Automated Depression Detection Using Deep Representation and Sequence Learning with EEG Signals.

Depression affects large number of people across the world today and it is considered as the global problem. It is a mood disorder which can be detected using electroencephalogram (EEG) signals. The manual detection of depression by analyzing the EEG signals requires lot of experience, tedious and time consuming. Hence, a fully automated depression diagnosis system developed using EEG signals will help the clinicians. Therefore, we propose a deep hybrid model developed using convolutional neural network (CNN) and long-short term memory (LSTM) architectures to detect depression using EEG signals. In the deep model, temporal properties of the signals are learned with CNN layers and the sequence learning process is provided through the LSTM layers. In this work, we have used EEG signals obtained from left and right hemispheres of the brain. Our work has provided 99.12% and 97.66% classification accuracies for the right and left hemisphere EEG signals respectively. Hence, we can conclude that the developed CNN-LSTM model is accurate and fast in detecting the depression using EEG signals. It can be employed in psychiatry wards of the hospitals to detect the depression using EEG signals accurately and thus aid the psychiatrists.

Electroencephalogram-Based Metastability in Mild Cognitive Impairment Alzheimer's Disease.

Background: In this study, we analyze metastability, a feature of brain dynamics in subjects experiencing mild cognitive impairment Alzheimer's disease (MCI-AD) under eyes open (EO) and eyes closed (EC) conditions. Alzheimer's disease (AD) is a critically prolonged brain disorder that interrupts neural synchronization and desynchronization. Thus, studying metastability under EO and EC conditions would help in understanding the cortical dynamics and its impact in early-stage AD. Methods: Metastability is investigated using three methods namely frequency variance analysis, Kuramoto order parameter, and through meta-state activation patterns. Frequency variance estimated from 21 electroencephalogram (EEG) channels was clustered into three regions namely anterior, central, and posterior to study the regional metastability analysis. Global metastability was assessed from Kuramoto order parameter and meta-state activation patterns by collating the 21 EEG channels. Results: Reduction in metastability was observed in central regions of MCI-AD subjects through the study of frequency variance analysis. There was a marked reduction in global metastability in the patient group under the resting EO condition. Reduction in meta-state activation properties such as temporal activation sequence complexity, modularity, and leap size in MCI-AD condition under the EO condition indicates an overall reduction in brain flexibility. Conclusion: Taken together, the study infers an underlying structural change in neuronal dynamics influencing the reduction of metastability under the MCI-AD condition. The study further revealed that this reduction in metastability is more pronounced in the EO condition.

Functional Connectivity and Complexity in the Phenomenological Model of Mild Cognitive-Impaired Alzheimer's Disease.

Background: Functional connectivity and complexity analysis has been discretely studied to understand intricate brain dynamics. The current study investigates the interplay between functional connectivity and complexity using the Kuramoto mean-field model. Method: Functional connectivity matrices are estimated using the weighted phase lag index and complexity measures through popularly used complexity estimators such as Lempel-Ziv complexity (LZC), Higuchi's fractal dimension (HFD), and fluctuation-based dispersion entropy (FDispEn). Complexity measures are estimated on real and simulated electroencephalogram (EEG) signals of patients with mild cognitive-impaired Alzheimer's disease (MCI-AD) and controls. Complexity measures are further applied to simulated signals generated from lesion-induced connectivity matrix and studied its impact. It is a novel attempt to study the relation between functional connectivity and complexity using a neurocomputational model. Results: Real EEG signals from patients with MCI-AD exhibited reduced functional connectivity and complexity in anterior and central regions. A simulation study has also displayed significantly reduced regional complexity in the patient group with respect to control. A similar reduction in complexity was further evident in simulation studies with lesion-induced control groups compared with non-lesion-induced control groups. Conclusion: Taken together, simulation studies demonstrate a positive influence of reduced connectivity in the model imparting a reduced complexity in the EEG signal. The study revealed the presence of a direct relation between functional connectivity and complexity with reduced connectivity, yielding a decreased EEG complexity.

MVME-RCMFDE framework for discerning hyper-responsivity in Autism Spectrum Disorders.

BACKGROUND: Autism Spectrum Disorder (ASD), characterized by impaired sensory processing, has a wide range of clinical heterogeneity, which handicaps effective therapeutic interventions. Therefore, it is imperative to develop potential mechanisms for delineating clinically meaningful subgroups, so as to provide individualised medical treatment. In this study, an attempt is being made to differentiate the hyper-responsive subgroup from ASD by analysing the complexity pattern of Visual Evoked Potentials (VEPs), recorded from a group of 30 ASD participants, in the presence of vertical achromatic sinewave gratings at varying contrast conditions of low (5%), medium (50%) and high (90%). METHOD: This study proposes a new diagnostic framework incorporating a novel signal decomposition method termed as Modified Variational Mode Extraction (MVME) and a multiscale entropy approach. MVME segments the signal into five constituent modes with less spectral overlap in lower frequencies. Refined Composite Multiscale Fluctuation-based Dispersion entropy (RCMFDE) is extracted from these constituent modes, thereby facilitating the identification of hyper-responsive subgroup in ASD. RESULTS: When tested on both simulated and real VEPs, MVME displays appreciable performance in terms of root mean square error and minimal spectral overlap in the lower frequencies, in comparison with the other state-of-the-art techniques. Relative Complexity analysis with RCMFDE exhibits a rising trend in 43%-50% of ASD in modes 1, 2, 3 and 4. CONCLUSION: The proposed MVME-RCMFDE approach is efficient in discriminating the hyper-responsive subgroup in ASD in multiple modes namely mode 1, 2, 3 and 4, which correspond to delta, theta, alpha and beta frequency bands of brain signals.

