (FC) information into the GSR-based network construction framework to achieve higher between-group separability while maintaining the merit of within-group consistency. Our method was based on an observation that the subjects from the same group have generally more similar FC patterns than those from different groups. To this end, we propose our new method, namely "strength and similarity guided GSR (SSGSR)", which exploits both BOLD signal temporal correlation-based "low-order" FC (LOFC) and inter-subject LOFC-profile similarity-based "high-order" FC (HOFC) as two priors to jointly guide the GSR-based network modeling. Extensive experimental comparisons are carried out, with the rs-fMRI data from mild cognitive impairment (MCI) subjects and healthy controls, between the proposed algorithm and other state-of-the-art brain network modeling approaches. Individualized MCI identification results show that our method could achieve a balance between the individually consistent brain functional network construction and the adequately maintained inter-group brain functional network distinctions, thus leading to a more accurate classification result. Our method also provides a promising and generalized solution for the future connectome-based individualized diagnosis of brain disease.

Single-Trial Recognition of Imagined Forces and Speeds of Hand Clenching Based on Brain Topography and Brain Network

To provide optional force and speed control parameters for brain-computer interfaces (BCIs), an effective feature extraction method of imagined force and speed of hand clenching based on electroencephalography (EEG) was explored. Twenty subjects were recruited to participate in the experiment. They were instructed to perform three different actual/imagined hand clenching force tasks (4 kg, 10 kg, and 16 kg) and three different hand clenching speed tasks (0.5 Hz, 1 Hz, and 2 Hz). Topographical maps parameters and brain network parameters of EEG were calculated as new features of imagined force and speed of hand clenching, which were classified by three classifiers: linear discrimination analysis, extreme learning machines and support vector machines. Topographical maps parameters were better for recognition of the hand clenching force task, while brain network parameters were better for recognition of the hand clenching speed task. After a combination of five types of features (energy, power spectrum of the autoregressive model, wavelet packet coefficients, topographical maps parameters and brain network parameters), the recognition rate of the hand clenching force task was 97%, and that of the hand clenching speed task was as high as 100%. The brain topographical and the brain network parameters are expected to improve the accuracy of decoding the EEG signal of imagined force and speed of hand clenching. A more efficient brain network may facilitate the recognition of force/speed of hand clenching. Combined features could significantly improve the single-trial recognition rate of imagined forces and speeds of hand clenching. The current study provides a new idea for the imagined force and speed of hand clenching that offers more control intention instructions for BCIs.

Feature selection using regularized neighbourhood component analysis to enhance the classification performance of motor imagery signals

In motor imagery (MI) based brain-computer interface (BCI) signal analysis, mu and beta rhythms of electroencephalograms (EEGs) are widely investigated due to their high temporal resolution and capability to define the different movement-related mental tasks separately. However, due to the high dimensions and subject-specific behaviour of EEG features, there is a need for a suitable feature selection algorithm that can select the optimal features to give the best classification performance along with increased computational efficiency. The present study proposes a feature selection algorithm based on neighbourhood component analysis (NCA) with modification of the regularization parameter. In the experiment, time, frequency, and phase features of the EEG are extracted using a dual-tree complex wavelet transform (DTCWT). Afterwards, the proposed algorithm selects the most significant EEG features, and using these selected features, a support vector machine (SVM) classifier performs the classification of MI signals. The proposed algorithm has been validated experimentally on two public BCI datasets (BCI Competition II Dataset III and BCI Competition IV Dataset 2b). The classification performance of the algorithm is quantified by the average accuracy and kappa coefficient, whose values are 80.7% and 0.615 respectively. The performance of the proposed algorithm is compared with standard feature selection methods based on Genetic Algorithm (GA), Principal Component Analysis (PCA), and ReliefF and performs better than these methods. Further, the proposed algorithm selects the lowest number of features and results in increased computational efficiency, which makes it a promising feature selection tool for an MI-based BCI system.