Classification of Emotional States Using EEG Signals and Wavelet Packet Transform Features.

In this work, an attempt has been made to classify arousal and valence states of emotion using time-domain features extracted from the Wavelet Packet Transform. For this, Electroencephalogram (EEG) signals from the publicly available DEAP database are considered. EEG signals are first decomposed using wavelet packet decomposition into θ, α, ß, and γ bands. Then featural, namely band energy, sub-band energy ratio, root mean of energy, and information entropy of band energy is estimated. These features are fed into various machine learning classifiers such as support vector machines, linear discriminant analysis, K-nearest neighbor, and random forest. Results indicate that features extracted from wavelet packet transform can predict the arousal and valence emotional states. It is also seen that Support Vector Machines perform the best for both arousal (f-m = 75.68%) and valence(f-m=57.53%). This method can be used for the recognition of emotional states for various clinical purposes in emotion-related psychological disorders like major depressive disorder.

Assessment of emotional states in EEG signals using multi-frequency power spectrum and functional connectivity patterns.

In this work, an attempt has been made to characterize arousal and valence emotional states using Electroencephalogram (EEG) signals and Phase lag index (PLI) based functional connectivity features. For this, EEG signals are considered from a publicly available DEAP database. Signals are decomposed into four frequency bands, namely theta (θ, 4-7 Hz), alpha (a, 8-12 Hz), beta (ß, 13-30 Hz), and gamma (γ, 30-45 Hz). Two features, namely relative PSD and PLI, are calculated from each band of signals with Welch's periodogram. Four classifiers, namely Random Forest (RF), Support Vector Machines (SVM), Linear Discriminant Analysis (LDA), and K-Nearest Neighbor (KNN), are employed to discriminate the emotional states. Results show that the proposed approach can differentiate emotional states using EEG signals. It is observed that there is strong functional connectivity in Fp1-02 and Fp2-Pz in all emotional states for different frequency bands. SVM classifier yields the highest classification performance for arousal, and RF yields the highest performance for valence in the y band. The combination of all features performs the best for the valence dimension. Thus, the proposed approach could be extended for classifying various emotional states in clinical settings. Clinical Relevance- This establishes PLI based approach for improved classification (fl = 74.77% for Arousal fl = 74.94 for valence) of emotional states.

A noise-robust sparse approach to the time-frequency representation of visual evoked potentials.

BACKGROUND: Visual evoked potential (VEP) offers a promising research strategy in the effort to characterise brain disorders. Pertinent signal processing techniques enable the development of potential applications of VEP. A joint time-frequency (TF) representation provides more comprehensive information about the underlying complex structures of these signals than individual time or frequency analysis. However, this representation comes at the expense of low TF resolution, increased data volume, poor energy concentration and increased computational time. Owing to the high non-stationarity and low signal-to-noise ratio of VEP, a TF representation that retains only the pertinent components is indispensable. METHOD: The objective of this study is to investigate and demonstrate the ability of various TF approaches to provide an energy-concentrated and sparse TF representation of VEP. The performance of each method has been assessed for its energy concentration and reconstruction ability on both simulated and real VEPs. Renyi entropy, computation time and correlation coefficient are chosen as the performance measures for the assessment. RESULTS: In comparison with the other state-of-the-art approaches, Synchroextracting transform (SET) exhibits the lowest Renyi entropy and the highest correlation coefficient, thereby ensuring a compact TF representation for the better characterisation of VEP signals. These results are also statistically verified through the Friedman test (p<0.001). CONCLUSION: SET assures a powerful TF framework with improved energy concentration at a faster pace while remaining invertible and preserving vital information.

Emotion Recognition in EEG Signals Using Decision Fusion Based Electrode Selection.