Investigating EEG Patterns for Dual-Stimuli Induced Human Fear Emotional State

Most electroencephalography (EEG) based emotion recognition systems make use of videos and images as stimuli. Few used sounds, and even fewer studies were found involving self-induced emotions. Furthermore, most of the studies rely on single stimuli to evoke emotions. The question of "whether different stimuli for same emotion elicitation generate any subject-independent correlations" remains unanswered. This paper introduces a dual modality based emotion elicitation paradigm to investigate if emotions can be classified induced with different stimuli. A method has been proposed based on common spatial pattern (CSP) and linear discriminant analysis (LDA) to analyze human brain signals for fear emotions evoked with two different stimuli. Self-induced emotional imagery is one of the considered stimuli, while audio/video clips are used as the other stimuli. The method extracts features from the CSP algorithm and LDA performs classification. To investigate associated EEG correlations, a spectral analysis was performed. To further improve the performance, CSP was compared with other regularized techniques. Critical EEG channels are identified based on spatial filter weights. To the best of our knowledge, our work provides the first contribution for the assessment of EEG correlations in the case of self versus video induced emotions captured with a commercial grade EEG device.

Electrophysiology Meets Printed Electronics: The Beginning of a Beautiful Friendship

Electroencephalography (EEG) and surface electromyography (sEMG) are notoriously cumbersome technologies. A typical setup may involve bulky electrodes, dangling wires, and a large amplifier unit. Adapting these technologies to numerous applications has been accordingly fairly limited. Thanks to the availability of printed electronics, it is now possible to effectively simplify these techniques. Elegant electrode arrays with unprecedented performances can be readily produced, eliminating the need to handle multiple electrodes and wires. Specifically, in this Perspective paper, we focus on the advantages of electrodes printed on soft films as manifested in signal transmission at the electrode-skin interface, electrode-skin stability, and user convenience during electrode placement while achieving prolonged use. Customizing electrode array designs and implementing blind source separation methods can also improve recording resolution, reduce variability between individuals and minimize signal cross-talk between nearby electrodes. Finally, we outline several important applications in the field of neuroscience and how each can benefit from the convergence of electrophysiology and printed electronics.

A Logarithmic Quantization-Based Image Watermarking Using Information Entropy in the Wavelet Domain

Conventional quantization-based watermarking may be easily estimated by averaging on a set of watermarked signals via uniform quantization approach. Moreover, the conventional quantization-based method neglects the visual perceptual characteristics of the host signal; thus, the perceptible distortions would be introduced in some parts of host signal. In this paper, inspired by the Watson's entropy masking model and logarithmic quantization index modulation (LQIM), a logarithmic quantization-based image watermarking method is developed by using the wavelet transform. Furthermore, the novel method improves the robustness of watermarking based on a logarithmic quantization strategy, which embeds the watermark data into the image blocks with high entropy value. The main significance of this work is that the trade-off between invisibility and robustness is simply addressed by using the logarithmic quantizaiton approach, which applies the entropy masking model and distortion-compensated scheme to develop a watermark embedding method. In this manner, the optimal quantization parameter obtained by minimizing the quantization distortion function effectively controls the watermark strength. In terms of watermark decoding, we model the wavelet coefficients of image by the generalized Gaussian distribution (GGD) and calculate the bit error probability of proposed method. Performance of the proposed method is analyzed and verified by simulation on real images. Experimental results demonstrate that the proposed method has the advantages of imperceptibility and strong robustness against attacks covering JPEG compression, additive white Gaussian noise (AWGN), Gaussian filtering, Salt&Peppers noise, scaling and rotation attack, etc.

Medical Image Synthesis with Deep Convolutional Adversarial Networks

Medical imaging plays a critical role in various clinical applications. However, due to multiple considerations such as cost and radiation dose, the acquisition of certain image modalities may be limited. Thus, medical image synthesis can be of great benefit by estimating a desired imaging modality without incurring an actual scan. In this paper, we propose a generative adversarial approach to address this challenging problem. Specifically, we train a fully convolutional network (FCN) to generate a target image given a source image. To better model a nonlinear mapping from source to target and to produce more realistic target images, we propose to use the adversarial learning strategy to better model the FCN. Moreover, the FCN is designed to incorporate an image-gradient-difference-based loss function to avoid generating blurry target images. Long-term residual unit is also explored to help the training of the network. We further apply Auto-Context Model to implement a context-aware deep convolutional adversarial network. Experimental results show that our method is accurate and robust for synthesizing target images from the corresponding source images. In particular, we evaluate our method on three datasets, to address the tasks of generating CT from MRI and generating 7T MRI from 3T MRI images. Our method outperforms the state-of-the-art methods under comparison in all datasets and tasks.

An adaptive CSP filter to investigate user independence in a 3-class MI-BCI paradigm

This paper describes the implementation of a Brain Computer Interface (BCI) scheme using a common spatial patterns (CSP) filter in combination with a Recursive Least Squares (RLS) approach to iteratively update the coefficients of the CSP filter. The proposed adaptive CSP (ACSP) algorithm is made more robust by introducing regularization using Diagonal Loading (DL), and thus will be able to significantly reduce the length of training sessions when introducing new patients to the BCI system. The system is tested on a 4-class multi-limb motor imagery (MI) data set from the BCI competition IV (2a), and a more complex single limb 3-class MI dataset recorded in-house. The latter dataset is produced to mimic an upper limb rehabilitation session, e.g., after stroke. The findings indicate that when extensive calibration data is available, the ACSP performs comparably to the CSP (kappa value of 0.523 and 0.502, respectively, for the 4-class problem); for reduced calibration sessions, the ACSP significantly improved the performance of the system (up to 4-fold). The proposed paradigm proved feasible and the ACSP algorithm seems to enable a user or semi user independent scenario, where the need for long system calibration sessions without feedback is eliminated.

Investigation of Optimal Afferent Feedback Modality for Inducing Neural Plasticity with A Self-Paced Brain-Computer Interface

Brain-computer interfaces (BCIs) can be used to induce neural plasticity in the human nervous system by pairing motor cortical activity with relevant afferent feedback, which can be used in neurorehabilitation. The aim of this study was to identify the optimal type or combination of afferent feedback modalities to increase cortical excitability in a BCI training intervention. In three experimental sessions, 12 healthy participants imagined a dorsiflexion that was decoded by a BCI which activated relevant afferent feedback: (1) electrical nerve stimulation (ES) (peroneal nerve-innervating tibialis anterior), (2) passive movement (PM) of the ankle joint, or (3) combined electrical stimulation and passive movement (Comb). The cortical excitability was assessed with transcranial magnetic stimulation determining motor evoked potentials (MEPs) in tibialis anterior before, immediately after and 30 min after the BCI training. Linear mixed regression models were used to assess the changes in MEPs. The three interventions led to a significant (p < 0.05) increase in MEP amplitudes immediately and 30 min after the training. The effect sizes of Comb paradigm were larger than ES and PM, although, these differences were not statistically significant (p > 0.05). These results indicate that the timing of movement imagery and afferent feedback is the main determinant of induced cortical plasticity whereas the specific type of feedback has a moderate impact. These findings can be important for the translation of such a BCI protocol to the clinical practice where by combining the BCI with the already available equipment cortical plasticity can be effectively induced. The findings in the current study need to be validated in stroke populations.

Modeling Mindsets with Kalman Filter

Mathematical models have played an essential role in interface design. This study focused on “mindsets”—people’s tacit beliefs about attributes—and investigated the extent to which: (1) mindsets can be extracted in a motion trajectory in target selection, and (2) a dynamic state-space model, such as the Kalman filter, helps quantify mindsets. Participants were experimentally manipulated to hold fixed or growth mindsets in a “mock” memory test, and later performed a concept-learning task in which the movement of the computer cursor was recorded in every trial. By inspecting motion trajectories of the cursor, we observed clear disparities in the impact of mindsets; participants who were induced with a fixed mindset moved the cursor faster as compared to those who were induced with a growth mindset. To examine further the mechanism of this influence, we fitted a Kalman filter model to the trajectory data; we found that system-level error-covariance in the Kalman filter model could effectively separate motion trajectories gleaned from the two mindset conditions. Taken together, results from the experiment suggest that people’s mindsets can be captured in motor trajectories in target selection and the Kalman filter helps quantify mindsets. It is argued that people’s personality, attitude, and mindset are embodied in motor behavior underlying target selection and these psychological variables can be studied mathematically with a feedback control system.

Abnormal cortical region and subsystem complexity in dynamical functional connectivity of chronic schizophrenia: A new graph index for fMRI analysis

OBJECTIVES

Schizophrenia is a predominant product of pathological alterations distributed throughout interconnected neural systems. Designing new objectively diagnostic methods are burning questions. Dynamical functional connectivity (DFCs) methodology based on fMRI data is an effective lever to investigate changeability evolution in macroscopic neural activity patterns underlying critical aspects of cognition and behavior. However, region properties of brain architecture have been less investigated by special indexes of dynamical graph in general mental disorders.

METHODS

Embracing the network dynamics concept, we introduce topology entropy index (TE-scores) which is focused on time-varying aspects of FCs, hence develop a new framework for researching the dysfunctional roots of schizophrenia in holism significance. In this work, the important process is to uncover noticeable regions endowed with antagonistic stance in TE-scores of between morbid and normal DFCs of 63 healthy controls (HCs) and 57 chronic schizophrenia patients (SZs).