Emotions are essential for the intellectual ability of human beings defined by perception, concentration, and actions. Electroencephalogram (EEG) responses have been studied in different lobes of the brain for emotion recognition. An attempt has been made in this work to identify emotional states using time-domain features, and probabilistic random forest based decision fusion. The EEG signals are collected for this from an online public database. The prefrontal and frontal electrodes, namely Fp1, Fp2, F3, F4, and Fz are considered. Eleven features are extracted from each electrode, and subjected to a probabilistic random forest. The probabilities are employed to Dempster-Shafer's (D-S) based evidence theory for electrode selection using decision fusion. Results demonstrate that the method suggested is capable of classifying emotional states. The decision fusion based electrode selection appears to be most accurate (arousal F-measure = 77.9%) in classifying the emotional states. The combination of Fp2, F3, and F4 electrodes yields higher accuracy for characterizing arousal (65.1%) and valence (57.9%) dimension. Thus, the proposed method can be used to select the critical electrodes for the classification of emotions.

Complex network analysis of MCI-AD EEG signals under cognitive and resting state.

OBJECTIVE: The purpose of this study is to characterize functional connectivity changes in mild cognitive impaired Alzheimer's disease (MCI-AD) under resting and cognitive task conditions. METHOD: EEG signals were recorded under resting states (Eyes closed (EC) and Eyes open (EO)) and cognitive states (Mental Arithmetic Eyes closed (MAEC) and Mental Arithmetic Eyes open (MAEO)) conditions. Functional connectivity metrics were calculated using weighted phase lag index (WPLI). Topological features of the functional connectivity network were analyzed through minimum spanning tree (MST) algorithm. Betweenness centrality was estimated in five different regions of the brain to study hub importance. RESULTS: An increase in values of eccentricity and diameter were observed in patient group in five frequency bands of delta, theta, alpha1, alpha 2 and beta bands under resting and cognitive states. A reduction in leaf fraction was observed in alpha 1 band of EO condition. The results indicated a reduction in functional integration in the brain network of MCI-AD patients. Analysis of MST parameters revealed a higher disintegrated network for the alpha band under EO protocol. The study of hub status in the network displayed an elevated hub status in the central region for the patient group under cognitive task. The study also observed increased integration in gamma band in MCI - AD subjects than healthy controls under cognitive load. CONCLUSION: Disintegration of functional network in alpha band under eyes open protocol and elevated hub strength in central region during cognitive task could be distinguishing features that could aid early detection of AD.

A hybrid method for artifact removal of visual evoked EEG.

BACKGROUND: The visual evoked Electroencephalogram (EEG) signals are useful indicators to explore the hidden neural circuitry in human brain. But these signals are highly contaminated with a plethora of artifacts arising from power interference, eye, muscle and cardiac movements. Since the interference components include neural activity also, the existing techniques result in the distortion of the underlying cerebral signals. NEW METHOD: To address the aforementioned problem, the current study proposes a hybrid method for denoising the visually evoked EEG responses. According to the proposed method, a cascade combination of digital filters, Independent Component Analysis (ICA) and Transient Artifact Reduction Algorithm (TARA) is utilized to suppress the artifacts. ICA technique automatically eliminates the ocular artifacts. The interference due to the remaining artifacts is removed through TARA. RESULTS: The artifact removal ability of the proposed heuristics is evaluated in terms of SNR, correlation coefficient and sample entropy. The ICA results exhibit an increase of 13.47 % in SNR values on simulated signals and 26.66 % on real data. The application of TARA on simulated and real signals results in further SNR gain of 6.98 % and 71.51 % respectively. Significant statistical difference is also observed in this method (p<0.05). COMPARISON WITH EXISTING METHODS: This approach outperforms previous methods based on wavelets, enhanced variants of empirical mode decomposition and earlier versions of total variation denoising. CONCLUSION: ICA-TARA effectively eliminates the major artifacts without compromising the interpretation of the underlying neural state in both simulated and real visual evoked EEG.

Event related potential analysis techniques for autism spectrum disorders: A review.

Autism Spectrum Disorders (ASD) comprise all pervasive neurodevelopmental diseases marked by deficits in social and communication skills, delayed cognitive development, restricted and repetitive behaviors. The core symptoms begin in early childhood, may continue life-long resulting in poor performance in adult stage. Event-related potential (ERP) is basically a time-locked electroencephalogram signal elicited by various stimuli, related to sensory and cognitive processes. The various ERP based techniques used for the study of ASD are considered in this review. ERP based study offers the advantage of being a non-invasive technique to measure the brain activity precisely. The techniques are categorized into three based on the processing domain: time, frequency and time-frequency. Power spectral density, coherence, phase synchrony, multiscale entropy, modified multiscale entropy, sum of signed differences, synchrostates and variance are some of the measures that have been widely used to study the abnormalities in frequency bands and brain connectivity. Various signal processing techniques such as Fast Fourier Transform, Discrete Fourier Transform, Short-Time Fourier Transform, Principal Component Analysis, Wavelet Transform, Directed Transfer Function etc. have been used to analyze the recorded signals so as to unravel the distinctive event-related potential patterns in individuals with ASD. The review concludes that ERP proves to be an efficient tool in detecting the brain abnormalities and connectivity issues, indicating the heterogeneity of ASD. Many advanced techniques are utilized to decipher the underlying neural circuitry so as to aid in therapeutic interventions for improving the core areas of deficits.