RESULTS

For the whole brain region levels, right olfactory, right hippocampus, left parahippocampal gyrus, right parahippocampal gyrus, left amygdala, and left cuneus in SZs are endowed with significantly different TE-scores. At brain subsystems level, TE-scores in DMN are abnormal in the SZs. Comparison with existing method(s): Topology entropy in DFCs is introduced to explore the dynamical information organization of diverse regions and their abnormal changes in mental illness. Several classical graph indexes (such as degree strength, betweenness, centrality) in the static brain network measure the region importance of FCs under senses of information integration and separation process. Although highly related to degree strength by comparing the corresponding values, topology entropy further explores the regions' aberrant adaptability of functional contact and function switching.

CONCLUSION

TE-scores of abnormal regions in SZs are associated to the passive apathetic social withdrawal, unusual thought content, disturbance of volition, preoccupation, active social avoidance and hallucinatory symptoms. Thought the strict contrastive study, it is worth stressing that topology entropy is a meaningful biological marker to excavating schizophrenic psychopathology.

Smart Doll: Emotion Recognition Using Embedded Deep Learning

Computer vision and deep learning are clearly demonstrating a capability to create engaging cognitive applications and services. However, these applications have been mostly confined to powerful Graphic Processing Units (GPUs) or the cloud due to their demanding computational requirements. Cloud processing has obvious bandwidth, energy consumption and privacy issues. The Eyes of Things (EoT) is a powerful and versatile embedded computer vision platform which allows the user to develop artificial vision and deep learning applications that analyse images locally. In this article, we use the deep learning capabilities of an EoT device for a real-life facial informatics application: a doll capable of recognizing emotions, using deep learning techniques, and acting accordingly. The main impact and significance of the presented application is in showing that a toy can now do advanced processing locally, without the need of further computation in the cloud, thus reducing latency and removing most of the ethical issues involved. Finally, the performance of the convolutional neural network developed for that purpose is studied and a pilot was conducted on a panel of 12 children aged between four and ten years old to test the doll.

Kernel Sparse Representation with Hybrid Regularization for On-Road Traffic Sensor Data Imputation

The problem of missing values (MVs) in traffic sensor data analysis is universal in current intelligent transportation systems because of various reasons, such as sensor malfunction, transmission failure, etc. Accurate imputation of MVs is the foundation of subsequent data analysis tasks since most analysis algorithms need complete data as input. In this work, a novel MVs imputation approach termed as kernel sparse representation with elastic net regularization (KSR-EN) is developed for reconstructing MVs to facilitate analysis with traffic sensor data. The idea is to represent each sample as a linear combination of other samples due to inherent spatiotemporal correlation, as well as periodicity of daily traffic flow. To discover few yet correlated samples and make full use of the valuable information, a combination of l₁-norm and l₂-norm is employed to penalize the combination coefficients. Moreover, the linear representation among samples is extended to nonlinear representation by mapping input data space into high-dimensional feature space, which further enhances the recovery performance of our proposed approach. An efficient iterative algorithm is developed for solving KSR-EN model. The proposed method is verified on both an artificially simulated dataset and a public road network traffic sensor data. The results demonstrate the effectiveness of the proposed approach in terms of MVs imputation.

Data-Driven Transducer Design and Identification for Internally-Paced Motor Brain Computer Interfaces: A Review

Brain-Computer Interfaces (BCIs) are systems that establish a direct communication pathway between the users' brain activity and external effectors. They offer the potential to improve the quality of life of motor-impaired patients. Motor BCIs aim to permit severely motor-impaired users to regain limb mobility by controlling orthoses or prostheses. In particular, motor BCI systems benefit patients if the decoded actions reflect the users' intentions with an accuracy that enables them to efficiently interact with their environment. One of the main challenges of BCI systems is to adapt the BCI's signal translation blocks to the user to reach a high decoding accuracy. This paper will review the literature of data-driven and user-specific transducer design and identification approaches and it focuses on internally-paced motor BCIs. In particular, continuous kinematic biomimetic and mental-task decoders are reviewed. Furthermore, static and dynamic decoding approaches, linear and non-linear decoding, offline and real-time identification algorithms are considered. The current progress and challenges related to the design of clinical-compatible motor BCI transducers are additionally discussed.

Jointly Learning Structured Analysis Discriminative Dictionary and Analysis Multiclass Classifier

In this paper, we propose an analysis mechanism-based structured analysis discriminative dictionary learning analysis discriminative dictionary learning, framework. The ADDL seamlessly integrates ADDL, analysis representation, and analysis classifier training into a unified model. The applied analysis mechanism can make sure that the learned dictionaries, representations, and linear classifiers over different classes are independent and discriminating as much as possible. The dictionary is obtained by minimizing a reconstruction error and an analytical incoherence promoting term that encourages the subdictionaries associated with different classes to be independent. To obtain the representation coefficients, ADDL imposes a sparse -norm constraint on the coding coefficients instead of using or norm, since the - or -norm constraint applied in most existing DL criteria makes the training phase time consuming. The code-extraction projection that bridges data with the sparse codes by extracting special features from the given samples is calculated via minimizing a sparse code approximation term. Then we compute a linear classifier based on the approximated sparse codes by an analysis mechanism to simultaneously consider the classification and representation powers. Thus, the classification approach of our model is very efficient, because it can avoid the extra time-consuming sparse reconstruction process with trained dictionary for each new test data as most existing DL algorithms. Simulations on real image databases demonstrate that our ADDL model can obtain superior performance over other state of the arts.

Data-Adaptive Spatiotemporal ERP Cleaning for Single-Trial BCI Implementation

This paper presents a data-adaptive approach to enhance the discriminative information of event-related potential (ERP) for the implementation of a brain-computer interface (BCI). The use of single-trial ERP in a real-time BCI application is challenging, due to its inherent noise contamination. Usually, multiple-trial ERPs are averaged to derive discriminative features of different classes by reducing their noise effects. Time-domain filtering is implemented here using an array wavelet transform. Sometimes, several channels can carry the signals, which are irrelevant to actual EPR information against the respective stimuli. A spatial filtering method based on clustering is introduced, to suppress such channels if any. Hence, the single-trial ERP is filtered in both the spatial and temporal domains to improve its discriminative features. The spatial-temporal discriminate analysis is employed to derive the features leading to the performance of target and non-target classification by using linear discriminant analysis. The proposed method is validated using a data set recorded from our experiments. The experimental results show that the performance of the proposed method is superior to that of the recently developed algorithms for single-trial ERP classification.

A 20-Questions-Based Binary Spelling Interface for Communication Systems

Brain computer interfaces (BCIs) enables people with motor impairments to communicate using their brain signals by selecting letters and words from a screen. However, these spellers do not work for people in a complete locked-in state (CLIS). For these patients, a near infrared spectroscopy-based BCI has been developed, allowing them to reply to “yes”/”no” questions with a classification accuracy of 70%. Because of the non-optimal accuracy, a usual character-based speller for selecting letters or words cannot be used. In this paper, a novel spelling interface based on the popular 20-questions-game has been presented, which will allow patients to communicate using only “yes”/”no” answers, even in the presence of poor classification accuracy. The communication system is based on an artificial neural network (ANN) that estimates a statement thought by the patient asking less than 20 questions. The ANN has been tested in a web-based version with healthy participants and in offline simulations. Both results indicate that the proposed system can estimate a patient’s imagined sentence with an accuracy that varies from 40%, in the case of a “yes”/”no” classification accuracy of 70%, and up to 100% in the best case. These results show that the proposed spelling interface could allow patients in CLIS to express their own thoughts, instead of only answer to “yes”/”no” questions.

Small lung nodules detection based on local variance analysis and probabilistic neural network

BACKGROUND AND OBJECTIVE

In medical examinations doctors use various techniques in order to provide to the patients an accurate analysis of their actual state of health. One of the commonly used methodologies is the x-ray screening. This examination very often help to diagnose some diseases of chest organs. The most frequent cause of wrong diagnosis lie in the radiologist's difficulty in interpreting the presence of lungs carcinoma in chest X-ray. In such circumstances, an automated approach could be highly advantageous as it provides important help in medical diagnosis.

METHODS

In this paper we propose a new classification method of the lung carcinomas. This method start with the localization and extraction of the lung nodules by computing, for each pixel of the original image, the local variance obtaining an output image (variance image) with the same size of the original image. In the variance image we find the local maxima and then by using the locations of these maxima in the original image we found the contours of the possible nodules in lung tissues. However after this segmentation stage we find many false nodules. Therefore to discriminate the true ones we use a probabilistic neural network as classifier.

RESULTS

The performance of our approach is 92% of correct classifications, while the sensitivity is 95% and the specificity is 89.7%. The misclassification errors are due to the fact that network confuses false nodules with the true ones (6%) and true nodules with the false ones (2%).

CONCLUSIONS

Several researchers have proposed automated algorithms to detect and classify pulmonary nodules but these methods fail to detect low-contrast nodules and have a high computational complexity, in contrast our method is relatively simple but at the same time provides good results and can detect low-contrast nodules. Furthermore, in this paper is presented a new algorithm for training the PNN neural networks that allows to obtain PNNs with many fewer neurons compared to the neural networks obtained by using the training algorithms present in the literature. So considerably lowering the computational burden of the trained network and at same time keeping the same performances.

Introducing chaos behavior to kernel relevance vector machine (RVM) for four-class EEG classification

This paper addresses a chaos kernel function for the relevance vector machine (RVM) in EEG signal classification, which is an important component of Brain-Computer Interface (BCI). The novel kernel function has evolved from a chaotic system, which is inspired by the fact that human brain signals depict some chaotic characteristics and behaviors. By introducing the chaotic dynamics to the kernel function, the RVM will be enabled for higher classification capacity. The proposed method is validated within the framework of one versus one common spatial pattern (OVO-CSP) classifier to classify motor imagination (MI) of four movements in a public accessible dataset. To illustrate the performance of the proposed kernel function, Gaussian and Polynomial kernel functions are considered for comparison. Experimental results show that the proposed kernel function achieved higher accuracy than Gaussian and Polynomial kernel functions, which shows that the chaotic behavior consideration is helpful in the EEG signal classification.

Knowledge-Based Neuroendocrine Immunomodulation (NIM) Molecular Network Construction and Its Application

Growing evidence shows that the neuroendocrine immunomodulation (NIM) network plays an important role in maintaining and modulating body function and the homeostasis of the internal environment. The disequilibrium of NIM in the body is closely associated with many diseases. In the present study, we first collected a core dataset of NIM signaling molecules based on our knowledge and obtained 611 NIM signaling molecules. Then, we built a NIM molecular network based on the MetaCore database and analyzed the signaling transduction characteristics of the core network. We found that the endocrine system played a pivotal role in the bridge between the nervous and immune systems and the signaling transduction between the three systems was not homogeneous. Finally, employing the forest algorithm, we identified the molecular hub playing an important role in the pathogenesis of rheumatoid arthritis (RA) and Alzheimer’s disease (AD), based on the NIM molecular network constructed by us. The results showed that GSK3B, SMARCA4, PSMD7, HNF4A, PGR, RXRA, and ESRRA might be the key molecules for RA, while RARA, STAT3, STAT1, and PSMD14 might be the key molecules for AD. The molecular hub may be a potentially druggable target for these two complex diseases based on the literature. This study suggests that the NIM molecular network in this paper combined with the forest algorithm might provide a useful tool for predicting drug targets and understanding the pathogenesis of diseases. Therefore, the NIM molecular network and the corresponding online tool will not only enhance research on complex diseases and system biology, but also promote the communication of valuable clinical experience between modern medicine and Traditional Chinese Medicine (TCM).

Sparse Group Representation Model for Motor Imagery EEG Classification

A potential limitation of a motor imagery (MI) based brain-computer interface (BCI) is that it usually requires a relatively long time to record sufficient electroencephalogram (EEG) data for robust classifier training. The calibration burden during data acquisition phase will most probably cause a subject to be reluctant to use a BCI system. To alleviate this issue, we propose a novel sparse group representation model (SGRM) for improving the efficiency of MI-based BCI by exploiting the intersubject information. Specifically, preceded by feature extraction using common spatial pattern, a composite dictionary matrix is constructed with training samples from both the target subject and other subjects. By explicitly exploiting within-group sparse and group-wise sparse constraints, the most compact representation of a test sample of the target subject is then estimated as a linear combination of columns in the dictionary matrix. Classification is implemented by calculating the class-specific representation residual based on the significant training samples corresponding to the nonzero representation coefficients. Accordingly, the proposed SGRM method effectively reduces the required training samples from the target subject due to auxiliary data available from other subjects. With two public EEG data sets, extensive experimental comparisons are carried out between SGRM and other state-of-the-art approaches. Superior classification performance of our method using 40 trials of the target subject for model calibration (Averaged accuracy = 78.2%, Kappa = 0.57 and Averaged accuracy = 77.7%, Kappa = 0.55 for the two data sets, respectively) indicates its promising potential for improving the practicality of MI-based BCI.